What are AI approvals?

AI approvals are intelligent decision-making steps built into multistage approval workflows in Copilot Studio. These steps allow AI to evaluate requests using your business rules and criteria, making consistent approve or reject decisions automatically. Unlike traditional rule-based automation, AI approvals can analyze unstructured data, interpret complex documents, and apply nuanced logic – while still allowing human reviewers to oversee and finalize critical decisions.

How do AI approvals work?

AI approvals integrate directly into your existing multistage approvals in Microsoft Copilot Studio (formerly known as advanced approvals). You can easily add AI stages – automated decision points – into your workflows to streamline routine approvals.

AI approvals in Copilot Studio follow a simple three-step process:

1. Define your decision criteria – Write clear instructions that guide the AI, such as “reject expense reports over $500 without manager pre-approval.”
2. Provide relevant inputs – Supply the AI with documents, images, text, or organizational knowledge it needs to make informed decisions.
3. Let AI decide – The AI evaluates each request, applies your rules, and returns an approve or reject decision with a clear explanation for its decision.

And the best part? Because you can combine AI stages with manual, human-driven stages, AI approvals foster true human-AI collaboration – where AI handles routine decisions, and humans stay in control of the final outcome.

Key benefits of AI approvals

AI approvals enable a plethora of benefits. Here are a few:

Empower employees to do more meaningful work

AI approvals reduce the burden of repetitive decision-making by automating routine tasks. This gives employees more time each month to focus on strategic, high-impact work. For example, instead of chasing signatures or verifying line items, finance teams can shift their attention to forecasting, budgeting, and negotiating better contracts.

Save hours with every approval

With AI approvals, what used to take days can now happen in seconds. Requests are processed instantly, even outside of business hours. That means no more waiting for someone to check their inbox; your workflows keep moving, around the clock.

Cut costs with faster approvals

When AI handles approvals instantly, your team avoids delays that lead to missed discounts or late fees. For instance, AI-driven invoice approvals can help finance teams secure early-payment savings and prevent costly penalties – without lifting a finger.

Make fewer error and promote better outcomes

Because AI follows your rules exactly, it reduces the chance of human error in approval workflows. That means fewer corrections, more reliable decisions, and stronger compliance. In the long run, this leads to better financial results, improved decision-making, and protects your organization’s reputation.

A digital transformation specialist summed it perfectly:

As someone deeply involved in digital transformation, I see [AI approvals] as a breakthrough for: faster workflows, fewer errors, better compliance, and more time for meaningful work

What can you use AI approvals for?

AI approvals are flexible and can be applied to many everyday workflows. Here are some common use cases:

• Expense reimbursement approvals: Approve or reject claims automatically when they meet policy thresholds, receipts are valid, and spending aligns with company guidelines.
• Purchase order (PO) approvals: Approve purchase requests when they fall within budget, meet vendor requirements, and match the appropriate authorization level.
• Vendor/supplier approvals: Accept or reject new vendor applications based on qualification criteria, compliance standards, and risk thresholds.
• Invoice processing approvals: Validate and approve invoices when documentation is complete, amounts are accurate, and payment terms are met.
• Document review approvals: Approve contracts, proposals, or reports if they meet formatting standards, include required elements, and are complete.
• Time-off request approvals: Automatically approve or flag time-off requests based on available balances, coverage needs, and scheduling rules.
• Travel authorization approvals: Automatically approve travel requests if they comply with policy, stay within budget, and are booked in advance.

Ready to get started?

AI approvals are available exclusively in agent flows within Microsoft Copilot Studio, giving you the power to automate approvals with intelligent decision-making.

To get started:

1. Familiarize yourself with AI approvals in our public documentation.
2. Open your agent flow in Copilot Studio and select the “Run a multistage approval” action from the AI capabilities section.
3. Configure your AI approval stages by writing clear instructions, uploading relevant documents or data, and defining your approval criteria.
4. Start simple with use cases like expense approvals or document reviews and expand to more complex use cases as you go along.
5. Deploy your workflow, knowing it integrates seamlessly with any human approval stages you include.

We have one question for you: What will you build?

Helpful resources

1. General information about multistage approvals: aka.ms/advancedapprovalsdoc
2. Documentation on AI approvals: aka.ms/aiapprovalsdoc
3. How to write instructions in your AI approvals: aka.ms/intelligentapprovalprompting
4. Multistage approvals (formerly, advanced approvals) blog: aka.ms/advancedapprovalsblog
5. Agent flows blog: Introducing agent flows: Transforming automation with AI-first workflows | Microsoft Copilot Blog

Tell us what you think

Tried AI approvals? We’d love to hear how it went. Use the thumbs up/down feedback tool in Copilot Studio or contact us directly. Your feedback helps shape the future of intelligent automation.

The post Automate decision-making with AI approvals in Microsoft Copilot Studio appeared first on Microsoft Copilot Blog.

What is request for information in Copilot Studio?

Request for information is a new human-in-the-loop action built into Copilot Studio agent flows. This capability enables automated workflows to pause execution and collect input from designated human reviewers. When your automated business process encounters a decision point requiring human expertise, input, or review, the RFI action sends a structured request to the right people via email. Subsequently, it collects their responses and seamlessly integrates that input back into the workflow automation.

