We’ve learned through our Customer Zero journey—where we use our technology and services first and communicate our learnings back to our product groups—that responsible AI (RAI) is an essential scaling function. It turns AI experimentation into AI systems that can operate safely across enterprise functions.  

From supply chain operations, to legal workflows, to customer support, and countless other essential business processes, we are learning that scaling AI requires operating discipline:

• Being intentional about the purpose of any AI solution
• Making its decisions observable and auditable
• Empowering teams by ensuring they have the expertise and processes necessary to make responsible AI decisions themselves

Responsible AI as a scaling function

Our Office of Responsible AI creates and implements standardized RAI policies and processes, which speed up reviews and allow for consistency across products. The goal is to build responsible AI into the design and development of agentic AI systems, rather than treat it as something added at the end.

“If you start with responsible AI principles by design, privacy by design, and security by design, the impact in terms of resources is extremely low. And the benefit will be projected across the entire lifecycle of your solution.”

Alberto Carrozzo, senior technical program manager, Microsoft Europe, the Middle East, and Africa (EMEA)

That approach is reflected in the work of Alberto Carrozzo and his team, who are experts at helping apply responsible AI practices across internal AI scenarios.

“If you start with responsible AI principles by design, privacy by design, and security by design, the impact in terms of resources is extremely low,” says Carrozzo, a senior technical program manager and an RAI leader for Microsoft Europe, the Middle East, and Africa (EMEA). “And the benefit will be projected across the entire lifecycle of your solution.”

Our responsible AI standard

Learn about responsible AI at Microsoft, including tools, practices, and policies.

As our teams across Microsoft apply more and more AI tools in their work every day, distinct risk profiles appear. Supply chain agents use business-confidential information and supplier data. Legal systems analyze sensitive records and must make defensible decisions about their relevance. Customer support tools put AI directly into conversations with external customers.

Carrozzo’s team has completed more than 2,000 internal-tool reviews so far, and they’re using their learnings to help prioritize and process a pipeline of thousands more.

Patterns from earlier work can direct reviewers toward cases that warrant closer attention, which gives our team a way to handle that volume consistently and efficiently.

“When you have thousands of agents, you need to use what you learned from the past,” Carrozzo says. “The more we learn, the more we need to use this knowledge to follow the risk and address it as fast as possible.”

Supply chain: Making agent behavior more predictable

Across our Customer Zero experience, several practices have emerged that can help teams with adoption and scaling work. The first may seem counterintuitive: Sometimes scaling AI means intentionally limiting what an AI system can do.

Our cloud supply chain organization—which sources, delivers, maintains, and ultimately decommissions the hardware behind all our company services—had grown its volume of business fivefold in recent years, while staffing remained flat. They needed to innovate.

The organization now has more than 100 enterprise-grade AI agents in production. Analysis that previously took several people days to assemble can now be automatically prepared and updated continuously, giving planners a real-time starting point to begin their work.

“If I build a planning agent, it shouldn’t be able to answer a question about the weather. It needs to be grounded in the fact that it was created to do a certain job.”

Jeevan Pathuri, vice president of software engineering, Cloud Supply Chain, Azure Hardware Systems and Infrastructure

As our teams put more agents into production, testing revealed a practical issue common with AI tools: An application sometimes allows through more of the model’s general capability than its business purpose requires. At enterprise scale, a broader range of capabilities outside the agent’s intended scope can lead to uncertainty with its potential behavior, making decisions less predictable and outcomes harder to anticipate.

Our Cloud Supply Chain group uses OneRAI to register AI capabilities, document models and evaluations, track approvals, and test those boundaries as part of its responsible AI process.

“With RAI, I get the benefit of running the Red Team Agent to make sure a tool is not doing things that are outside of its boundary,” says Jeevan Pathuri, a vice president of software engineering in Cloud Supply Chain, part of Azure Hardware Systems and Infrastructure at Microsoft. “If I build a planning agent, it shouldn’t be able to answer a question about the weather. It needs to be grounded in the fact that it was created to do a certain job.”

Defining what an agent should and shouldn’t do up front reduces ambiguity, minimizes rework, and accelerates deployment by making system behavior more predictable.

A narrow scope also helps employees see what the system was built to handle and where their own judgment should be applied. AI-generated recommendations remain simply inputs for a decision—the person using the agent still owns the outcome.

Legal review: Making outputs verifiable

Litigation and regulatory investigations often require our legal department to review large volumes of data for relevance, privilege, personally identifiable information, security risks, and confidentiality concerns.

After search and machine-learning filters reduce the total volume of material, attorneys from our Corporate, External and Legal Affairs (CELA) department may still need to read tens of thousands of documents individually.

“The AI-generated reasoning needed to be reproducible and explainable. We had to make sure that the insights were verifiable—that it was not a hallucination and was accurately tied to the document itself.”

EJ Bastien, general manager of discovery programs, Microsoft Corporate, External and Legal Affairs (CELA)

Generative AI creates an opportunity to apply case-specific instructions across a large collection of documents, classify and summarize content, and preserve a written reason for each decision. But in highly sensitive legal work, getting an answer isn’t enough. Attorneys need to understand why the AI reached its conclusion and be able to verify it against the underlying evidence.

“The AI-generated reasoning needs to be reproducible and explainable,” says EJ Bastien, a general manager of Discovery Programs in Microsoft CELA. “We had to make sure that the insights were verifiable—that it was not a hallucination and was accurately tied to the document itself. This is so that any time an attorney looks at the output, they have a written explanation for why a given document would be considered relevant or not.”

CELA’s agentic system was built in Azure using Azure OpenAI so that internal documents would not be exposed to a public model. The agent is restricted to a curated document set, rather than searching broadly. Attorneys can inspect the explanation, find the supporting material, and test whether the interpretation holds up before they make a legal decision.

Bastien describes the RAI operating principle as “trust but verify.” Citations help users confirm what a source says and can reveal other repositories that may need to be preserved or collected. A strictly scoped corpus of documents and durable outputs support a review method that legal teams can explain.

Customer support: Keeping responsible AI current as technology evolves

In customer support, the challenges come in real time as our teams handle customer calls from nearly 200 countries in more than 60 languages. Some of these cases can be highly specialized, and only a limited number of support engineers have the expertise to handle them.

“There may be five experts in the world on a particular issue, and they all speak English,” says Bill Fillmore, a senior technical program manager in Customer Service and Support. “But we now have an AI-powered translation platform that enables us to communicate with that customer in their chosen language.”

However, putting AI directly into customer conversations creates responsible AI challenges. The system interacts with customers and customer data while technology and its security, privacy, and governance requirements continue to evolve. Addressing the RAI requirements only once, before deployment, isn’t enough.

“RAI is a priority for us. We make sure that the guardrails are there first, as far as security, privacy, and governance go.”

Josh Wong, senior product manager, Customer Zero Engineering

Ensuring that customer support use cases like front-door problem solving and partially autonomous support are grounded in responsible AI principles has been a game changer for the team.

“RAI is a priority for us,” says Josh Wong, a senior product manager in Customer Zero Engineering. “We make sure that the guardrails are there first, as far as security, privacy, and governance go. Whenever there are updates to those things, we prioritize making sure we incorporate them into our AI solutions. That’s non-negotiable at Microsoft.”

That practice gives the team a way to keep the solution aligned with our requirements as both the technology and its use case evolve. Rather than treating responsible AI as a one-time approval, we continually incorporate updated requirements into the AI solution.

Looking ahead to the Frontier future

With the lightning pace of change that comes with agentic development, we’re not standing still at Microsoft when it comes to responsible AI work.

Pathuri’s Cloud Supply Chain group is progressing from individual agents to end-to-end supply chain workflows driven by AI. Bastien sees Microsoft’s legal teams prototyping more case-specific AI tools while preserving the same obligations around evidence and human judgment. Wong’s and Fillmore’s teams are exploring AI for case routing, front-door problem solving, and autonomous support functions.

“These guardrails solve the challenge of scaling AI safely without slowing teams down. Our goal is not to block innovation, but to make safe AI adoption the default path.”

Alberto Carrozzo, senior technical program manager, Microsoft Europe, the Middle East, and Africa (EMEA)

As AI takes on a larger role in the enterprise, the lessons from our Customer Zero experience become more important. Our teams cannot eliminate every risk when developing AI-powered solutions, but they can establish evidence, ownership, and feedback mechanisms that reinforce responsible AI principles to support stronger data governance without losing sight of the larger goals.

This is where responsible AI becomes a scaling function.

“Things like mandatory risk assessments before deployment, permission-based data access, continuous telemetry and monitoring, and automated detection of security and responsible AI risks are key to effective AI deployment,” Carrozzo says. “These guardrails address the challenge of scaling AI safely without slowing teams down. Our goal is not to block innovation, but to make safe AI adoption the default path.”

The post Scaling AI responsibly at Microsoft: Lessons from our Customer Zero journey appeared first on Inside Track.

Here at Microsoft, data powers our reporting, supports our decision-making, and helps our leaders understand what’s happening across our teams responsible for our enterprise data.

But agentic AI has raised the stakes, making data quality more consequential. Frontier Firms like ours depend on high-quality data to support agents, automation, and employee decision making.

Today’s AI models can produce sophisticated summaries, answer complex questions, and help employees work more efficiently. Agents and reasoning systems increasingly connect data across business functions, perform multistep tasks, and help automate decisions and workflows.

That shift has led our teams in Microsoft Digital, the company’s IT organization, to ask a new question: What does it mean to have data that’s fully AI-ready, and where are we on this journey?

Measuring data quality is just one part of the answer.

“World-class AI agents start with AI-ready data, with the right foundation. That means more than simply making data available. It requires quality, governance, and consistent standards so that AI systems can understand the data, trust it, and use it effectively.”

Naval Tripathi, principal engineering manager, Microsoft Digital

In Microsoft Digital, we’re creating a repeatable framework for AI-ready data that combines governance, discoverability, semantics, context, ownership, lineage, security, and quality into an overall foundation. This framework helps AI systems interpret enterprise data and use it appropriately.

“World-class AI agents start with AI-ready data, with the right foundation,” says Naval Tripathi, principal engineering manager in Microsoft Digital. “That means more than simply making data available. It requires quality, governance, and consistent standards so that AI systems can understand the data, trust it, and use it effectively.”

Treating AI readiness as a business capability                         

One of the key lessons from our focus on AI-ready data here in Microsoft Digital is that readiness can’t be treated as a one-time cleanup project.

“We decided to focus on continuous data readiness, rather than a one-time certification process. You really can’t say, ‘OK, it’s all done now.’ We have to keep learning about and updating metadata and governance information to improve how AI uses data.”

Aishwarya Dinde, senior software engineer, Microsoft Digital

Data changes. Systems evolve. New business requirements emerge. AI capabilities continue to advance. So, we are approaching readiness as an ongoing business capability.

“We decided to focus on continuous data readiness, rather than a one-time certification process,” says Aishwarya Dinde, a senior software engineer in Microsoft Digital. “You really can’t say, ‘OK, it’s all done now.’ We have to keep learning about and updating metadata and governance information to improve how AI uses data.”

These data-readiness signals help determine whether the data is discoverable, reliable, understandable, and appropriate for AI-driven scenarios. Rather than just looking at a data-quality score, we’re creating a systematic way to identify where investments in data will likely have the greatest impact.

AI changed the standards for quality data

Many data platforms were originally designed to support analytics and reporting, which are human-centered activities. These platforms were designed primarily for human analysis, not for AI agents that need explicit business context.

“In the past, data was something humans used to get information and provide business insights. With AI, we are expecting the models to reason over the data. There is a need to ensure the data is ready for that and represents the business correctly.”

Miguel Uribe, principal PM manager, Microsoft Digital

In that era, humans supplied the context around these applications of data. Business analysts understood company terminology. Data engineers knew where information came from. Subject matter experts filled in gaps that weren’t captured inside the data itself.

AI systems have changed all this.

“In the past, data was something humans used to get information and provide business insights,” says Miguel Uribe, a principal PM manager in Microsoft Digital. “With AI, we are expecting the models to reason over the data. There is a need to ensure the data is ready for that and represents the business correctly.”

AI systems need explicit guidance about what data represents and how it should be used. Column names, metadata, ownership information, business definitions, descriptions, lineage records, and usage guidance all help AI understand what the data means and how it should be used.

In one example from our networking environment, an internal system used “Pigeon,” an internal name for a networking system, in its title. Without that context, an AI system interpreted “Pigeon” literally and incorrectly inferred that the data set was about birds rather than network infrastructure.

This is where human effort is needed to make sure that context is present and accurate.

“There’s a lot of retrofitting that needs to be done—human effort to contextualize the data,” says Diego Baccino, principal software engineering manager in Microsoft Digital. “AI doesn’t understand the business context. That’s where the human capital becomes really valuable. Once the context is added, then AI can really take it forward.”

This doesn’t mean that AI doesn’t play a role in the data-readiness process. AI-driven tools are increasingly used to detect data problems, and people decide what to do about them. We regularly run automated checks against our readiness criteria, flagging data that is stale, incomplete, or inconsistent with its documentation..

AI can also draft data descriptions for engineers to review and correct. This approach helps teams scale readiness work while keeping people accountable for accuracy and business context.

This realization about the importance of context led the Microsoft Digital data council and AI Center of Excellence—two vital aspects of our Frontier Firm move into agentic AI—to expand their focus beyond traditional governance and data-quality efforts. The goal was to create a common understanding of what AI-ready data means for the different teams that are responsible for our data across the enterprise.

“Metadata and context are part of the product at this point,” says Aira Agrawal, a product manager in Microsoft Digital. “Just having the data available isn’t enough anymore. Without that context, it’s not usable in the age of AI.”

As a result, our teams in Microsoft Digital are increasingly working to make data context discoverable and reusable, reducing the dependence on tribal knowledge that previously existed only in conversations and scattered documentation.

Measuring readiness before scaling AI

As groups across Microsoft have increased their investments in AI, our leaders are now asking questions about data readiness. Our mission in Microsoft Digital is to come up with accurate ways to make this readiness measurable.

“Our progression towards AI-ready data is going to be one of those key metrics that get looked at month over month. It’s becoming a super-critical metric for the organization, to ensure that all of our solutions are leveraging AI-ready data.”

Myron Wan, principal group product manager, Microsoft Digital

We now evaluate our data across multiple dimensions, including quality, governance, discoverability, interoperability, accessibility, observability, and contextual understanding. These dimensions help teams understand where gaps exist and where remediation efforts should be prioritized.

Historically, data quality was more of a technical concern discussed by engineering teams. Microsoft Digital is elevating data readiness to a business conversation tied directly to AI outcomes.

