Microsoft Reviews Contracts with Microsoft Azure AI Services

At Microsoft, our Finance team is leveraging technology to simplify workstream prioritization and automate qualitative assessments.

The Revenue team within our Finance division is composed of accounting professionals with product, industry, and region-specific expertise. These professionals review and document accounting conclusions on the company’s major contracts with its customers, an activity that involves coordination across a global team and the assessment of numerous legal documents.

As our contract portfolio grew, our time spent on workstream management and completing review documentation increased in parallel. Team leads read through each deal’s document stack to assign contracts to their staff and decide the order of priority. The initial assessment of each deal’s construct and assignment process required lengthy, time-consuming efforts that didn’t help us complete our final documentation. Further, a subset of our reviews resulted in standard conclusions that did not require subject matter experts to apply professional judgement.

There had to be a more efficient way to work through the reviews. With an open mind, a companywide focus on automating for scale, and access to Microsoft Azure’s industry-leading technology, we set out on a journey to find a solution.

At the core, this was a document processing opportunity, and we knew Azure AI technologies could do this super fast and provide a streamlined and consistent process.

—Mathew Abraham, technical program manager, Corporate Accounting

Abraham smiles as he stands in front of a full-size wall screen. — Microsoft Azure AI has significantly sped up and streamlined financial contract reviews, says Mathew Abraham, a technical program manager on the Corporate Accounting team.

Our Revenue team engaged our Intelligent Transformation Finance (ITF) team to design a solution. ITF started by interviewing our subject matter experts with the goal of using their knowledge to automate aspects of contract review function. The outcome was a platform with the ability to read documents and assess the likelihood that a contract would require an accounting assessment. The platform is powered by team member expertise input and machine learning.

Today the team makes daily use of the solution to manage its global contract queue. The tool replaces the manual pre-read by performing a precursory document review and delivers insights that produce time savings and streamlines global coordination and prioritization. The platform is now an integral part of our contract review function’s tool ecosystem.

“At the core, this was a document processing opportunity, and we knew Azure AI technologies could do this super fast and provide a streamlined and consistent process,” says Mathew Abraham, a technical program manager on the Corporate Accounting team who worked on the project.

There were two other significant benefits that the team realized. First, it provided more insights than was expected, so the team needed to figure out which insights to pull forward and look at while doing manual reviews. Second, it was very good at integrating into other systems, which helped the team with automation.

We are a small team and in order to drive big impact at scale we needed to leverage tools that could accelerate our development while being compliant and timeless. That’s why we turned to Azure.

—Dana Borgmann, technology program manager, Intelligent Transformation Finance

“It’s truly amazing what the Azure suite of technologies makes available,” Abraham says.

What started with the age-old time savings value proposition, produced a best-in-class document review solution that addresses business challenges across multiple organizations within Microsoft. A deep dive of the solution architecture best tells the story.

Getting the design right

“We are a small team and in order to drive big impact at scale we needed to leverage tools that could accelerate our development while being compliant and timeless. That’s why we turned to Azure,” says Dana Borgmann, a technology program manager on the Intelligent Transformation Finance team.

Lee and Borgmann stand together in a hallway of a Microsoft building. They are smiling at the camera. — Sam Lee and Dana Borgmann are on the team that helped the Finance team streamline the contract review process. Lee is a senior data scientist on the Intelligent Transformation Finance team and Borgmann is a technology program manager on the Intelligent Transformation Finance team.

The team approached designing a solution with a few key principles in mind.

Firstly, they wanted to leverage as many off-the-shelf products as possible. This would buy both scale and speed, as well as ensure the design could take advantage of ongoing improvements by the Microsoft product teams. Microsoft Azure services are continually being improved and those improvements are implemented automatically to the cloud services. So, if the team based their design on these services, this ensured they could count on them being supported, compliant, scalable, and easily implemented. With a small team, this approach gave access and the benefits of all of Microsoft’s engineering via the Azure services.

Secondly, the solution needed to be continually improving and learning from its own previous experiences. So it was essential that the design enabled continuous improvement. We also needed the ability to demonstrate how accurate the model was for implementation and the life of the solution.

Lastly, the team recognized early on, this technology could be applied to many types of contracts and documents and so, rather than build a one-off solution, it was designed as a platform. Then it could more quickly adapt to new use cases and document types.

Intelligent Transformation Finance team’s design principles and list of Microsoft Azure and Microsoft 365 services and tools leveraged. — Here are the Intelligent Transformation Finance team’s design principles and a list of the Microsoft Azure and Microsoft 365 services and tools they leveraged in building this solution.

Turning to Microsoft Azure Services

With these principles in mind, the team set out to knit the needed Microsoft Azure services together. We started by finding the right solution to read the contracts. The solution had to achieve an elevated level of accuracy, to translate into multiple languages, and to apply these across all types of documents quickly. For this, we turned to Microsoft Azure Cognitive Services.

