Delivering Regulatory Insights 92% Faster Using AWS with Amazon Finance Technology

Each day, Amazon processes millions of transactions, and Amazon Finance Technology (Amazon FinTech) manages the complex financial infrastructure that supports these operations. However, different finance teams faced increasingly complex workloads as the business continued to scale.

Using Amazon Web Services (AWS), Amazon FinTech has transformed some of its most time-consuming processes using artificial intelligence (AI) and machine learning (ML). These automated solutions have streamlined operations, accelerated decision-making, and empowered finance professionals to focus on value-added tasks.

Amazon FinTech provides technology to help Amazon’s Finance & Global Business Services teams support the growth, expansion, and restructuring of Amazon’s businesses.

“We build and maintain the solutions that drive business growth, maintain compliance with financial and tax reporting, and facilitate strategic financial analysis,” says Sujeeth Kumar, senior software development manager at Amazon. “Our mission is to equip Amazon teams with tools that maximize time spent on high-value insights, reduce operational costs, and elevate the customer experience.”

In 2020, Amazon FinTech began to explore AI/ML by helping teams determine tax rates for millions of accounts payable transactions. Using Amazon SageMaker, a service used to build, train, and deploy ML models at scale, the team created a hybrid model that helped automate this process while maintaining accuracy. This solution reduced inference time from 2 seconds to 300 milliseconds and sparked a wave of AI-powered innovations across the organization.

“AI/ML has proven highly effective in automating time-consuming, manual financial tasks such as data reconciliation, compliance checks, and fraud detection,” says Piyush Paritosh, software development manager at Amazon. “We’ve improved efficiency while minimizing the potential for errors.”

Amazon FinTech collaborated with Amazon’s Value-Added Tax (VAT) Policy and Planning team to launch World Wide Watch (WWW), a generative AI–powered scanning solution. WWW helps track and manage VAT regulation updates at Amazon’s scale—across geographies, tax types, and business models. It identifies, analyzes, and prioritizes regulatory updates and uses generative AI to summarize key policy changes, as well as highlight potential impacts and risks to Amazon’s businesses. The solution reduces the average time to gain insights for decision-making by 92 percent, from 26 minutes to just 2 minutes per update. By facilitating faster, more accurate decision-making, the solution also empowers tax teams to focus more on strategy and implementation, along with reducing reliance on external services. WWW is powered by Amazon Bedrock, a solution that helps build and scale generative AI applications with foundation models.

The system uses large language models through Amazon Bedrock to summarize articles and extract key regulatory information. WWW then generates concise summaries that include critical details, such as the dates that these changes go into effect. The solution uses Amazon SageMaker to deploy embedding models, which help capture the meaning and context of regulatory documents. Amazon OpenSearch Service, which facilitates the near real-time search, monitoring, and analysis of business and operational data, combines keyword and semantic analysis to identify and group related regulatory updates.

Using WWW, the VAT Policy and Planning team can gain insights into regulatory updates 92 percent faster, enhancing decision-making processes. “Our tax experts approved 80 percent of the suggested titles, summaries, and impact predictions generated by WWW; the system’s analysis consistently meets their standards,” says Kumar. “Information has been cut from 8,000 words to only 250, and processing time has been reduced to just 16 seconds per 1,000 words. This decrease in cognitive load helps our team stay sharp and focused when assessing regulatory changes.”

For Amazon’s tax, legal, and business teams, contract volumes increase annually. To facilitate the review of these documents, Amazon FinTech created the Contract Review and Interpretation System Powered by AI (CRISP). This solution uses a mix of AWS services to automate contract analysis while maintaining accuracy and compliance standards.

CRISP stores contracts in Amazon Simple Storage Service (Amazon S3), an object storage service, and then uses Amazon OpenSearch Service to process and index the documents using a hierarchical chunking algorithm. Claude 3 Sonnet on Amazon Bedrock analyzes these chunks to extract citations, classify terms, and generate explanations. The entire workflow runs on Amazon Elastic Container Service (Amazon ECS), a fully managed container orchestration service.

“With the launch of CRISP using AWS, we have drastically reduced the time and effort required by our teams to analyze information,” says Paritosh. “The time spent per contract has decreased by 80 percent, which facilitates faster decision cycles. Our teams can review complex contracts with less mental fatigue and maintain high-quality analysis throughout their workday. Additionally, the high human acceptance score of 91.37 percent proves that CRISP is accurate and reliable.”

Using a similar architecture pattern, Amazon FinTech focused on driving efficiencies in its finance operations, too. The team used Amazon Bedrock to build a conversational chatbot that serves as a single, unified interface for securely accessing customer account information and internal documentation. Using AWS services, Amazon FinTech has achieved an accuracy rate of over 90 percent. Further, this solution has freed Amazon analysts from searching through hundreds of lengthy documents for standard operating procedures while helping empower their decision-making.

Looking ahead, Amazon FinTech plans to expand its use of AI and ML across more aspects of financial management. The team envisions implementing AI-driven financial planning and forecasting systems, enhancing risk management capabilities, and developing more sophisticated decision support tools for strategic planning. For example, Amazon FinTech is building an AI-powered research solution that lets finance experts perform analyses across volumes of documents with proof of citations.

“AI will reshape financial management by improving efficiency, accuracy, and adaptability,” says Kumar. “It will empower finance teams to make more data-driven, forward-thinking decisions, enhance risk management, and streamline compliance processes, all while providing employees with an opportunity to invest these time savings in more strategic, innovative tasks.”