“Our progression towards AI-ready data is going to be one of those key metrics that get looked at month over month,” says Myron Wan, principal group product manager in Microsoft Digital. “It’s becoming a super-critical metric for the organization, to ensure that all of our solutions are leveraging AI-ready data.”

If a particular data set isn’t ready for its intended AI use, leaders can weigh the business value against the investment needed to address the gaps. “If your data score is telling you it’s not ready, then you can pose the question: Should we fast-track this and invest more in making the data AI-ready?” Uribe says. “That enables us as business decision-makers to prioritize where we’re going to put our money and our AI investments.” 

Using AI to breakdown business silos

Across Microsoft, teams are exploring scenarios where AI systems help identify problems, gather information, and support operational decision-making. Those scenarios depend heavily on trusted data foundations.

“We’re moving into a world where our data sets are not going to be in silos anymore, especially as all our organizational data moves into Microsoft Fabric. Networking, HR, finance, they’re all going to have their data in Fabric. This means we can tell stories across the business with connectors and APIs, unlocking more advanced analytics.”

Aira Agrawal, product manager, Microsoft Digital

Future AI systems may increasingly operate across domains, drawing information from disparate groups such as IT, HR, finance, operations, and other areas. That requires data that’s accessible and that also meets this new, higher bar for readiness.

We’re already seeing opportunities to use AI to connect information across previously separate domains, creating richer business insights and supporting more advanced forms of analysis.

“We’re moving into a world where our data sets are not going to be in silos anymore, especially as all our organizational data moves into Microsoft Fabric,” says Aira Agrawal, a product manager on the Infrastructure Data Lake team. “Networking, HR, finance, they’re all going to have their data in Fabric. This means we can tell stories across the business with connectors and APIs, unlocking more advanced analytics. It’s a really cool thing we’re enabling.”

Our long-term goal is to give business teams trusted data and shared foundations so they can create AI-enabled workflows without depending on central teams for every solution. Instead, we want individual business teams to be able to create their own AI-enabled workflows that operate across domains, knowing that the data they’re relying on is AI-ready.

The future of AI-ready data

Our work with AI-ready data continues to evolve.

The company is still refining readiness models, governance approaches, access patterns, and future capabilities. Our subject matter experts acknowledge that some challenges remain unresolved, including how best to provide business context at scale and how readiness should continue evolving alongside AI technologies and regulatory requirements.

“Whatever technologies you’re using, don’t miss the boat on contextualizing your data and making it AI-ready. The longer you wait, the longer it takes to realize AI value.”

Diego Baccino, principal software engineering manager, Microsoft Digital

Yet several key lessons have already emerged:

• Visibility matters. Organizations need to know what data they have and who owns it.
• Governance remains critical, but governance alone isn’t enough.
• Semantics and business context increasingly determine whether AI can use data effectively.
• Readiness must become repeatable across teams rather than dependent on individual experts.

And perhaps one of the most important lessons: Don’t wait to get started. Data readiness is too crucial a factor for business success in the age of AI to put off dealing with the challenge.

“Whatever technologies you’re using, don’t miss the boat on contextualizing your data and making it AI-ready,” Baccino says. “The longer you wait, the longer it takes to realize AI value.”

The post AI-ready data: Building the foundation for enterprise AI at Microsoft appeared first on Inside Track.

And those were definitely important—but we were leaving out a huge factor: trust.

This finding emerged from a series of user-research studies our team in Microsoft Digital—the company’s IT organization—conducted internally after early adoption rates of the agent were not as strong as we’d hoped. The studies revealed that our employees’ reactions to agentic support were categorically different than other software rollouts.

“We needed to change the agentic support scenarios to create trust and empathy—a more humanized approach. That was a big paradigm shift for us.”

Olivier Cherel, senior business process manager, Microsoft Digital

People were making decisions about using the Employee Self-Service Agent based on more human factors, including emotion, expectations, habits, and previous experiences with support. Those quintessentially human factors played a significant role in whether the employee opted for agentic support or chose a different service modality, even when the technology itself was clearly capable of meeting their needs.

Our research highlighted this crucial distinction: technical quality mattered, but it wasn’t enough. It was about the employee’s total support journey.

“We realized after doing the research that ESS was not completely ready for prime time yet,” says Olivier Cherel, a senior business process manager in Microsoft Digital’s employee support organization. “We needed to change the agentic support scenarios to create trust and empathy—a more humanized approach. That was a big paradigm shift for us. And I think it’s going to be the same for any organization that wants to implement an IT help-desk agent.”

This process ultimately gave us a practical framework for designing agentic support experiences that employees choose to use first, and return to repeatedly. These are lessons that our customers can apply as they introduce AI support agents of their own.

Read about how we developed and deployed the Employee Self-Service Agent at enterprise scale across Microsoft.

Agentic employee support: A social experience

Our research began with a broad question: What was limiting employee adoption and acceptance of the Employee Self-Service Agent?

We needed to explore how our employees interacted with different modes of support, what shaped their expectations, and where friction could be found in the process.

“We had the chance to work with researchers and designers at Designit, a vendor that conducted qualitative research, UX design work, and adoption analysis,” Cherel says. “The findings helped us better understand how employees experienced support, and also what would make an agent-driven experience more useful.”

Experts at Designit—a global experience innovation company that is the strategic design arm of Wipro—conducted interviews and focus groups, analyzed Microsoft employee feedback, examined support workflows, and studied interactions between employees and the agent across multiple geographic regions. The findings pointed toward factors that extended well beyond product functionality.

“We started realizing there was a lot more under the hood than we expected,” says Chaitrasri Rao, executive design director at Designit. “Trust was a major issue. People approached an AI agent very differently than the way they approached traditional software.”

“With traditional support chatbots, it was like a search engine—you plugged in your problem and it spat out a solution. With ESS, it’s really more about how to engage in conversation with these end users so that they get as much as they can out of it.”

Kevin Verdeck, senior IT service manager, Microsoft Digital

When a user interacts with a traditional application—such as a spreadsheet, a dashboard, or a ticketing system—they have a clear mental model: They are operating a machine. They expect predictable, deterministic outputs.

When that same user interacts with an AI agent, something shifts. They don’t operate it; they talk to the agent. They ask questions. They negotiate. They even say “please” and “thank you.” Users treat AI agents less like tools and more like interlocutors—somewhere between a colleague and a customer service representative.

This was a big leap for the team developing the Employee Self-Service Agent. They realized that user trust was not based on technical reliability. It has more to do with social alignment.

“With traditional support chatbots, it was like a search engine—you plugged in your problem and it spat out a solution,” says Kevin Verdeck, senior IT service manager, Microsoft Digital. “With ESS, it’s really more about how to engage in conversation with these end users so that they get as much as they can out of it. The user needed to be aware that if ESS didn’t understand them the first time, they can reword the issue and provide more context in the conversational interface.”

What’s more, our employees were seeking indications that the system understood—and even empathized with—their situation.

“We realized that ESS was too much of a machine,” Cherel says. “It was not human enough. So that’s something that we really prioritized and addressed.”

This insight reframed the work. The team shifted from asking how to improve a support tool to asking how to improve the overall support experience.

Understanding context in the support experience

The Designit research uncovered a second finding that proved equally important: the overall support context matters.

“We had to build trust, that ESS is not just a dead end for the user. It can help them self-solve their issue, but it also needs to tell them, ‘Hey, if I can’t fix it, I’ll hand you over to a live agent.’”

David Finney, director of IT Service Management, Microsoft Digital

Most employees didn’t make the Employee Self-Service Agent their first stop when they needed help. Instead, they typically tried a number of other resources before opening the agent. They’d consult coworkers, search through documentation, explore existing tools, and/or try their own troubleshooting steps.

As a result, people were sometimes frustrated when the agent made suggestions for possible fixes that they had already tried. We needed to be sensitive to what the employee was experiencing before using the Employee Self-Service Agent and demonstrate its tangible value to connect user context to known outages or problems, provide relevant self-help, and enable seamless escalation to a live agent when users need human support.

“We had to build trust, that ESS is not just a dead end for the user,” says David Finney, director of IT Service Management in Microsoft Digital. “It can help them self-solve their issue, but it also needs to tell them, ‘Hey, if I can’t fix it, I’ll hand you over to a live agent.’ This helps create trust, so the employee will turn to ESS as the first point of entry—because they know that it won’t leave them hanging.”

This broader perspective influenced the design recommendations that followed. Researchers concluded that employees evaluated support experiences across the entire process rather than judging each interaction in isolation.

When context was lost, employees noticed. When systems forced them to retrace their steps, frustration increased. When support channels felt connected, confidence improved.

“Maybe the biggest learning was that we needed to earn users’ trust before expecting them to consistently turn to self-service tools,” says Cherel. “It’s fine if they initially used ESS mainly as a trusted path to other support options. But if we could offer relevant self-help consistently, and help them escalate quickly when it’s not useful, users will increasingly choose self-service because it works—not because we force them into it.”

Turning research into design decisions

The next challenge involved translating these research insights into practical improvements in the Employee Self-Service Agent.

In addition to the above insights, the team identified specific situations that repeatedly caused employees to lose confidence in the agent. Some employees felt trapped in troubleshooting loops. Others encountered responses that seemed disconnected from what they had already tried. Some were uncertain when escalation was available or what would happen next.

The resulting work focused on helping the Employee Self-Service Agent acknowledge previous troubleshooting efforts, recognize employee context, communicate limitations clearly, and explain next steps. Researchers also documented a series of trust-breakdown scenarios, and then paired each scenario with specific behaviors the agent should demonstrate.

“The first thing we asked ourselves was, ‘How can we make the Employee Self-Service Agent more trustworthy?'” says Monserrat Oliver Sanchez, a senior design researcher with Designit who contributed to the research. “So we defined a personality for ESS that incorporated the qualities that should characterize the agent, and we created guidelines for how it should behave in different situations.”

The emphasis remained on practical employee needs. We examined how the agent should respond when it lacked confidence in an answer, when a user had already completed troubleshooting steps, and when escalation would provide a better outcome than continued self-service.

Ultimately, the research findings were turned into design guidance that could be tested, refined, and incorporated into future improvements.

The results we’ve gotten so far from the Employee Self-Service Agent are significant: a 30% overall reduction in IT support tickets being created and a 50% rate of employees starting their support journey with the agent (up from 27% before we began this trust-building initiative). We continue to pursue our goal of having every employee support need begin with a digital self-service experience.

Putting the user first

As our journey with the Employee Self-Service Agent progressed, the team arrived at a broader view of adoption.

Employees expected agentic support tools to work. They assumed that accurate information and reliable performance were baseline requirements. But embracing an AI-driven support agent depended on additional considerations that went beyond technical competence.

“It’s not about speed or performance; it’s more about continuity. It’s about ensuring that the user can do their work, that they can overcome the problems and challenges they have during their day seamlessly. That’s when we start building AI support agents that people will choose to rely on.”

Alex Calero, lead service designer, Designit

Employees wanted relevance. They asked for systems that recognized where they were in the support process and responded accordingly. And they looked for interactions that respected their previous efforts and connected naturally to the next step.

In other words, they wanted AI to treat them like a human being that needed understanding and help. It’s all part of designing software tools that put the user first.

“At the end of the day, the user should always be at the center of the equation,” says Alex Calero, a lead service designer with Designit. “The tool is there to support them. It’s not about speed or performance; it’s more about continuity and relevance. When a user encounters an IT, HR, or facilities issue, they shouldn’t spend time deciding where to go. They should know that they can simply ask ESS and it will take it from there. That is the ultimate measure of a human-centered AI adoption strategy.”

The post Building trust in AI-powered support at Microsoft: Our Employee Self-Service agent story   appeared first on Inside Track.

Opportunity records need updates. Forecast comments require review. Investment programs and licensing options live in different places. Preparing for a meeting can mean searching across customer relationship management records, email, and meeting notes.

To tackle this challenge, we developed Sales Agent in Microsoft 365 Copilot. This AI-driven platform gives our sellers one conversational entry point, allowing them to pull all the data and context they needed for a deal into one space.

But information retrieval was just the first step. As agentic AI has gotten more powerful, we realized it had the potential to fundamentally change how our sellers completed their day-to-day work. That’s the promise of becoming an AI Frontier Firm.

We also knew that just creating an amazing new AI sales platform for our employees wasn’t enough. Sales Agent needed to connect with how our employees actually worked, or it would never change long-standing work habits. Rethinking work is difficult, and AI hesitancy is real.

“If you have a broken process, just laying AI on top of it is not going to fix it. We had to fundamentally redesign the process to drive the outcomes we wanted to get to. And that’s where Sales Agent comes in.”

Rene Vejlby, senior business program manager, Business Innovation and AI Transformation, Small Medium Enterprises and Channel

One of our biggest Customer Zero takeaways throughout this process is that connecting AI to the work that matters for each role makes a huge difference in adoption rates. If you show your employees exactly how the technology will benefit them in their work, they will navigate the transition to a new way of working and become evangelists for AI in the workplace.

Rethinking how sales work is done

Our internal research has shown that about 70 percent of a seller’s time was being spent on non-customer-facing activities—researching accounts, preparing for meetings, following up afterward, coordinating across the team, and sending proposals and quotes.

The reason for this could be traced back to sheer information sprawl. Our sellers were navigating at least 15 different CRM, tracking, and communication tools on a daily basis, with the data they needed scattered across all of them.

This was a perfect challenge for AI to address. But without changing the way our employees worked, just adding another piece of technology—AI or not—was not the answer, our experts realized.

“If you have a broken process, just laying AI on top of it is not going to fix it,” says Rene Vejlby, a senior business program manager in the Small Medium Enterprises and Channel (SME&C) group. “We had to fundamentally redesign the process to drive the outcomes we wanted to get to.”

That meant setting a clear destination first: a hub where a seller could pull in contextual details and complete common tasks without ever leaving the interface.

“The goal has always been to keep our sales force in their flow of productivity while also keeping our systems of record up to date. Sales Agent gives sellers a single place to work without hopping from system to system.”

Steve Cohen, senior director of digital specialists, SME&C

Getting there was painstaking. Vejlby and his peers took every single workflow that our sellers used and broke it down step by step, asking of each one what AI could take over, what could be eliminated entirely, and what sellers could be doing with that recovered time instead.

“The goal has always been to keep our salesforce in its flow of productivity while also keeping our systems of record up to date,” says Steve Cohen, a senior director of digital specialists in SME&C. “Sales Agent gives sellers a single place to work without hopping from system to system.”

The Sales Agent user experience builds on previous platforms we’ve developed, like Viva Sales and Microsoft 365 Copilot for Sales. Crucially, it adds a chat-based “front door” that can retrieve information, interpret context, and support seller actions through natural language.

But delivering on that ambition meant rebuilding what Sales Agent itself was designed to do.