Microsoft Azure Cognitive Services are cloud-based artificial intelligence (AI) services that help developers build cognitive intelligence into applications without having direct AI, data science skills, or knowledge. They are available through REST APIs and client library SDKs (Software Development Kits) in popular development languages. The services enable developers to easily add cognitive features into their applications with cognitive solutions that can see, hear, speak, and analyze.

We leveraged Microsoft Azure Form Recognizer, Microsoft Azure Optical Character Recognizer (OCR), and Microsoft Azure Translator. These combined services can be quickly configured to diverse types of documents, accurately read both handwritten and typed text, and do so in PDFs, Microsoft Word documents, images, and many other document types while translating them in real time in over 100 languages.

By utilizing Azure Cognitive Services, Applied AI, and Azure Machine Learning, we can focus attention on the custom risk scoring model. A handful of people can deliver significant business impact because we have the full power of Microsoft product engineers and data scientists behind us.

—Sam Lee, senior data scientist, Intelligent Transformation Finance

Once they had read the documents, we then needed to make searching on them easy to enable the revenue accountants to quickly find key words, phrases, or any other terms. The solution needed to find the correct response in all document types and to understand a user’s intent both as a searcher and in the documents themselves. For this, we relied on Microsoft Azure Cognitive Search, which is the only cloud search service with built-in AI capabilities that enriches all types of information to help you identify and explore relevant content at scale.

You can use cognitive skills for vision, language, and speech, or use custom machine learning models to uncover insights from all types of content. The service offers semantic search capability, which uses advanced machine learning techniques to understand user intent and contextually rank the most relevant search results. This enabled us to access the full content of the documents and make all that content searchable and usable in their next design element – determining the likelihood that a contract required an in-depth assessment.

“By utilizing Azure Cognitive Services, Applied AI, and Azure Machine Learning, we can focus attention on the custom risk scoring model,” says Sam Lee, a senior data scientist on the Intelligent Transformation Finance team. “A handful of people can deliver significant business impact because we have the full power of Microsoft product engineers and data scientists behind us.”

Using Microsoft Azure Machine Learning

We relied on Microsoft Azure Machine Learning, which empowers data scientists and developers to build, deploy, and manage high-quality models. It accelerates time to value with industry-leading machine learning operations (ML Ops), open-source interoperability, and integrated tools.

The challenges we are tackling within Microsoft finance are not unique. Every finance team is looking for ways to accelerate innovation and drive more efficient business impact. That is what makes this story so powerful—all the tools we used are available to everyone through Azure.

—David Koscheski, finance director and team lead, Intelligent Transformation Finance team

The platform is designed for responsible AI applications in machine learning. This was chosen because it enabled the team to still move quickly even with a custom model while ensuring complete governance, security, and it integrates directly into an ML Ops methodology.

Monitoring, Power Applications, and Microsoft Azure Database

Now that all the AI and ML needs were covered, ITF had to integrate these into a complete process workflow including a web interface, storage, and ongoing monitoring. For these elements we called upon the Microsoft Power Applications, Microsoft Azure Database services, and Microsoft Azure Monitor. Combined, these enabled us to quickly build out the flow and interfaces and covered all the scaling and infrastructure needs.

High-level technology architecture and document review process flow for our solution. — This is a high-level technology architecture and document review process flow for our solution.

“The challenges we are tackling within Microsoft finance are not unique,” says David Koscheski, leader of the Intelligent Transformation Finance team. “Every finance team is looking for ways to accelerate innovation and drive more efficient business impact. That is what makes this story so powerful—all the tools we used are available to everyone through Azure.”

We are being challenged to do more with less but that does not mean we stop innovating. Investing in repeatable solutions like this to scale efficiencies that are cost effective and highly compliant is critical. With so many document and contract types within our business we have just begun to realize the full potential and savings.

—Gerard Morisseau, senior leader, Finance Applications

All businesses must process documents for various reasons. Our scenario was to address contract reviews, but there are many other document types where this sort of solution and technology can be applied. Our team has converted this internal solution into a platform and, with configuration, it can also be leveraged in various other scenarios. Think of the possibilities in finance, education, insurance, legal, retail, social services, travel and hospitality, utilities and so much more! Now think of the document scenarios in your business. Your company, most likely, has the in-house expertise to build a similar experience.

“We are being challenged to do more with less but that does not mean we stop innovating. Investing in repeatable solutions like this to scale efficiencies that are cost effective and highly compliant is critical,” says Gerard Morisseau, senior leader of our Finance Applications team. “With so many document and contract types within our business we have just begun to realize the full potential and savings.”