Amazon's New AI Reads Invoices and Learns on the Job

Amazon has introduced a new AI-powered document processing system that could save businesses significant time and money. According to IDC research, document challenges cost organizations nearly $20,000 per information worker annually and reduce productivity by over 21%. Traditional systems require manual setup for each document format and break when layouts change. Amazon's new system uses AI agents that automatically classify, extract, validate, and learn from documents like invoices, purchase orders, and insurance claims.

The system works by using eight specialized AI agents that collaborate. When a new document type arrives, the agents analyze it, generate extraction instructions, validate the results, and learn from any corrections. Over time, it builds a knowledge base of document formats, so similar documents are processed automatically with minimal human intervention. For example, when a new purchase order from a vendor was uploaded, the system initially failed because the vendor's SKU format didn't match the company's internal format. But the AI diagnosed the issue, added a transformation rule, and on the second try, successfully validated all line items in under 4 minutes.

This matters because it can drastically reduce the time employees spend on document-related tasks—up to 30% of their time—and cut error rates from 1-4% to near zero. The system is built on Amazon Bedrock AgentCore, a platform for building and scaling AI agents. It requires deployment on Amazon Web Services, which takes about 15-20 minutes. Once set up, it provides a web interface for uploading documents and reviewing results. The system improves over time through feedback loops and can be evaluated when changing models to ensure quality.

The catch is that it's not a simple plug-and-play solution; it requires some technical setup and is currently only available on AWS. However, for businesses dealing with hundreds of suppliers and varying document formats, the potential savings and efficiency gains are substantial. This could be a game-changer for back-office operations, allowing staff to focus on more strategic work.