From information retrieval to workflow completion

When an earlier incarnation of Sales Agent came out a bit more than two years ago, the focus was information retrieval—reflecting the abilities of Copilot at the time. Agentic AI use cases were just starting out, and it was considered a breakthrough simply to get a seller’s CRM data and our wider data estate into one chat experience.

That was no small thing, given the many different systems and hundreds of reports our sellers were otherwise navigating to piece the same picture together.

But eventually, information retrieval became table stakes. Microsoft Cowork and other Copilot agents could do the same work—and do it well, deeply integrated across Microsoft 365—which meant Sales Agent needed a different reason to exist.

“Our charter kind of evolved from focusing on pure information retrieval to become more about workflow and action completion. Today, Sales Agent’s unique value is as an entry point for sellers to complete end-to-end workflows that are aligned to their jobs to be done.”

Ajay Nair, principal group engineering manager, Commercial Engineering and AI

It found one in doing tasks rather than fetching information. Today, Sales Agent works across up to 25 different systems of record, making updates and running workflows, so a seller can create a CRM opportunity, submit an investment request, or build a deal book without ever leaving the conversation.

This was a big transition, according to Ajay Nair. He’s the engineering owner for Sales Agent at Microsoft, and his team is in charge of internal implementation of the product for our own sellers.

“Our charter kind of evolved from focusing on pure information retrieval to become more about workflow and action completion,” Nair says. “Today, Sales Agent’s unique value is as an entry point for sellers to complete end-to-end workflows that are aligned to their jobs to be done.”

Delivering that functionality meant tapping the capabilities of Copilot Studio. Nair’s team used technologies like Model Context Protocol (MCP) to make Sales Agent a single “front door” for agentic services that other Microsoft teams had already built. The routing is invisible: Sellers type a prompt, and Sales Agent works out which agent to invoke.

AI adoption is a mindset shift

When it comes to change management in the AI space, we couldn’t count on traditional approaches that we might have used in the past when rolling out a new tool.

Our sellers may already feel they’re being successful using their existing methods, so they ask, “Why should I change?” There is also a learning curve and a real insecurity about the larger impact of AI that must be addressed.

“Sellers need to think about what outcome they want from AI, and then they have to craft their prompt around that. We have to get people used to that shift in mindset.”

Susan Neece Robien, senior director, AI Transformation, Adoption, and Change

It’s not a simple case of, Look what this new platform can do. Instead, as adoption specialist Susan Neece Robien explains, it’s more of a shift in how our employees think about their work.

“As we like to say, you have to ‘think differently, feel differently, and do differently,’” says Neece Robien, a senior director of AI transformation in Microsoft Sales. “Sellers need to think about what outcome they want from AI, and then they have to craft their prompt around that. We have to get people used to that shift in mindset.”

Because of this, much of our work around Sales Agent adoption started with the people, not the technology.

“I talked to one hard-nosed seller about his experience with Sales Agent, and he said that because of it he was no longer having to work on presentations at two in the morning. This made him a better husband and father, a better person.”

Jacqueline Stein, senior director, Learning and Skilling, Worldwide Learning

An important thing to remember is that people typically do their best learning when they are motivated to do so, and that means understanding how they will benefit from the new knowledge. Once they grasp that, it unlocks the willingness to change and use AI in their everyday work.

“It’s about the human motivation,” says Jacqueline Stein, a senior director of learning and skilling for Microsoft Worldwide Learning. “I talked to one hard-nosed seller about his experience with Sales Agent, and he said that because of it he was no longer having to work on presentations at two in the morning. This made him a better husband and father, a better person. And he can relay that to his customers as his lived experience.”

Role-based adoption

Broad capability descriptions and generic prompt training could create awareness of Sales Agent, but they did not always answer the question a seller would ask first: “How will this help with my job?”

So the team created specific role pages and validated them with people doing the work. Neece Robien described developing content for about 20 field roles, a task that included checking sales guidance with account executives. These scenarios had to represent work that happened often enough, with a benefit large enough, to earn attention.

“What matters is the flow of work and the tools that support it,” Neece Robien says. “Leading with the workflow makes the change more meaningful.”

Our Frontier Accelerator adoption program is designed to use peer learning and role-based practices to cultivate enthusiasm for AI applications among our sales professionals. The benefits include these five outcomes:

Time with customers

Reclaim time for customer interactions by reducing admin and prep tasks

Stronger teams

Collaborate more effectively across internal teams, with peers teaching peers

More energy for sellers

Bring greater energy and value to your customer engagements

Increased employee engagement

Boost engagement through visible daily AI practice

Frontier mindset

Cultivate an approach where AI is the default catalyst for innovation and growth

The teams also moved beyond one-way demonstrations. In peer huddles, sellers now discuss how they approach account planning or another job responsibility, then examine where Sales Agent can help. Our champions add credibility, because they can test realistic scenarios with customer and opportunity data that our adoption professionals may not have.

Regional adoption leaders can also provide local context. They address familiar habits, explain why the change matters, and create space for sellers to learn from one another.

Stein, a regional adoption leader based in the United Kingdom, described how teams are moving toward more dynamic learning and experimentation in the flow of work.

“It’s not just adoption programs aimed at sellers,” Stein says. “It’s sellers learning together, and us creating a framework and a blueprint for them to do that.”

An evolution in AI measurement

Sales Agent is also a great example of our changing approach to AI measurement strategy.

Our early reporting focused on monthly, weekly, and daily use of Sales Agent. Those measures showed whether people opened the product, but they didn’t reveal whether a deal moved faster or a seller spent less time on customer updates as a result.

“The fact that someone uses the agent does not necessarily mean that they’re driving business impact. Now that we’ve got a decent level of adoption, we’re pivoting toward looking at the business outcomes that are actually being driven through the product, rather than mere usage.”

Ajay Nair, principal group engineering manager, Commercial Engineering and AI

Today, we’re shifting our measurement toward actual business outcomes, including deal velocity and reduced seller toil. That change requires teams to identify the process they aim to improve before agent deployment, then decide what evidence will indicate a meaningful difference has been made.

“The fact that someone uses the agent does not necessarily mean that they’re driving business impact,” Nair says. “Now that we’ve got a decent level of adoption, we’re pivoting toward looking at the business outcomes that are actually being driven through the product, rather than mere usage.”

Our larger experience points to a significant Customer Zero insight: A capable agent and a thoughtful adoption plan are parts of the same implementation. Great engineering of the agent gives sellers a useful path through their workflows. Role-based scenarios, peer learning, and results-driven measurement help ensure stickiness and change the way our sellers actually carry out their day-to-day work. And the result is better business outcomes.

The post Driving seller adoption of Sales Agent through role-based change management at Microsoft appeared first on Inside Track.

It worked, up to a point. Individual developers got faster. But the gains stopped there.

“We quickly identified that improving the individual productivity of a developer was not resulting in a boost to team productivity. That was our hard lesson.”

Sudhakar Sadasivuni, principal group engineering manager, Microsoft Digital

The problem wasn’t the tools, but rather the process around them. The traditional software development lifecycle was built for human handoffs, and every handoff introduces a gap in intent.

“We quickly identified that improving the individual productivity of a developer was not resulting in a boost to team productivity,” says Sudhakar Sadasivuni, a principal group engineering manager in Microsoft Digital, who leads our AI-native engineering initiative. “That was our hard lesson.”

Eventually, our leadership posed the question: What would the lifecycle look like if it were designed from the start as a collaboration between humans and AI agents?

Exploring that concept led us to spec-driven development (SDD), an AI-native engineering approach that treats specifications as the primary artifact in the software development lifecycle. Instead of relying on code, documents, or prompts as the main source of truth, our teams now anchor development around a living specification, which captures the business intent and acceptance criteria before implementation begins.

Our engineers can now generate code faster, accelerate testing, and move from idea to implementation more quickly than ever before, thanks to the power and speed of AI.  

This is a step towards Frontier Firm ways of working, where humans provide oversight and direction and AI tools handle task execution. And because we serve as the company’s Customer Zero, the solution is a model that our customers can learn from and use in their own engineering organizations.

Shifting left

By shifting critical thinking and collaboration to the requirements phase of development, SDD resolves ambiguities, surfaces edge cases, and gets everyone on board early in the process, dramatically reducing late-stage chaos.

“Speed without direction is just expensive chaos—and we learned that the hard way through vibe coding. After burning our hands on those traps, we realized that SDD gives you both velocity and vector.”

Mridul Verma, senior software engineer, Microsoft Digital

The result is a development process that helps teams align stakeholders earlier, reduce rework, and make more effective use of AI coding agents while maintaining a clear connection to business goals.

Another advantage of SDD, according to Mridul Verma, a senior software engineer in Microsoft Digital, is that it focuses not only on velocity, but also on vector. (Velocity is how fast your development moves, while vector reflects the direction you’re moving.)

“Speed without direction is just expensive chaos—and we learned that the hard way through vibe coding,” Verma says. “After burning our hands on those traps, we realized that SDD gives you both velocity and vector.”

Why specs matter more in the AI era

Traditional software development often treats specifications as planning documents. After development begins, these documents can quickly become outdated as requirements evolve and implementation details change.

“In the AI era, the best dev teams aren’t the ones that generate the most code. It’s about how they’re best able to preserve intent.”

Vignesh Vijayaraghavan, senior software engineer, Microsoft Digital

SDD flips that model.

The specification becomes a living artifact that evolves alongside the product, serving as a shared reference point for product managers, software engineers, architects, designers, testers, and AI coding agents.

That shift is especially important in AI-native development environments. While AI tools can generate large amounts of code quickly, they depend on clear instructions and guardrails. Teams that can’t consistently communicate intent typically discover that increased coding velocity simply produces more variability and additional rework.

“In the AI era, the best dev teams aren’t the ones that generate the most code,” says Vignesh Vijayaraghavan, a senior software engineer in Microsoft Digital. “It’s about how they’re best able to preserve intent.”

There are five pillars that drive the transformational value of spec-driven development, based on our experience at Microsoft:

Spec as the source of truth

The spec becomes the single authoritative reference for all stakeholders.

Living, executable artifacts

Specs evolve alongside the project, remaining synchronized with the code and tests. They are co-authored by various stakeholders.

AI-assisted automation

SDD leverages AI coding agents to generate, test, and validate code, accelerating development and enhancing productivity.

Human validation and collaboration

Automation is central to SDD, human expertise is crucial for reviewing specs, refining requirements, and validating outcomes.

Predictability and measurability

SDD enables teams to track progress, measure outcomes, and improve predictability.

How the SDD workflow works

At Microsoft Digital, our developers and PMs begin by creating specifications that capture business goals, user requirements, edge cases, and acceptance tests. These specifications are version-controlled and continuously updated as the project evolves. Rather than serving as static documentation, they guide every major decision throughout the lifecycle.

“The spec becomes the shared source of truth for the entire team. Instead of handing requirements from one group to another, everyone collaborates around a living document that guides development and keeps decisions aligned.”

Garima Gaurav, senior product manager, Microsoft Digital

It’s important to note that before creating the specs, the team agrees on a constitution: A shared set of architectural principles, governance requirements, security standards, and development limitations that everyone adheres to. Developing this constitution is a team-wide exercise rather than an individual one, allowing employees and AI agents to work from the same facts and the same context.

“The spec becomes the shared source of truth for the entire team,” says Garima Gaurav, a senior product manager in Microsoft Digital. “Instead of handing requirements from one group to another, everyone collaborates around a living document that guides development and keeps decisions aligned.”

Our teams use GitHub Spec Kit to support the process. This workflow includes six major stages:

1. Define the business problem and expected outcomes
2. Clarify ambiguities and unanswered questions
3. Create a technical implementation plan
4. Break work into traceable tasks
5. Validate alignment between requirements and plans
6. Implement and test the solution

“As teams mature, they learn the right level of granularity, and breaking large requirements into multiple specs becomes a key practice for maintaining quality and preserving intent.”

Ajeya Kumar, principal software engineer, Microsoft Digital

Because each stage builds on the spec, teams can trace implementation decisions back to their related business requirements more easily than with traditional code-first approaches. AI coding agents enhance the spec and then generate code, tests, documentation, and other tasks from it. (Specs are version-controlled in the same repository to aid tracking and avoid confusion.) Engineers focus on validating intent, reviewing outputs, and refining requirements.

“Effective SDD adoption relies on well-scoped specifications,” says Ajeya Kumar, a principal software engineer in Microsoft Digital. “Smaller, focused specs generate outputs that are easier to review and validate, enabling teams to build incrementally rather than all at once. As teams mature, they learn the right level of granularity, and breaking large requirements into multiple specs becomes a key practice for maintaining quality and preserving intent.”

How SDD impacts different roles

Adopting spec-driven development across our organization involved more than just introducing new tools. Existing roles were tightly coupled to the old software development lifecycle, so some had to change while other new roles and skills were created. Microsoft Digital teams documented these changes at every step, because the shift in mindset is what makes SDD work.

Overall, we found that the biggest change was behavioral: Specifications stopped being documents that simply supported development and actually became the artifact that drove it.

Here’s how Microsoft’s shift toward an SDD-driven approach has impacted various roles across the company:

Leadership

For engineering leaders, success is no longer measured solely by development activity. Leaders need to create an environment that rewards clarity, alignment, and collaboration before actual implementation begins. Dev teams are encouraged to invest time in resolving ambiguities and documenting expectations early in the software development lifecycle because this reduces costly rework later.

Leaders also play an important role in driving adoption. Our dev teams say that SDD delivered the greatest value when it became a standard way of working rather than an optional process that individuals can bypass. The adoption has to be deliberate.

“It’s not something where you decide, today I woke up and I’ll start using SDD,” says Apoorv Gupta, a principal software engineer in Microsoft Digital. “It doesn’t work like that. It’s a mindset shift and a learning curve.”

Developers

For developers, the transition to SDD often requires breaking long-standing work patterns. For years, software engineers were encouraged to move quickly from requirements to implementation. Under SDD, their first step is developing a complete understanding of the problem, including defining success, before any code is created.

“As developers, we get excited and want to jump directly into coding,” Verma says. “Changing that habit was really tricky.”

That change runs deeper than just a habit. The developers must make sure the agent fully understands the task before starting the coding process.

“SDD makes you clarify and refine your thoughts first for agents, not for yourself,” Sadasivuni says. “This is agent-first solutioning, not human-first solutioning.”

Our Microsoft Digital engineers now spend more time refining requirements, identifying edge cases, evaluating plans, and validating AI-generated outputs. Their role has shifted away from producing every line of code to making sure that code implementation accurately reflects intent.

Program and product managers

Our program and product managers continue to perform many of their traditional responsibilities, including stakeholder management, prioritization, and roadmap planning. What’s changed is the degree of ownership that they maintain throughout the development lifecycle.