Unlike approvals with yes/no decisions only, RFI in Copilot Studio offers flexibility and dynamic functionality. You can request specific inputs with different data types, including text, numbers, dates, email addresses, and yes/no responses. Furthermore, you can customize each input with placeholder text, dropdown options, or multi-select capabilities to ensure you get exactly the information your workflow needs.

How does request for information work in agent flows?

Request for information integrates directly into your Copilot Studio agent flows. Here’s how it works:

• Configure your request – First, define the title, message, and assignees for your information request, along with the specific inputs you need from human reviewers.

• Collect responses – Next, the workflow pauses while designated humans receive an email request and provide the required information through an intuitive form interface.

• Continue processing – Finally, once responses are submitted, the workflow resumes automatically, making the human input available as parameters for subsequent actions.

Request for information use cases

Here’s how RFI transforms real business workflow scenarios:

Financial services compliance automation: During loan applications, workflows can pause to collect additional verification documents from compliance officers. This includes income verification or regulatory approvals, helping to ensure thorough review without manual process monitoring.

Insurance claims processing workflow: When automated claim review identifies missing documentation, RFI can request specific details from field adjusters or assessors – for example, damage assessments, repair estimates, or links to photo evidence – before continuing with automated claim processing.

Legal document review workflow: Contract approval workflows can request specific input from legal counsel on complex clauses or compliance requirements that need professional judgment. This helps to maintain both efficiency and legal rigor.

The beauty of RFI in Copilot Studio lies in its workflow flexibility. You can combine multiple input types, assign requests to multiple reviewers (with first-response processing), and configure fields as optional or required based on your specific business automation needs.

Key benefits of request for information feature

Enhanced workflow quality control: By incorporating human oversight at critical decision points, you help ensure that automated processes maintain high accuracy and appropriate judgment for complex or sensitive business cases.

Reduced manual workflow monitoring: Instead of constant supervision, you can strategically pause automated workflows only when human input is genuinely needed. This reduces overall manual oversight requirements.

Improved business decision making: Human reviewers provide context-aware insights and handle edge cases that may be challenging for automated systems alone in your Copilot Studio workflows.

Scalable workflow review process: Multiple assignees can be designated for requests, providing more coverage and redundancy while maintaining efficient response times in your automation.

Getting started

Request for information is available exclusively in Copilot Studio agent flows. This gives you the power to seamlessly blend human expertise with automated workflow efficiency.

To get started with RFI in your workflow automation:

1. Familiarize yourself with RFI capabilities in our comprehensive Copilot Studio documentation
2. Open your agent flow in Copilot Studio and add the “Request for information” action from the Human in the loop connector
3. Configure your workflow request by defining the title, message, assignees, and required input fields for your automation
4. Deploy confidently, knowing RFI integrates seamlessly with your existing Copilot Studio agent flow actions

We have one question for you: What human expertise will you strategically incorporate into your automated workflows?

Helpful Resources

• Request for information documentation
• Agent flows introduction

Tell us about your experience

Have you tried request for information in your Copilot Studio workflows? We’d love to hear about your automation experience through the feedback tools in Copilot Studio or by contacting us directly. Your insights help us continue improving intelligent workflow automation capabilities that truly serve your business needs.

Ready to transform your workflow automation? Start incorporating request for information into your Copilot Studio agent flows today and discover the power of strategic human-AI collaboration.

The post Introducing request for information in Copilot Studio agent flows appeared first on Microsoft Copilot Blog.

Approvals are foundational to almost every business process. And now, advanced approvals offer a better way to build these crucial approval workflows. “Advanced approvals,” said one customer, a regional IT manager at a global manufacturing firm, “will unlock and unblock the complex processes we need in our organization.”

Key features of advanced approvals

Copilot Studio has launched advanced approvals with two powerful capabilities: multi-stage and conditional approvals (more coming soon!). These are designed to unlock new scenarios and meet the needs of even the most sophisticated workflows.

Multi-stage approvals

With multi-stage approvals in agent flows, you can configure complex approval processes that require multiple levels of sign-offs. Each stage has its own set of stakeholders who make approval decisions, and each stage can be separately configured to meet the specific needs of each level of stakeholders. This feature reflects the real-world approval workflows of organizations.

Conditional approvals

You can also add conditions between stages to account for often-needed flexibility in decision-making. Use conditions so that the agent flow automatically approves or rejects approvals, moves to another stage, or skips stages based on specific criteria. This feature enhances the adaptability and efficiency of the approval process as circumstances change.

What’s next?

This public preview is just the beginning of advanced approval capabilities that bring additional flexibility and control to your teams. Based on customer feedback, the Intelligent Automations team is working on expanding capabilities.

Advanced approvals are available now in public preview, so put them to work in agent flows today and start building smarter, more agile workflows that adhere to the ways your specific organization makes decisions. As always, your feedback is welcome and encouraged, because your insights help shape what’s next. 

Learn more and get started with the Advanced approvals in agent flows page on Microsoft Learn.

The post Build advanced approvals in agent flows  appeared first on Microsoft Copilot Blog.