“In spec-driven development, PMs become the owners of the spec; they’re no longer just responsible for planning and prioritization. Everything starts with the spec, so they become the starting point for the entire system.”

Apoorv Gupta, principal software engineer, Microsoft Digital

In an SDD environment, product managers become stewards of the specification itself. They help define success criteria, resolve ambiguities, document business requirements, and check to make sure that priorities remain visible throughout implementation. The specification becomes a shared contract among stakeholders and a common source of truth for the entire team.

“In spec-driven development, PMs become the owners of the spec; they’re no longer just responsible for planning and prioritization,” Gupta says. “Everything starts with the spec, so they become the starting point for the entire system. If the specs are right, then everything falls into place.”

Architects

Architects establish the guardrails that govern the development process. One of the SDD first activities is defining a constitutionthat captures architectural principles, governance requirements, security standards, and development constraints. Architects are responsible for creating and maintaining those guardrails.

Because architectural decisions are made earlier in the process, teams can identify issues before implementation begins rather than discovering them during code reviews or testing cycles.

Looking ahead

Spec-driven development continues to evolve at Microsoft. But our development teams already view it as a vital component of AI-native engineering. The goal is to create a system that preserves intent as work moves from business requirements to implementation.

“SDD gives PMs a level of empowerment that wasn’t there before. You can move from an idea to a prototype, show it to customers and leaders, and start generating feedback much faster. Once teams experience that, they keep creating value, generating new ideas, and building on what they’ve learned.”

Gopal Panigrahy, principal product manager, Microsoft Digital

As AI coding agents become more capable, clear specifications become more valuable. Teams that can define requirements, establish guardrails, and maintain shared ownership of intent will be better positioned to scale AI-assisted engineering without sacrificing quality or governance.

For Microsoft Digital, SDD represents an important step toward Frontier Firm ways of working, where humans provide direction, judgment, and accountability while AI helps accelerate execution. Preserving intent, not simply generating code, is what enables teams to move faster while continuing to deliver outcomes that meet business needs.

“SDD gives PMs a level of empowerment that wasn’t there before,” says Gopal Panigrahy, a principal product manager in Microsoft Digital. “You can move from an idea to a prototype, show it to customers and leaders, and start generating feedback much faster. Once teams experience that, they keep creating value, generating new ideas, and building on what they’ve learned.”

The post Engineering the Frontier Firm: Sharing our AI-native approach to software development appeared first on Inside Track.

At Microsoft, we’re developing a growing agent ecosystem that offers diverse options for your workplace demands. This collection of tools acknowledges that there is no single “right” way to build an agent.

“The best agent architecture isn’t the most powerful one. It’s the simplest one that delivers value today while giving you the flexibility to grow tomorrow.”

Patrice Pelland, partner engineering manager, HR Employee Experience

Microsoft 365 Copilot, declarative agents, Copilot Studio, Microsoft Fabric data agents, and Microsoft Foundry agent service all represent valid patterns that are optimized for different kinds of work. Newer ideas like scout-style personal agents and Frontier Tuning (how we fine-tune our agent compliance) offer even more flexibility.

When choosing the best agentic technology for you, the rule of thumb is simple: Select the lightest architecture that fits your user experience, data needs, action model, and governance requirements.

“The best agent architecture isn’t the most powerful one,” says Patrice Pelland, a partner engineering manager in HR Employee Experience. “It’s the simplest one that delivers value today, while giving you the flexibility to grow tomorrow.”

Start with the surface, not the stack

The fastest way to get agent architecture wrong is to begin with technology instead of the workload. A better starting point is the user surface—where the work happens and how people expect to interact with it.

Our agent ecosystem is broad, but each of the core options map cleanly to a different kind of challenge:

• Microsoft 365 Copilot is the default entry point for AI in daily work, embedded across apps like Teams, Outlook, and Word. It excels at summarization, drafting, and reasoning over Microsoft Graph data without requiring users to leave their most familiar tools.
• Microsoft Copilot Cowork takes Microsoft 365 Copilot beyond just answering questions to actually executing on work assignments. Designed for complex, multi-step tasks, it can plan, coordinate, and carry out work across Microsoft 365 applications and connected systems while keeping users in control through approval checkpoints and oversight.
• Declarative agents extend Microsoft 365 Copilot with scoped instructions, knowledge, and lightweight actions. They’re the simplest way to create focused, role-based assistants while staying entirely within the Copilot experience.
• Copilot Studio agents provide a low-code environment for building workflow-driven agents with connectors, triggers, approvals, and multi-channel delivery. They’re best suited for business processes that need structure, governance, and integration across systems.
• Microsoft Foundry Agent Service is a code-first platform for building and operating custom agents with full control over models, orchestration, and runtime. It fits engineering-heavy scenarios that require scalability and tailored architecture.
• Microsoft Fabric data agents specialize in governed analytics over business data, using semantic models, ontologies, and curated datasets. They are designed to answer questions about KPIs, entities, and operations with trusted, auditable context rather than general document retrieval.
• Microsoft Scout and Autopilots introduce always-on personal agents that can coordinate work in the background across Microsoft 365. They focus on continuity—tracking commitments, preparing context, and proactively surfacing what matters next.
• Frontier Tuning enables organizations to teach agents how work actually gets done using company-specific data, workflows, and evaluation signals. It’s a later-stage capability that refines behavior inside compliance boundaries, rather than serving as a starting point.

Taken together, these are layers in a system. The practical decision is to pick the one closest to the user.

Picking the right agentic tool

Use Microsoft 365 Copilot as the default front door for broad productivity and Microsoft 365-grounded experiences. Extend it with declarative agents, connectors, plugins, or APIs when a scenario needs more domain-specific knowledge or actions. Do not treat plain Copilot as the primary architecture when the real need is a departmental specialist, governed workflow, analytics specialist, custom channel, or custom runtime. Those needs fit Copilot Studio, Fabric, or Foundry patterns better.

Use Microsoft Cowork when you want to complete a meaningful piece of work across multiple steps, files, applications, and systems. Cowork is best suited for long-running tasks that require planning, orchestration, and execution over time, while remaining grounded in Microsoft 365 context. Do not use Cowork as a replacement for workflow platforms, analytics specialists, or custom agent architectures. When a task requires governed business processes, specialized analytics, or deep runtime customization, Cowork should coordinate with other agentic tools rather than replace them.

Use Copilot Studio when governed workflow, connectors, approvals, and channel reach matter as much as the chat experience. Do not use it as the default for a small Microsoft 365 Copilot extension or for engineering-first scenarios that require custom model choice, runtime hosting, private networking, or deep orchestration; those fit declarative agents or Foundry better.

Use Microsoft Foundry when low-code or Microsoft 365-native layers hit real limits around orchestration, runtime control, model selection, deployment architecture, or engineering operations. Do not start here for a simple business workflow, no/low-code service agent, or focused Microsoft 365 assistant unless custom runtime control is truly the differentiator.

Use Fabric data agents when you’re looking for analytics expertise in a broader architecture. Fabric IQ, semantic models, and ontology can help translate raw data into business meaning. Do not use them as a general enterprise assistant or workflow engine: they are read-only, analytics-focused, and not intended for broad action-taking, unstructured document help, or custom model/runtime control.

Use Microsoft Scout when you need a personal front-door and coordination layer. It should not serve as a replacement for every enterprise agent architecture. When work crosses into governed process, business analytics, custom orchestration, or domain systems, Scout should delegate to Copilot Studio, Fabric data agents, MCP tools, or Foundry specialists.

Use Frontier Tuning only after the basics are in place: clear workflows, governed data, meaningful evaluation signals, and operational maturity. It is a later-stage optimization for institutional know-how, not the first step for a lightweight assistant or a substitute for sound architecture, data governance, or human-in-the-loop controls.

AI-ready data: The foundation of effective agents

The success of an AI agent depends less on the model it uses and more on the quality of the data that grounds it. Agents can only reason effectively when they have access to information that is trusted, discoverable, secure, and connected to the broader decision-making context. Organizations that invest in AI-ready data create the foundation for more reliable and useful AI experiences.

“The competitive advantage in the age of AI won’t come from access to models alone. Organizations also need to prioritize AI-ready data, which allows agents to reason accurately, securely, and at enterprise scale.”

Ravi Goripathy, principal software engineering manager, HR Employee Experience

AI-ready data goes beyond standard measures of data quality, such as accuracy, completeness, and consistency. It also includes the semantics, governance, and contextual information that help AI systems understand what the data means, how it should be used, and the way it relates to other information across the business. High-quality data is essential, but AI systems need data that is also meaningful, connected, and trustworthy.

As organizations deploy more copilots and agents, data readiness increasingly becomes the limiting factor for success, rather than model availability. A strong AI-ready data foundation helps agents produce better results, apply business context consistently, and operate within appropriate security and governance boundaries. The most effective AI strategies treat data readiness as a core architectural requirement.

“The competitive advantage in the age of AI won’t come from access to models alone,” says Ravi Goripathy, a principal software engineering manager in HR Employee Experience. “Organizations also need to prioritize AI-ready data, which allows agents to reason accurately, securely, and at enterprise scale.”

Building reliable agent systems

As soon as an agent moves from answering questions to taking actions, architecture matters more.

A useful pattern is to separate responsibilities into layers:

MCP tools play a key role here by exposing actions as controlled, reusable capabilities. Instead of giving an agent broad access to systems, you expose bounded functions with defined inputs, permissions, and telemetry. This improves reliability, governance, and auditability.

Similarly, not every interaction belongs in free-form chat. Tasks that require review, approval, or comparison—like approving a budget or evaluating candidates—benefit from structured UI components. MCP apps let you surface those experiences inside conversational flows without rebuilding entire applications. Rather than sending users to a separate system to complete a task, MCP apps can render forms, grids, dashboards, and other transactional experiences directly within the conversation, enabling users to review information, approve requests, update records, and take action in context.

As solutions mature, new patterns emerge. Always-on personal agents—like Scout-style Autopilots—can monitor commitments, prepare context, and proactively surface issues. In parallel, Frontier Tuning introduces a way to teach agents how work actually gets done in a specific organization, using real workflows, evaluation signals, and internal knowledge.

These are powerful capabilities, but they’re not starting points. They’re extensions of an architecture that already works.

A layered approach beats a ‘super agent’

It’s tempting to imagine a single agent that handles context, analytics, workflows, and actions. In practice, a more durable approach is a set of loosely connected specialists that each focus on a specific responsibility.

A Microsoft 365–native assistant can serve as the front door for user interaction, delegating analytics to Fabric data agents, actions to Copilot Studio workflows or MCP tools, and more complex orchestration to Foundry-based agents when needed. This separation makes systems easier to govern, test, and evolve over time.

Before moving to production, architecture should include clear ownership, permission design, governance controls, human approval for sensitive actions, evaluation processes, and monitoring. These are all essential aspects of building an agent system that can be trusted.

Deciding where to start

When faced with multiple valid options, a simple decision path helps:

• If a packaged Microsoft experience already meets most of the need, use it.
• If the experience should live in Microsoft 365, start with Copilot and a declarative agent.
• If the core problem is workflow—approvals, routing, integration—lean toward Copilot Studio.
• If the value depends on governed business data, bring in Fabric data agents early.
• If you need deep control over orchestration, models, or runtime behavior, step up to Foundry.

And across all of these: add structure only where the workload demands it. Use MCP tools for actions, structured UI where chat breaks down, and human review where risk increases.

Most importantly, resist the urge to start at the most complex layer. Many successful solutions begin as simple Microsoft 365 extensions and evolve incrementally into multi-layered systems.

Looking ahead: From tools to systems of work

The trajectory of agent architecture is clear. We’re moving from isolated assistants toward interconnected systems that combine personal productivity, business data, workflows, and continuous coordination.

“Agentic AI is shifting from isolated capabilities to connected systems of work. The organizations that lead will be those that design for evolution—starting simple, learning continuously, and scaling with intention.”

Ravi Goripathy, principal software engineering manager, HR Employee Experience

In the near term, that means more layering: personal agents coordinating work, workflow agents handling processes, analytics agents grounding decisions, and platform agents orchestrating across systems. Over time, capabilities like Frontier Tuning will make these systems more adaptive, learning from how organizations actually operate while staying within governance boundaries.

But the core principle is unlikely to change. The most successful teams won’t be the ones who adopt the most advanced architecture first. They’ll be the ones who start with the smallest viable solution, validate it in real workflows, and scale intentionally—adding complexity only when it delivers clear value.

“Agentic AI is shifting from isolated capabilities to connected systems of work,” Goripathy says. “The organizations that lead will be those that design for evolution—starting simple, learning continuously, and scaling with intention.”

The post Start light, scale intentionally: Choosing the right Microsoft agent architecture appeared first on Inside Track.

But Radoi didn’t let it faze her. Even though she grew up without a computer in her home and studied criminal law in college, she found herself drawn toward a career in technology. After about a dozen years of working in IT roles for different companies, she was invited to apply at Microsoft Romania.

So, she interviewed and received an offer. At the exact same time, she learned (to her surprise) that she was pregnant. With this twist, she thought maybe that her offer would be withdrawn.

“I was pretty afraid to tell Microsoft about the pregnancy, because my experience with other companies in Romania is that they are not always supportive of women with children,” Radoi says.

But instead of uncertainty or rejection, she was met with support and understanding.

“It was amazing,” Radoi said of her future Microsoft manager’s reaction to her disclosure. “After a quick conversation with HR, he just said, ‘Welcome to Microsoft. We want you as part of our team, and this pregnancy doesn’t change that. It’s not an issue for us.’ It was the first time I’d received that kind of professional support and encouragement as a woman in IT.”

In the eight years since that conversation, Radoi has thrived in the supportive Microsoft culture and forged a successful career here.

“I really love what I’m doing here at Microsoft. I love my leaders, and I trust them. That’s huge for me. It’s allowed me to grow and do my best work.”

Adelina Radoi, senior IT field manager, Microsoft Digital

She now serves as a senior IT field manager for Microsoft Digital—the company’s IT organization—based in Bucharest. It’s been a great match, and she continues to learn and develop as a professional.

“I really love what I’m doing here at Microsoft,” Radoi says. “I love my leaders, and I trust them. That’s huge for me. It’s allowed me to grow and do my best work.”

Having a growth mindset

From the beginning, Radoi stood out for her ability to learn quickly and take on new challenges—even without a traditional technical background.

“Adelina is one of those people that consistently models what we call a growth mindset,” says Simon Price, who works with Radoi as a senior director of IT Field Management in Microsoft Digital. “She’s a resilient, energized individual who is always looking for opportunities to drive improvements. And she sees the wider picture really well.”

Radoi enjoys the collaborative aspect of her job, and she works seamlessly with her both her coworkers in Romania and Microsoft employees in other regions around the globe. Her strong communication skills really allow her to shine in this area.

“Her default mode is to collaborate with others and partner with them; she’s built a strong network in Microsoft Romania but also at a global scale, reaching out and building new connections.”

Simon Price, senior director, IT Field Management, Microsoft Digital

Radoi is a people person who speaks three languages—Romanian, French, and English—while understanding Italian, Spanish, and Portuguese (and is working on her Japanese). These traits and skills have allowed her to make a strong impression across Microsoft Digital.

“Her default mode is to collaborate with others and partner with them; she’s built a strong network in Microsoft Romania but also at a global scale, reaching out and building new connections,” Price says. “Others constantly talk about how great a partner she is, and how she’s willing to speak up and share her perspective.”

One project on which Radoi exemplified this mindset was I Am New, a portal for brand-new Microsoft employees where they can get up and running technically with minimal assistance. She helped come up with the concept and worked on the team that drove the initial development of the portal.

“I’m incredibly proud of I Am New,” she says. “The idea grew from a challenge many of us were seeing across Microsoft: How do we create the best Day 1 experience for new hires? Romania was one of the early voices raising this need, and the feedback we received locally closely mirrored insights from colleagues across other Microsoft countries. What started as a shared challenge became a collaborative solution that now helps new hires feel connected, informed, and empowered from day one.”

Price says that the project showed the kind of outsized impact that Radoi has had in her work.

“It gave her the opportunity to work within a global team and develop a solution that would help every new Microsoft employee,” he says. “She embraced the challenge, partnered effectively with others, and had a large-scale impact on the effort.”

Serving as a mentor and role model

After going through her own challenges as a woman breaking into the IT field, Radoi knew she wanted to help mentor others in the same situation. That’s why she has played a pivotal role in several women-in-tech groups at the company.

“I’ve really enjoyed being a leader in these groups and serving as a mentor for other women,” she says. “As someone who faced unconscious bias in her career, I want to foster a nurturing environment and create a more inclusive workplace. I mentor new graduates and speak at various diversity events to contribute to the growth and empowerment of women in the field.”

One of Radoi’s own female mentors at the company is Silvina Olkies, a senior director of Global End User Support Services and Employee Experience.

“Adelina’s very genuine and brings a real warmth to the workplace. She’s proactive and is always focused on giving back in different ways. And she’s gone through complex challenges and embraced transformation throughout her career.”

Silvina Olkies, senior director of Global End User Support Services and Employee Experience, Microsoft Digital

Olkies has gotten to know Radoi by working with her on various employee resource group initiatives over the years. She admires how Radoi brings her authentic self to her job and strives for human connection and community.

“Adelina’s very genuine and brings a real warmth to the workplace,” Olkies says. “She’s proactive and is always focused on giving back in different ways. She’s gone through complex challenges and embraced transformation throughout her career.”

Olkies remembered fondly that Radoi even composed and performed a special song to celebrate her 25^th anniversary at the company, exhibiting some of her many talents. (Radoi also studied ballet seriously in her youth and credits this experience with contributing to her discipline and work ethic.)

“She’s the only field manager I know that has had an office project going on just about every day of her entire time on the job. Her site has grown almost double in size during her time here, and it’s kept her extremely busy.”

Anders Jepsen, director, Field IT Management, Microsoft Digital

Connecting Microsoft with the world

Microsoft Romania is growing fast, and that has given Radoi plenty of challenges during her first eight years with the company. Part of her role involves coordinating the technology aspects of new buildings, and she has had an extraordinary number of those to handle so far.

“She’s the only field manager I know that has had an office project going on just about every day of her entire time on the job,” says Anders Jepsen, who has been Radoi’s manager since she started with the company. “Her site has grown almost double in size during her time here, and it’s kept her extremely busy.”

As a rising star in the tech field, Radoi has had offers to move to jobs in other parts of the world. It would be a natural step for an ambitious career professional.

“I’m so happy that Microsoft Romania exists. For me, it’s very important to remain in my country. I love my people; I love Romania.”

Adelina Radoi, senior IT field manager, Microsoft Digital

But she’s decided to stay in Romania, where she’s raising her daughter and doing all she can to bring new technologies—such as AI—to her native country.

“I’m so happy that Microsoft Romania exists. For me, it’s very important to remain in my country. I love my people; I love Romania,” she says. “A lot of the best Romanian minds are going overseas. But if we are all leaving Romania, who is going to do something for this country? So I’ve turned down the offers to leave.”

She’s further demonstrated this commitment with her outreach work for Microsoft. She has served as the Microsoft representative at various government meetings and discussion panels in Romania, dealing with issues such as privacy, child safety, and AI. She’s also done podcasts and other engagements.

“She’s constantly looking for new opportunities to widen her scope,” Jepsen says. “Whether it’s traditional customer engagements, diversity and inclusion events, speaking at universities, or other outreach, she’s a role model for the field manager position. Her biggest challenge is knowing her own limits.”

As a mother, the work on child safety and AI is particularly close to Radoi’s heart.

“Children are really curious and excited about AI, and it’s everywhere,” she says. “As a country, I think we need to do more to teach about AI, and to make sure the digital world is safe for our kids. So I’ve gotten involved in that work, brainstorming how to develop the best technology safeguards for our children.”

And her advice when she speaks to young people about careers in technology, or mentors others who aim to follow in her footsteps? She tells them to set ambitious goals and remain true to themselves.

“Follow your heart and instincts—believe in yourself and create your own image,” she says. “Compare yourself with who you were one, two, or three years ago. What you have you learned? How have you improved yourself? If you like the answer, then you are on the right path.”

The post Rising to the challenge: How Adelina Radoi makes a difference in IT at Microsoft appeared first on Inside Track.

With an environment comprising 100,000 network devices (including access points) and over 900 buildings (including data centers), which supports 350,000 users and over 1 million connected devices generating 300,000 incidents per year, traditional methods of network management can often be insufficient.  

To operate at this scale, we’re building and deploying AI solutions that help us manage our network more efficiently, respond to issues faster, and reduce manual work. This is a critical part of our role as the team that’s responsible for powering, protecting, and transforming our digital employee experience across devices, applications, and hybrid infrastructure at Microsoft.

“As the organization that acts as Customer Zero for the company, we are leading with AI to help power Microsoft and keep things like our network infrastructure secure and reliable. We must aggressively experiment and adopt AI agents to foster innovation and model the future of service engineering.”

Brian Fielder, vice president, Microsoft Digital

Within Microsoft Digital, our AIOps and Network Infrastructure Copilot (NiC) AI solutions have emerged as transformative new tools for enhancing network performance. These tools take advantage of AI-powered capabilities by turning data and knowledge into powerful insights and, eventually, meaningful actions to run the industry’s most secure and reliable enterprise network. They also share a common data architecture comprised of millions of telemetry points but perform unique AI functions in both running and supporting the network.

AIOps is an automation solution that uses data insights to prevent and resolve network issues before they become impactful. NiC is an interactive experience that helps network practitioners, and a variety of other personas more easily interact in natural language with complex network services to get rapid answers and intelligent suggestions.  

“As the organization that acts as Customer Zero for the company, we are leading with AI to help power Microsoft and keep things like our network infrastructure secure and reliable,” says Brian Fielder, vice president of Microsoft Digital. “We must aggressively experiment and adopt AI agents to foster innovation and model the future of service engineering.”

AIOps: Transforming how we deliver operational excellence 

AIOps integrates AI into our IT operations by automating and enhancing various aspects of network management. This approach significantly reduces manual intervention, allowing network engineers to focus on higher-value tasks. Key components of AIOps include: 

• Automated ticketing and remediation: AIOps automates the creation and resolution of tickets, reducing the time and effort required to manage incidents. This automation is particularly beneficial in environments with high ticket volumes, ensuring timely and efficient resolution of network issues. 
• Ticket noise reduction: AIOps uses AI-powered ticket correlation, suppression, and enrichment capabilities to significantly reduce ticketing noise, enabling engineers to concentrate on the most critical issues.  
• Automatic remediation: AIOps executes automatic troubleshooting and remediation actions on behalf of engineers to mitigate issues and keep outage duration and the associated business impact to a minimum. This is accomplished based on troubleshooting knowledge and successful remediation steps executed. 
• Postmortem report generation: AI-powered tools generate detailed postmortem reports for network incidents, providing insights into the root causes and recommended remediation steps. This capability enhances the learning process and helps prevent future occurrences. 

Network Infrastructure Copilot (NiC): The everyday AI assistant 

NiC is an AI assistant designed to support network engineers in managing complex network environments. NiC provides powerful data insights and documentation, enabling engineers to design, configure, analyze, and troubleshoot network issues using natural language queries. Key features of NiC include: 

• Data insights: NiC helps engineers extract valuable data insights from various sources, such as wikis, SharePoint libraries, troubleshooting guides, and the infrastructure data lake (IDL). This capability streamlines data analysis and enhances decision-making to take impactful actions. 
• Network artifacts interpretation: NiC understands key artifacts such as device logs and configurations, providing engineers with concise and relevant information. This capability greatly reduces the cognitive load on engineers to access and process the most critical data and insights required to manage a complex network environment. 
• Simplified network observability: NiC enables non-networking personas (such as conference room technicians and facilities managers) to get quick glances at the health and configuration of their services without requiring deep understanding of network protocols and taxonomies. 

“NiC and other AIOps agents have dramatically reduced the time engineers spend searching through documentation and other network artifacts to yield actionable insights, slashing effort from 25 minutes to under 5 minutes.”

Phil Suver, principal group product manager, Microsoft Digital

“NiC and other AI Ops agents have dramatically reduced the time engineers spend searching through documentation and other network artifacts to yield actionable insights, slashing effort from 25 minutes to under 5 minutes,” says Phil Suver, a principal group product manager in Microsoft Digital.

“By staying close to our teams’ real needs, we were able to turn AI opportunities into practical solutions, delivering near-term value while laying the groundwork for lasting innovation.”

Anand Meduri, principal PM manager, Microsoft Digital

Driving impact through AIOps and next-gen AI agents 

These AI-driven solutions have significantly improved network reliability in several ways: 

• Efficiency gains: The automation of routine tasks and the provision of actionable insights have drastically reduced the cognitive load on network engineers. In recent years, network practitioners have cumulatively saved over 20,000 hours on network infrastructure management. These efficiency gains free up engineers to focus on strategic initiatives that further enhance network performance.
• Rapid issue detection and resolution: AI-powered anomaly detection and automated remediation ensure that potential issues are identified and resolved before they impact network performance. This proactive approach minimizes downtime and enhances overall network reliability.

“By staying close to our teams’ real needs, we were able to turn AI opportunities into practical solutions, delivering near-term value while laying the groundwork for lasting innovation,” says Anand Meduri, a principal PM manager for Microsoft Digital.

Lessons learned in building and deploying AIOps agents

To deliver sustainable business value at enterprise scale, we adopted an iterative approach focused on rapid experimentation, measurable outcomes, and continuous learning. As we evolved from automation-centric operations to an agent-driven operating model, three major themes emerged:

“By embedding AIOps and AI agents into our operational fabric, we are transforming manual workflows into autonomous, scalable digital labor. This is accelerating our journey toward a human-led, Frontier Firm future.”

Suvodip Moitra, senior product manager, Microsoft Digital

• Prioritization and value realization: Not every workflow benefits equally from AI. We achieved the greatest impact by focusing on high-volume, repetitive, and time-sensitive operational activities such as incident triage, outage analysis, troubleshooting, and cross-system coordination. Prioritizing these use cases enabled us to demonstrate tangible productivity gains while building confidence in agent-led operations.
• Human-agent collaboration and change management: The transition to agentic operations required more than technology adoption—it demanded new ways of working. Success depended on positioning agents as trusted digital teammates, augmenting engineers rather than replace them. Continuous feedback loops, user engagement, and incremental rollout strategies were critical to help us drive adoption and improve agent effectiveness over time.
• Scalability, autonomous operations, and the Frontier Firm future: As our operational footprint continues to grow, we are moving from automated workflows to a scalable ecosystem of intelligent AI agents that can reason, coordinate, and act across the incident lifecycle. By embedding agents such as Smart Bonding, Outage Insights, and On-Demand Troubleshooting into daily operations, we are reducing human toil, accelerating resolution, and laying the foundation for a Frontier Firm operating model where engineers focus on strategy and innovation, while digital labor drives operational execution at scale.

“By embedding AIOps and AI agents into our operational fabric, we are transforming manual workflows into autonomous, scalable digital labor,” says Suvodip Moitra, a senior product manager in Microsoft Digital. “This is accelerating our journey toward a human-led, Frontier Firm future.”

The post Enhancing Microsoft network reliability with AIOps and Network Infrastructure Copilot appeared first on Inside Track.

Enter the power of sophisticated data tools and agentic AI. 

A great example of this came when our Microsoft Digital engineers and product managers were trying to get a handle on our sprawling software licensing landscape—thousands of third-party tools that our employees rely on in their work.

“We realized there were all these fragmented, scattered repositories of licensing data across many teams, all with different ownership,” says Ahmed Musa, a senior software engineer in Microsoft Digital, the company’s IT organization. “There was no visibility into what contracts existed or how they were being used. We needed a single solution.”

The answer was IntelLicense, an enterprise-wide intelligence platform that collects product information, licensing contracts, cost data, employee usage telemetry, and supplier details in one system. This all comes together in our Software Asset Management (SAM) portal, where we deliver enterprise-grade governance through a modern user experience.

“IntelLicense is a game-changer for us. Before, internal software licensing was manual and labor-intensive—it could take months to make sense of the data. Now, an answer that used to take up to six months for us to track down can be generated immediately using this platform we’ve created.”

Ahmed Musa, senior software engineer, Microsoft Digital

Our IntelLicense platform uses the advanced capabilities of agentic AI to answer queries and generate insights on the data, giving us much greater understanding and visibility while enabling us to reduce license duplication, simplify procurement, and cut spending.

This benefits our employees who need to license software, our software asset managers, and our Procurement team, which manages the process on our back end to make sure the company isn’t wasting money and resources—especially considering the company makes significant annual investments in third-party software licensing.

This solution shows how at Microsoft we’re constantly looking for ways to apply AI to help solve enterprise-level challenges at scale—the hallmark of a Frontier Firm.  

“IntelLicense is a game-changer for us,” says Musa, the principal architect for the project. “Before, everything around internal software licensing was manual and labor-intensive,—it could take months to make sense of the data. Now, an answer that used to take up to six months for us to track down can be generated immediately using this platform we’ve created.”

Uncovering the challenge

With more than 200,000 employees working across over 100 countries worldwide, attempting to centralize information at an organization the size of Microsoft is never easy. The state of our third-party software licensing system was no different.

“As we looked beyond just employees finding software and deeper into the process, we began to see the challenge was also about purchasing, how we dealt with suppliers, and how we managed the licenses at a higher level. That’s where we saw the big opportunity.”

Revanth Chandra Pydimarri, senior product manager, Microsoft Digital

We began this journey nearly three years ago. The first big need we identified came from employee feedback that indicated it was difficult to figure out how to identify and license third-party software tools. But as we began to analyze the larger picture, we realized that the problem was much more layered and complex.

“As we looked beyond just employees finding software and deeper into the process, we began to see the challenge was also about purchasing, how we dealt with suppliers, and how we managed the licenses at a higher level,” says Revanth Chandra Pydimarri, a senior product manager in Microsoft Digital. “That’s where we saw the big opportunity.”

But by expanding the scope of the project, we were setting off on a long and technically daunting quest.

“I think we counted 19 different systems that contained relevant licensing data. Working with all the different teams to pull that data together was the first big challenge we had to go after.”

Jay Selveraj, principal software engineering manager, Microsoft Digital

Tackling the data first

The first step was to gain visibility into all our third-party software contracts, our suppliers, and the actual product usage across the company. But our teams were operating in silos, each maintaining their own agreements and data about software licenses.

“This was fundamentally a data problem,” says Jay Selveraj, a principal software engineering manager in Microsoft Digital. “The enterprise data for license management is highly distributed, non-standard, and spread across the company. I think we counted 19 different systems that contained relevant licensing data. Working with all the different teams to pull that data together was the first big challenge we had to go after.”

To fully understand the software asset management process, Selveraj and Chandra Pydimarri were charged with creating a journey map to show the steps, dependencies, and stakeholders involved.   

“It was a very daunting task for us,” Selveraj says. “We identified so many different bottlenecks. And that’s when we decided we can’t just troubleshoot the existing process—we needed to build a new platform that would span the enterprise.”

To accomplish this, they turned to Microsoft Fabric, which at the time was a relatively new product. Fabric provided the power and flexibility needed for this kind of project.

“Microsoft Fabric was designed as a unified data platform for engineers, making it an ideal fit for this project.”

Misrak Ararso, senior software engineer, Microsoft Digital

And as Customer Zero for Microsoft, we were excited to be early adopters of Fabric (it had just gone into public preview). The fact that we were able to try it out on a real enterprise challenge we were facing was both a strategic advantage and a bonus.

“Microsoft Fabric was designed as a unified data platform for engineers, making it an ideal fit for this project,” says Misrak Ararso, a senior software engineer in Microsoft Digital, who also worked on IntelLicense. “It has great features like Data Wrangler, which allowed us to drill down on the data and clean it up quickly. We also used Microsoft OneLake, which meant we avoided having to duplicate data before working on it.”

Ararso appreciates how Microsoft Fabric continues to evolve with new AI capabilities, making it an increasingly powerful and beneficial tool for data engineering.

“Early adoption wasn’t always smooth,” she says. “We encountered challenges, sharing feedback when we did, and we benefited from improvements as the platform matured alongside our implementation.”

Reducing waste and saving money in procurement

Before we developed IntelLicense, our Procurement team at Microsoft also struggled to answer basic questions about our software licenses.

“It was very difficult to gauge usage, consolidate agreements, and do cost optimization. And when we tried to audit our contracts and move licenses around, it all had to be done manually and took a lot of time and effort. IntelLicense addresses that.”

Rasa Amiri, senior sourcing manager, Financial Operations

The Procurement team is responsible for negotiating contracts, pricing, and terms and conditions with thousands of different suppliers. However, it can be difficult to negotiate volume discounts and manage the other aspects of licensing if you don’t have a holistic view across the enterprise.

In a typical example, one group at Microsoft might purchase 20 software licenses from a particular supplier, but then only use 15 of them. Another team needs 5 licenses, but they have no idea that there are unused licenses they could tap from the other group, so they purchase their own. And when an employee moves teams or leaves the company, their software licenses often go unused rather than get reassigned.

“It was very difficult to gauge usage, consolidate agreements, and do cost optimization,” says Rasa Amiri, a senior sourcing manager in our Financial Operations group. “And when we tried to audit our contracts and move licenses around, it all had to be done manually and took a lot of time and effort. IntelLicense addresses that.”

IntelLicense structure

The IntelLicense solution consists of three parts: a UX layer, an agentic AI layer, and a data layer.

According to Amiri, one helpful feature of IntelLicense is the ability to see if a software license is not being used, and then directly contact that employee (or license owner) to say, “Hey, it looks like you’re not using this license. Can we reallocate it?”

“We have a real-time dashboard called the SAM portal that we can now use for that, focused on our top 200 suppliers,” Amiri says. “Now, every time we negotiate a deal, it’s uploaded into IntelLicense with all the details—the cost, the contract, the number of licenses. Not only does it help us with reallocation, it helps us quickly resolve issues we have with suppliers who want to charge us for overuse.”

Musa agrees.

“The IntelLicense platform can identify overlapping tools already in use and surface relevant alternatives, enabling more informed, cost-efficient decisions across the organization,” he says.

Introducing these kinds of efficiencies can quickly generate significant cost savings at an organization the size of Microsoft. Our internal data shows that IntelLicense drove substantial savings in software licensing fees over the last fiscal year. And we have greater ambitions for the future—Chandra Pydimarri cited industry studies that show up to 20% of third-party software spending is unnecessary. That’s huge potential savings for an enterprise organization.

“As we evolved the platform, we realized that users don’t want just another dashboard—they need decision intelligence. They need a system that can connect signals across datasets, surface actionable insights, and guide decisions in real time, so they can move faster and act with confidence.”

Urvi Sengar, senior software engineer, Microsoft Digital

Adding an AI layer

The Software Asset Management portal was a strong foundation for centralizing licensing data, but we wanted to take the solution further.

It was one thing to centralize and surface data with all the relevant data about third-party software licenses. It was a whole different challenge to build a system that helped the user understand the data, ask the right questions, and turn insights into decisions.

“As we evolved the platform, we realized that users don’t just want another dashboard—they need decision intelligence,” says Urvi Sengar, a senior software engineer in Microsoft Digital. “They need a system that can connect signals across datasets, surface actionable insights, and guide decisions in real time, so they can move faster and act with confidence.”

So Sengar and her fellow engineers set to work adding an agentic layer to IntelLicense that could provide those AI-driven insights. They used Microsoft Foundry to create a multi-agent solution that could handle all the various needs users of the system might have.

“With the multi-agent architecture that we followed, we have a workflow manager that delegates any user query to specialized agents,” Sengar says. “One agent handles license management, another deals with supplier management, another can support audit scenarios. Each agent understands the user’s intent and can call any deterministic workflows when needed.”

Sengar sees the agentic layer as the transformation of IntelLicense from a reporting tool into an intelligent system that can deliver contextual, on-demand insights and help users take action through workflows. It also aligns with the growing expectations for a more conversational, Copilot-like AI experience, where users can ask questions naturally and receive meaningful, actionable responses in real time.

“The platform delivers proactive insights through the portal while also enabling users to explore them on-demand, in the context of their work,” Sengar says.

She goes on to describe a scenario where a software asset manager is in the middle of negotiating a contract with a supplier. If they have a new idea, question, or strategy they want to validate, the portal can surface relevant recommendations and insights right away. If they want to go deeper, they can use the chat interface to ask questions and get instant access to the latest context-aware information in a dynamic way, without having to leave the flow of their work.

“That’s where we see the future of AI-driven work going,” Sengar says. “It’s using AI not only to surface insights, but to help people explore them further, act on them, and make better decisions faster.”

Applying intelligence across the enterprise

Large enterprise organizations like Microsoft face this kind of challenge in many areas: how to maximize efficiency by centrally managing a process that is scattered across many different teams and systems, with data that is often inaccessible or systems that are incompatible. Teams have often developed different ways of accomplishing the same task and are reluctant to change.

“We want to make the biggest difference for the company—that’s the ultimate goal. At the end of the day, we want to make sure there is plenty of cost savings produced. Beyond that, we want to apply as much intelligence as possible to the problem, so that AI is impacting all aspects of the process.”

Senthil Selveraj, principal group product manager, Microsoft Digital

Our approach is to apply AI where it makes sense, using continuous improvement principles to guide us. We also look to our AI councils to make sure that we’re following best practices.

This ensures that when we at Microsoft Digital tackle something like software licensing, we’re going to achieve a transformational result that will pay big dividends across the company.

“We want to make the biggest difference for the company—that’s the ultimate goal,” says Senthil Selveraj, a principal group product manager in Microsoft Digital. “At the end of the day, we want to make sure there is plenty of cost savings produced. Beyond that, we want to apply as much intelligence as possible to the problem, so that AI is impacting all aspects of the process. That’s where we’ll see the largest, most impactful benefits.”

The post Taming software licensing sprawl at Microsoft with an AI-driven solution appeared first on Inside Track.

At that scale, Python dependencies have emerged as a significant source of risk for us—representing the third-largest vulnerability surface across the company.

The good news is that we have strong visibility into these vulnerabilities, with tools that continuously detect and surface risks across our codebases. The bad news is that turning those insights into action required a complex remediation process.

Updating a single code package often caused changes across multiple interdependent libraries. This required coordinated updates, validation, and testing to maintain system stability.

“When AI arrived, I saw it as a great opportunity to finally fix a very complex problem we had: The level of entanglement involved in Python code dependencies. A simple script wasn’t going to resolve it—you needed the power of AI.”

Humberto Arias, senior product manager, Microsoft Digital

Multiply this by thousands of projects throughout our enterprise, and vulnerabilities accumulated much faster than we could resolve them. To address this challenge, we turned to AI.

Microsoft Digital—the company’s IT organization—has developed an AI-powered solution called Python Dependency Remediation. Designed to work directly within the developer workflow, this solution analyzes dependency chains, applies required updates, and automatically adjusts the code. This enables our engineers to remediate vulnerabilities quickly and consistently at enterprise scale.

“I’ve worked for years in the vulnerability management space at Microsoft,” says Humberto Arias, a senior product manager in Microsoft Digital. “When AI arrived, I saw it as a great opportunity to finally fix a very complex problem we had: The level of entanglement involved in Python code dependencies. A simple script wasn’t going to resolve it—you needed the power of AI.”

The tool has shown so much promise that we have begun releasing it externally, so that millions of Python developers around the world can take advantage of it.

“I used to have this problem all the time. I upgrade one library, and then I’ve got to upgrade 17 other things, and something else breaks, and now my code is completely different.”

Rich Chiodo, principal software engineer, Python and Tools for AI

Get started

Try using the Python Dependency Remediation tool

Flexibility leads to dependencies and risk

Python is a very flexible language, which is why it’s so popular among software developers. But that same flexible nature—it can be used across a wide range of scenarios—also means it forms deeply interconnected dependency chains. When one code library is updated, it can trigger changes across many others.

“I used to have this problem all the time,” says Rich Chiodo, a principal software engineer on the team responsible for Python Tools and AI. “I upgrade one library, and then I’ve got to upgrade 17 other things, and something else breaks, and now my code is completely different.”

“Developers avoid the upgrades because the dependency web is so complex. This means the vulnerabilities accumulate over time and can become a real security risk.”

Chintan Sheth, principal engineering manager, Viva Glint

Because the code is so interdependent and remediation is time-consuming, many developers skip updating their code packages, which can lead to security vulnerabilities.

Security compliance was often seen as a burden because it slows people down.

“Developers avoid the upgrades because the dependency web is so complex,” says Chintan Sheth, a principal engineering manager on the Viva Glint product team. “This means the vulnerabilities accumulate over time and can become a real security risk.”

“After my manager mentioned it, I reviewed the idea on the hackathon page, and it looked really interesting to me. So I jumped in, and we created a prototype and a demo video with a quick solution. That’s how it started.”

Shiva Krishna Gollapelly, senior software engineer, Microsoft Digital

Hacking our way to a solution

Like some of the best internally developed tools and processes, Python Dependency Remediation came out of a Microsoft hackathon project. These grassroots events allow our engineers to tackle interesting technical challenges in a collaborative, creative way.

“After my manager mentioned it, I reviewed the idea on the hackathon page, and it looked really interesting to me,” says Shiva Krishna Gollapelly, a senior software engineer in Microsoft Digital and the lead developer on the project. “So I jumped in, and we created a prototype and a demo video with a quick solution. That’s how it started.”

The fact that this solution came from a hackathon highlights the ideas-driven culture that we promote at the company.

“This really speaks to our special culture of innovation,” says Snigdha Bora, a principal group engineering manager for Employee Experience. “After this emerged from the hackathon, our developers realized it could solve a problem at scale—that it was worth taking through the full development cycle so we can release it for all of Microsoft, and maybe beyond.”

Solving the issue with one click (and AI)

Because the challenge was not detecting vulnerabilities but fixing them, we had to rethink how we addressed Python dependencies.

“The extension automatically finds the right updates and then fixes the vulnerabilities, so developers don’t need to do the research, the manual upgrades and fixes, run test cases, debugging—all those things that used to take so much time. With our solution, it’s just one button click and it does all of that automatically.”

Shiva Krishna Gollapelly, senior software engineer, Microsoft Digital

In the past, when engineers received a vulnerability notification, they would have to step outside their development workflow and address the issue. What was needed was a solution that could be enacted within their normal workflow—integrating remediation directly into the tools they were already using.

So, we created the Python Dependency Remediation extension for Visual Studio Code, a common Python development environment. Once installed, engineers can address vulnerabilities in the flow of their work.

“The extension automatically finds the right updates and then fixes the vulnerabilities, so developers don’t need to do the research, the manual upgrades and fixes, run test cases, debugging—all those things that used to take so much time,” Gollapelly says. “With our solution, it’s just one button click and it does all of that automatically, with the help of AI.”

The extension uses the APIs built into Visual Studio Code to connect with any AI model the user has access to. (If there is no AI model available, Gollapelly explains, the extension will still make the package updates but won’t do the remediation fixes to the code.) It also produces a report of the changes for the developer to review in case there’s a snag that needs troubleshooting.

“This tool removes a significant burden from our developers,” Bora says. “We are shifting the entire remediation process left, embedding it early in the development workflow. Developers can review the changes and move forward immediately, making the whole process more efficient.”

“We’ve upgraded the library with new methods, calls, and structures. Now, let’s make sure everything works, check for errors in the code, etc. That’s the gap we’re bridging with AI.”

Angel Saldivia, software engineer, SharePoint

The result is that fixes and upgrades that used to take multiple hours of developer time now take minutes, and the code is much more reliable.

What the agent does in this solution is help close that loop, something that the engineer used to have to do.

“We’ve upgraded the library with new methods, calls, and structures,” says Angel Saldivia, a software engineer on the SharePoint product team. “Now, let’s make sure everything works, check for errors in the code, etc. That’s the gap we’re bridging with AI.”

From Customer Zero to global impact

One of the powerful things about working at Microsoft is that you get to help develop technology tools that can change the world. This is the case with Python Dependency Remediation as well.

“We realized this technology had much broader value. There are hundreds of millions of Python users worldwide, so the impact could be massive.”

Snigdha Bora, principal group engineering manager, Employee Experience

As Bora explains, while the solution was being developed it was presented to Guido van Rossum, the creator of Python (and a Microsoft employee). He immediately saw the incredible potential of the concept.

“He suggested that we could take this solution to the world, not just to Microsoft,” Bora says. “We realized this technology had much broader value. There are millions of Python users, so the impact could be massive.”

To help make this happen, Microsoft Digital approached Graham Wheeler, a principal group engineering manager on the Python and Tools for AI team. Wheeler’s team is responsible for shipping Pylance, a development extension for Visual Studio Code used by more than 180 million developers worldwide.

“One of the things we could do was provide a jumping-off point for this extension, so that when users installed Pylance they’d be prompted to also download Python Dependency Remediation. It can help raise awareness, because many users don’t actually do the dependency scanning and updating that they should.”

Graham Wheeler, principal group engineering manager, Python and Tools for AI

Wheeler and his team are in the process of incorporating the Python Dependency Remediation extension as an option during Pylance installation. This will open up a convenient vector for getting the tool in front of a huge audience, potentially revolutionizing Python development.

“One of the things we could do was provide a jumping-off point for this extension, so that when users installed Pylance they’d be prompted to also download Python Dependency Remediation,” Wheeler says. “It can help raise awareness, because so many users don’t actually do the dependency scanning and updates that they should. So, we’re helping with that challenge.”

Beyond Python, the AI-powered technology behind this extension might be applied to other dependency challenges as well. What started as a simple hackathon project could have huge ramifications for the future of software development.

“This solution can easily be adapted to other libraries, other programming languages,” Gollapelly says. “Whether you’re talking about C#, Angular, React, or another language, the concept is the same. The implications are vast.”

The post Taming our Python dependencies at Microsoft with AI appeared first on Inside Track.

Today, the company has grown into a global organization with more than 200,000 employees. They all rely on Microsoft Digital—the company’s IT organization—to provide the tools, technologies, and solutions that empower them to accomplish more every day.

The need for digital transformation

The history of information technology is one of constant evolution, and the pace of change has never felt greater than it does right now. The AI capabilities and other groundbreaking innovations unveiled in the last few years show the potential to radically transform our world and change the way we think about and operate all IT services.

When the world pivoted to remote online work and collaboration because of the COVID-19 pandemic, it was just one example of how digital transformation doesn’t always happen in a straight line or on a predictable schedule. Our company’s history of shaping and adapting its IT organization to the latest challenges faced by employees and partners is no different; marked by big bets and strategic shifts that reflect our ever-changing world.

Mapping our IT journey

Today, Microsoft Digital is the team that powers, protects, and transforms the digital employee experience across all devices, applications, and hybrid infrastructure at the company. Using our deep knowledge and experience in enterprise IT, we’re pivoting to help lead the company’s AI transformation while also sharing our journey with customers so they can take advantage of this generational opportunity to reshape their businesses and IT operations.

To understand where we’re going, it helps to take a look at where we’ve been. This article explores the details of the major eras of our IT history and then shifts to examine the trendlines and technological innovations that are shaping Microsoft now.

On-Premises IT (founding to 2009)

It’s useful to break the history of our IT operations into different eras. For the first three decades or so from its founding in 1975, Microsoft operated with on-premises IT systems. This era was characterized by the setup, operation, and maintenance of onsite physical technology—servers, datacenters, and other hardware infrastructure.

During this time, IT roles were narrowly defined. IT team members functioned primarily as “order-takers,” with limited influence over strategic decisions.

Because funding was inconsistent, our IT organization had limited growth opportunities and relied on vendors for development work. Gaps were filled in with “shadow IT,” where internal teams would sometimes procure their own hardware or software without formal IT approval or standards.

We established security as an early priority for the company. Cofounder Bill Gates launched the Trustworthy Computing initiative more than two decades ago, an effort emphasizing the importance of security, privacy, and reliability across Microsoft products and services both internally and externally.

Our On-Premises IT era established the foundation that would become crucial to the company’s future digital transformation.

All in on the cloud: The Cloud and Culture era (2010-2018)

Cloud computing marked the next significant shift in the history of IT at Microsoft. It began in 2010 under the leadership of CEO Steve Ballmer, signaling a major break with the previous era of physical IT infrastructure and an important step toward today’s distributed-computing world.

The launch of the cloud computing platform then known as Windows Azure heralded this new era, as we transitioned away from an IT philosophy focused on the Windows desktop client toward a more platform-agnostic view. Cloud computing offered extensive advantages for customers and for our own IT environment, in terms of cost, performance, security, and scalability.

We started our journey by moving productivity workloads (Exchange and SharePoint) to the cloud. Then, we shifted new development to Azure and optimized modern applications to run in the cloud. We also moved existing applications targeted for migration to virtual machines. Today, more than 98% of our IT workloads run on Azure.

Cultural transformation

Another important shift during this era was the profound cultural transformation at Microsoft sparked by new CEO Satya Nadella, who rose to the top job at the company in 2014. Nadella had previously run the Microsoft cloud computing and enterprise group, so he was already steeped in the idea of transformational change at the company.

“Achieving our mission requires us to evolve our culture. It all starts with a growth mindset—a passion to learn and bring our best every day to make a bigger difference in the world.”

Satya Nadella, CEO, Microsoft

Before Nadella’s ascension, Microsoft had long been known for its extremely competitive, “know-it-all” culture. Employees succeeded by showcasing their own individual achievements and how their accomplishments exceeded their peers.

Nadella changed this ethos by championing a growth mindset, encouraging employees to be “learn-it-alls” rather than “know-it-alls.” The shift included placing new importance on how employees contributed to the success of others, a value that was incorporated into individual performance reviews. Nadella made this transformation his personal mission and directed leadership to propagate the new philosophy at all levels across the organization.

“Achieving our mission requires us to evolve our culture,” Nadella says. “It all starts with a growth mindset—a passion to learn and bring our best every day to make a bigger difference in the world.”

The combination of the shift to cloud computing infrastructure and overhauling the company culture helped set the stage for the major technological innovations to come.

A new vision: The Modern Engineering era (2018-2023)

For years, IT at Microsoft had been order takers, doing what the business requested with limited ability to impact strategic priorities. That changed as we shifted to become a modern engineering organization. With support from our executive leadership, IT was elevated to become a peer engineering function at Microsoft.

Rather than simply taking orders, the team was empowered to lead with a strong vision for the future. In fact, leading with vision is the primary hallmark of our Modern Engineering era. As we moved into this era, we needed a clearly articulated view of our goals as an IT organization aligned to the needs of our business partners, as well as the resources needed to achieve them.

Role transformation

Transitioning to become a modern engineering organization required Microsoft Digital to adapt our legacy approach to IT.

Operating an engineering organization in a cloud environment meant new roles, new skills, and a new mindset. With no need to manage physical hardware, our modern IT professionals were freed to work more closely with business partners, requiring greater strategic acumen. The team was now focused on DevOps, Agile program management, and user-centric design principles.

User-centric, coherent design

Our design philosophy puts the user—an employee or guest—at the heart of every decision we make at Microsoft Digital.

The goal of this approach is to make tasks that might have previously caused friction to become simpler. Instead of dealing with disconnected systems, user-centric design introduces consistent and logical flow between services. This makes it easier for people to access services, learn how to use them, and then put them to good use.

Microsoft also embraces coherent design across all our products. A familiar look and feel, along with consistent usage patterns, accelerates employee usage and adoption. 

Embracing work-from-anywhere capability

During the pandemic, when our workforce was still fully remote, our organization was already starting to think about what the new hybrid workplace would look like when people started returning to the office. We identified three key dimensions of the employee experience:

• Physical spaces: We partner with Global Workplace Services to plan and deploy meeting spaces with amazing digital capabilities that support an inclusive approach to hybrid productivity.
• Digital capabilities: We keep employees productive and their digital environment safe and secure, no matter where they’re located or how they connect.
• Culture: A strong partnership with HR ensures that digital experiences support our company culture.

Managing shadow IT with a culture of trust

Shadow IT is the unknown and unmanaged set of applications, services, and infrastructure that are developed and managed outside standard IT policies. Shadow IT typically crops up when engineering teams are unable to support the needs of non-engineering partners, a situation that could arise from a lack of available capacity or the need for specialized domain solutions. 

While earlier eras of our IT history focused on trying to prevent shadow IT, we are now concentrating on managing it. We use Azure best practices to optimize shadow IT and Microsoft 365 governance policies to ensure that our corporate security, privacy, and accessibility standards are met. We empower our employees to create whatever they need within our tenant, including PowerApps, SharePoint sites, Teams channels, or agents, mitigating the need for “shadow” solutions while also providing visibility into how our employees are using our own technology.

Learn how optimizing our Microsoft Azure usage is helping us manage our Shadow IT.

The Era of AI (2023 to present)

The latest chapter in the history of our organization’s digital transformation is defined by the integration of AI and agents into IT operations. AI is revolutionizing how Microsoft does IT at enterprise scale, driving efficiency and innovation across the board. From the apps, workflows, and services that power our employee experience to the network, infrastructure, and devices that enable employee productivity, our AI-focused investments provide a solid foundation for the innovations that we are constantly implementing. As we look at the future of Microsoft Digital, we’re focused on four key priorities: security, service fundamentals, acting as Customer Zero, and AI-powered innovation. We’re working to excel in all four domains with the help of our industry-leading AI capabilities.  

“Our mission is to power and protect Microsoft, and that starts with an unwavering commitment to the Secure Future Initiative.”

Brian Fielder, vice president, Microsoft Digital

Securing our future

Security is our highest priority. The Microsoft Secure Future Initiative aligns every team with a shared approach, common priorities, and consistent milestones to harden our security posture across all products and services.  

“Prioritizing security above all else is critical to our company’s future,” Nadella says. “Every task we take on—from a line of code to a customer or partner process—is an opportunity to help bolster our own security and that of our entire ecosystem. If you’re faced with a tradeoff between security and another priority, your answer is clear: Do security.”

The Secure Future Initiative is built on three core principles: Secure by design, secure by default, and secure operations. As the company’s IT organization, we work relentlessly to fulfill the key pillars of the Secure Future initiative across all our systems, including:

• Safeguarding identities and secrets
• Protecting tenants and isolating production systems
• Securing networks and engineering systems
• Enhancing threat detection
• Expediting response and remediation

“Our mission is to power and protect Microsoft, and that starts with an unwavering commitment to the Secure Future Initiative,” says Brian Fielder, vice president of Microsoft Digital.

Secure Future Initiative | Microsoft

Foundations: Service fundamentals

The second pillar is to maintain the highest standards of service fundamentals. These are the essential capabilities and practices that enable us to deliver reliable, secure, and compliant services companywide. Adhering to the highest standards of service fundamentals ensures that our organization continues to play a critical role in running the company’s business and enabling innovation, agility, and resilience in a fast-changing and competitive environment.

Customer Zero

The third pillar is acting as Customer Zero for Microsoft’s most important products and services, like Copilot Studio, Microsoft Teams, and Agent 365. In Microsoft Digital, we take pride in being the first customer for a wide variety of Microsoft products and services, relentlessly focusing on our own employee experience to create products that enable every person on the planet to achieve more.

Being Customer Zero means forging a deep partnership between our IT organization and product engineering groups to envision the right experiences, co-develop innovative solutions, and then listen to and act on insights gathered from our employees. We work together to stay grounded in the way our employees use our products every day, so your employees can benefit from our insights prior to external product launches.

Read about how we’re improving our employee experience through our Customer Zero focus.

AI-powered innovation

The final pillar of this era is innovating with AI to transform the digital experience at Microsoft. By doing all the fundamental work detailed above—security, foundations, and Customer Zero—extremely well, we gain the confidence and earn the trust necessary to embed AI across our full portfolio of services. We do this over three key dimensions: core IT services, employee experiences, and corporate functions.

Core IT services: Transforming and securing our network and infrastructure

We’re focused on using AI to infuse data-driven intelligence into every part of our infrastructure and network operations. This allows us to optimize operations and increase security while simultaneously improving outcomes.

Examples include:

• Network observability and governance: Ensuring data accuracy, eliminating non-compliant hardware and software, and real-time updates
• Securing endpoints: Device management, asset management, and patching
• Zero Trust networking: Isolating device classes and limiting attacker’s movements across the network
• Network access: Azure VPN, identity management, and Secure Access Workstation (SAW) infrastructure security

Learn how we’re transforming our enterprise IT operations at Microsoft.

Core IT services: Tenant management

We manage one of the most complex tenants anywhere. Governance today is a somewhat fragmented experience, with no clear mechanism for IT to safely enable self-service asset creation for sites, Teams, groups, Power Apps, and so on. These unmanaged assets increase the risk of over-sharing sensitive data and compromise the health and security of our IT environment.

In the world of AI, security through obscurity is no longer a viable option. This means data hygiene, permission management, and data protection are essential to providing trustworthy AI tools that don’t overexpose sensitive content, while still providing quality responses.

Read about one way we’re improving security by protecting elevated-privilege accounts at Microsoft.

Core IT services: Support

We’re using generative AI to transform the way our employees interact with our support services. IT issues will be either auto-remediated or resolved remotely and instantly through conversational, personalized, and contextualized solutions, often without a human agent’s intervention.

We’ll accomplish this with a focus on the following:

• User experience: Our employees are using the AI-powered Employee Self-Service Agent to access personalized, accurate, and cost-effective issue resolution. Future goals include implementing a seamless transition to a human agent while the user stays within the agentic Copilot experience.
• Human agent experience: Operational efficiency and automation are being integrated into the Service Operations Workspace. The service includes chat and incident summarization that recommends best next actions and drafts contextual answers to queries.

Find out how we’re transforming IT support at Microsoft with AI and the Employee Self-Service Agent.

Defragmenting the employee experience

The second dimension where we’re implementing our AI vision to make a difference is our employee experience. Our vision is to deliver a unified, connected, and personalized experience where users can access employee data, tools, and insights from one place.

“We see AI as the key to unlocking the full potential of our employees, delivering personalized experiences that empower us to work smarter, faster, and happier—unleashing the innovation and collaboration necessary for our success.”

Vijaya Alaparthi, principal group product manager, Microsoft Digital

One of the key ways we’re doing this is with Microsoft 365 Copilot, which functions as a “UI for AI” across our employee tools and services. An example is our Employee Self-Service Agent, an AI-driven tool based on Copilot that helps employees more efficiently find context-specific answers to their questions using natural language queries.

“We see AI as the key to unlocking the full potential of our employees, delivering personalized experiences that empower us to work smarter, faster, and happier—unleashing the innovation and collaboration necessary for our success,” says Vijaya Alaparthi, a principal group product manager in Microsoft Digital.

To achieve our vision, we’re building a workplace where AI defragments the employee experience by:

• Providing contextual support in the flow of work
• Reducing the number of sites and apps an employee must remember
• Using Microsoft 365 Copilot as the “UI for AI,” making it simple for employees to find information, take action, and even fully automate certain repeatable tasks

Corporate functions growth

Our third major priority in Microsoft Digital is to improve how we support the company’s corporate functions organizations, including legal and real estate and facilities.

“With AI, we have so many new ways to innovate. From optimizing building occupancy, to streamlining commute services, to automating contract and document management, we have incredible potential to make our corporate functions more efficient and impactful.”

Becky West, principal group product manager, Microsoft Digital

This is a particular challenge, as these teams are being asked to do more with less today; Microsoft can no longer afford to grow operational costs at the same rate as in the past.

AI is playing a fundamental role in transforming the business workflows of our corporate functions partners while improving operational efficiency, user productivity, regulatory and corporate compliance, and data-driven decision making. It’s revolutionizing the way they operate by automating repetitive and time-consuming operational tasks.

“With AI, we have so many new ways to innovate,” says Becky West, a principal group product manager in Microsoft Digital. “From optimizing building occupancy, to streamlining commute services, to automating contract and document management, we have incredible potential to make our corporate functions more efficient and impactful.”

Some of the corporate functions taking advantage of AI capabilities and related increased efficiencies include:

• Real estate and facilities: In supporting the technology needs for more than 500 company buildings worldwide, we are poised to use AI and related innovations to implement cost savings in the areas of workspace systems, facilities management, and space management.

Find out how we’re transforming facility operations at Microsoft with AI maps.

• Travel and expense: Our plan is to work for near-elimination of the traditional expense reporting process through AI-based and touchless experiences, driving simplification and productivity gains.

Check out how OneExpense transformed our employee expense reporting.

• Legal: Our vision for integrating AI into Corporate, External, and Legal Affairs (CELA) includes more discoverable legal findings, better corporate document management with the Docufy platform, enhanced engagement with Microsoft Philanthropies, and accelerated support for business-critical functions such as immigration, contracting, and insider trading compliance.

Read how AI is revolutionizing the way we support corporate functions at Microsoft.

Agentic AI: Becoming a Frontier Firm

This era of AI in IT has quickly morphed into a world in which agents are having major impacts across the enterprise. Microsoft Digital plays a central role in helping the company embrace this change and transform into a Frontier Firm: an organization that has deeply embedded AI and agents into its operations, products, and culture

As a Frontier Firm, we go beyond simply adopting AI as a discrete tool or additional technology. We’re actively integrating intelligent systems, rich data platforms, and human knowledge into a unified operating model, where automation, decision making, and innovation combine to spark acceleration at scale. Agentic AI is a core enterprise capability for us, powering everything from employee productivity to customer experiences and strategic decisions.

As Microsoft progresses into this agentic AI future—where autonomous or semi-autonomous AI agents understand context, take actions, and collaborate alongside humans—Microsoft Digital has played a lead role in deploying these capabilities internally. We’ve led the early adoption of tools like Microsoft 365 Copilot, Azure AI services, and custom-built agents that help us automate repetitive tasks, surface insights, and orchestrate workflows across systems while enforcing strict governance policies. Examples include AI-powered agents that assist in IT service management, network monitoring, and enterprise knowledge retrieval, which allow employees to focus on higher-value work and maximize their individual impact.

As AI agents continue to grow in power and functionality and become more deeply integrated into the daily workflows of knowledge professionals, Microsoft IT will maintain our leadership role and operate at the bleeding edge of this technological revolution. 

A catalyst for change and growth

Microsoft’s digital transformation is a story of evolutionary change, resilience, and adaptation across multiple eras of information technology. From our origins as a traditional IT organization to becoming a modern engineering organization focused on driving AI-powered innovation, we in Microsoft Digital remain a catalyst for change within the company and our industry.

With our insights born from customer and employee obsession, we’re committed to streamlining IT operations while prioritizing security, revolutionizing user services, and facilitating corporate functions growth and development. All with the overarching goal of making Microsoft employees everywhere more productive while showing our customers and partners what’s possible as we move forward together into the future of IT.

The post Digitally transforming Microsoft: Our IT journey appeared first on Inside Track.

“The legacy system was highly fragmented, with more than 50 device options, localized processes, data scattered across different systems, manual touchpoints, and limited visibility into order status. This created a lot of friction across the company, as well as higher costs.”

Aniruddha Das, principal PM manager, Microsoft Digital

But as our organization scaled globally, this system became increasingly complex and inefficient.

Across more than 200,000 employees in over 100 countries and regions, our device procurement and inventory management devolved into a patchwork of systems and practices. There was no top-level management, no consistent lifecycle model, and no standardized experience for employees needing to order a new device.

“The legacy system was highly fragmented, with more than 50 device options, localized processes, data scattered across different systems, manual touchpoints, and limited visibility into order status,” says Aniruddha Das, a principal PM manager in Microsoft Digital, the company’s IT organization. “This created a lot of friction across the company, as well as higher costs.”

Working in partnership with Procurement, we built myDevice—a companywide program that centralizes primary-device procurement and lifecycle workflows within a single managed platform. myDevice brought together previously distributed data and processes, giving us reliable visibility into each employee’s primary device and enabling predictable, data-driven refresh planning at scale—along with the operational insights needed to run the end-to-end experience.

“We’ve moved from a fragmented device procurement experience with many touchpoints to one conversation with one agent. Basically, we can tell our employees that their next device is just a conversation away.”

Mukul Singhal, partner engineering manager, Microsoft Digital

Beyond solving some of these basic data challenges, we also needed to address harder, at‑scale problems in the broader procurement experience: helping employees confidently pick the right device without getting overwhelmed by options, aligning eligibility and catalog choices to an employee’s role and location, and more.

With the new Employee Device Information (EDI) agent—affectionately referred to as “Eddie”—our employees will soon be able to use an AI chat interface to quickly explore recommended device options based on their job role and work needs, compare different models, and order a new device with a single click. This streamlined approach will reduce frustration, save time, and cut our costs as we automate a formerly manual process.

“We’ve moved from a fragmented device procurement experience with many touchpoints to one conversation with one agent,” says Mukul Singhal, a partner engineering manager in Microsoft Digital. “Basically, we can tell our employees that their next device is just a conversation away.”

Creating a unified approach with myDevice

The transformation began about three years ago, when our Procurement partners brought their vision for myDevice to us in Microsoft Digital. The stated goal was to standardize employee device procurement, inventory, and lifecycle management across the company.

“Our decentralized process was an industry outlier, and we realized we needed to pivot toward a standardized global process. We wanted to give our employees a single entry point, with consistent guidance across the enterprise.”

Angela Pearson, senior procurement program manager, Microsoft Procurement

Built on top of our internal ServiceNow platform, myDevice brought together previously disconnected workflows into a single consistent experience for employees. It introduced a common catalog of devices, standardized refresh cycles, and centralized ordering and fulfillment processes.

It was evident that such a system was badly needed.

“Our decentralized process was an industry outlier, and we realized we needed to pivot toward a standardized global process,” says Angela Pearson, a senior procurement program manager in Microsoft Procurement. “We wanted to give our employees a single-entry point, with consistent guidance across the enterprise. That was the first big win.”

Purchasing devices for a global workforce the size of Microsoft is not an insignificant budget item. And our fragmented system meant higher costs for the organization.

“The decentralized procurement model allowed teams to move fast, but it also led to inefficiencies,” Das says. “Devices were often purchased at the end of the fiscal cycle, because there was budget, but they didn’t always match real requirements and would sometimes go unused.”

This list maps out the evolution of our device procurement program here at Microsoft:

Legacy state

• Distributed experience: Inconsistent, inefficient processes across countries and business groups
• Zero visibility: Lack of proper asset tracking, accounting, or device management
• High friction and high cost: Significant manual effort required across multiple touchpoints

Foundation: myDevice 1.0

• Unified global experience: One consistent process across all geographies and divisions
• Persona-based: Guided buy for productivity needs based on job role
• Streamlined workflows: Optimized automated processes reducing manual touchpoints
• Enhanced visibility: Improved data management and asset tracking entering the company

Future vision: myDevice 2.0

• Next-gen interface: AI-driven, seamless employee interaction
• Supply chain resiliency: Built to weather disruption
• AI-assisted workflows: Intelligent automation across backend systems and processes

Creating a solution at the scale required for myDevice was a challenge. Relevant data was spread across systems, refresh cycles were inconsistent, and key processes, such as determining which devices should be refreshed, were often manual and time-intensive.

“The data was distributed across multiple systems, and each group was doing their own cycle,” says Amit Raghuwanshi, a principal software engineering manager in Microsoft Digital. “Bringing all the relevant data together was a huge task and big part of the solution.”

Three scenarios for device procurement

Our employees interact with three different workflows that involve obtaining a new primary work device:

• New hire provisioning
• Device refresh (for aging machines)
• Replacement for damaged devices

Each of these core scenarios now follows a standardized process, replacing the ad hoc methods that previously varied by team and geography.

“If a new hire’s device arrives a week after they join the company, that’s not a good experience. We want them to have it on day one.”

Ashok Bagade, principal product manager, Microsoft Digital

Establishing a clear device lifecycle was a big step. For example, Microsoft employees are eligible for a new PC (called a refresh) every four years. Now, with myDevice, roughly 20,000 employees each quarter automatically get a refresh invite, presuming available budget. This takes the burden off IT, admins, and other support staff who were previously tasked with managing this process manually.

The myDevice system is also improving the process for new hire device procurement. The goal has always been for each new Microsoft employee to have a computer waiting for them on their first day, but the gaps in the previous system often made meeting this standard difficult.

“If a new hire’s device arrives a week after they join the company, that’s not a good experience,” says Ashok Bagade, a principal product manager in Microsoft Digital. “We want them to have it on day one. Today, we are only able to do that around 70% of the time. We want to take that rate up to 98%.”

To make that happen, we’re in the process of shifting from a build-to-order device model to an inventory-based model. This means that our suppliers will have a stockpile of our most frequently requested PCs on hand, based on data we have compiled. This approach is designed to reduce the device procurement time from four to six weeks down to less than two weeks, Bagade says.

By centralizing and standardizing device management, myDevice created the foundation for the next phase of transforming this process: intelligence.

Adding AI for a seamless procurement experience

The second part of reimagining our employee device procurement at Microsoft was on the front end, which is the digital interface our people experience when they need to choose their new device.

Even with a unified platform in place, the experience of selecting and ordering the device could be time consuming and disjointed. Employees could browse a list of available computers, but choosing the right one still required research and consultation with colleagues and led to cognitive overwhelm.

“The challenge we saw is that people take a long time to complete their purchase because they don’t know which device to select,” Bagade says. “They get a bunch of options, and they’re not sure which one is right for them.”

So we built EDI (or “Eddie”), an AI-powered agent created with Microsoft Copilot Studio, that makes the process much more efficient. Eddie is transforming a static workflow into a conversational experience.

With Eddie, employees get a guided, conversational buying experience instead of a form-led journey. The agent validates the employee’s eligibility, captures the required inputs, and then offers role-aligned device choices directly in the chat—complete with comparison details and a clear selection path. Once confirmed, the employee submits the request and receives a trackable ServiceNow reference, which reduces follow-ups and helps employees stay informed on the process end-to-end.

“Eddie not only gives a consistent purchasing experience—it also guides you,” Bagade says. “You can ask questions and, through that conversation, narrow down your options.”

The agent is underpinned by a multi-agent architecture where specialized components handle different aspects of the process—catalog browsing, comparison, recommendations, and ordering—coordinated by an orchestration layer.

“With Eddie, the agent validates the information and, in many cases, executes the change. It can literally happen in minutes.”

Aniruddha Das, principal product manager, Microsoft Digital

Reducing complexity and increasing visibility

EDI also addresses operational pain points that extend beyond purchasing. One early challenge focused on correcting primary device records, which is an essential but previously manual process.

“Before we developed this solution, something as simple as correcting an employee’s primary device required multiple tickets and human validation,” Das says. “It could take weeks—and in some cases months—to resolve. With Eddie, the agent validates the information and, in many cases, executes the change. It can literally happen in minutes.”

Another critical improvement is visibility. Previously, once a device request was submitted, it effectively disappeared into a system that offered limited transparency.

“After placing a request, employees had no way to check on the status—it went into a black hole,” Das says. “Now, Eddie can look at the same data that Procurement has and provide answers immediately.”

Looking ahead: From systems to conversations

With myDevice and EDI in place, we’re now focused on the next phase: transforming device management into a fully conversational, intelligent experience. The vision is to move beyond portals entirely and toward an agent-first model where employees interact with systems through natural language.

“Our view for the future is that you won’t need to visit a UI or a system to accomplish a task, you’ll just start talking to an AI agent.”

Amit Raghuwanshi, principal software engineering manager, Microsoft Digital

This includes expanding capabilities such as predictive refresh cycles, deeper personalization, and tighter integration with supply chain and inventory systems. Efforts are also underway to reduce fulfillment times through innovations like centralized inventory management and forecasting.

At the same time, the experience will continue to evolve toward greater simplicity.

“Our view for the future is that you won’t need to visit a UI or a system to accomplish a task, you’ll just start talking to an AI agent,” Raghuwanshi says.

For employees, that shift means less friction, faster decisions, and a more intuitive experience.

“We know how busy our employees are, and we don’t want them spending time on lower-priority tasks. When it’s time to acquire a new device, the experience should be seamless from start to finish, without added complexity or cognitive burden. We’re making steady progress toward that reality.”

Anna Adams, director of hardware programs and operations, Microsoft Procurement

For our organization, it represents a broader transformation—one where AI not only improves existing processes but fundamentally reshapes our workflows and the way that work gets done across all functions.

This means that our employees have one less process that they have to puzzle through and figure out, which allows them to focus on their higher-value work.

“We know how busy our employees are, and we don’t want them spending time on lower-priority tasks,” says Anna Adams, director of hardware programs and operations in Microsoft Procurement. “When it’s time to acquire a new device, the experience should be seamless from start to finish, without added complexity or cognitive burden. We’re making steady progress toward that reality.”

The post Meet ‘Eddie,’ our agent for putting new PCs in the hands of our employees appeared first on Inside Track.