Buildsimple revolutionizes IDP using AWS serverless technologies

With over 400 customers in highly regulated industries, such as finance and healthcare, Buildsimple helps companies improve document-processing accuracy and reliability while accelerating critical processes. Its solution facilitates rapid implementation and consistent operations, providing business stakeholders with modern technology that addresses individual use cases. For example, some customers need to incorporate document formats—beyond contracts and invoices—that vary by domain and market.

Furthermore, companies typically require technical experts to install and maintain third-party IDP software as part of a costly 12-month process to get to production. To address these challenges, ISR Information Products, an AWS Partner, launched Buildsimple as a new approach to IDP. “We wanted to build, run, and maintain the software for our customers instead of bringing one-to-one solutions into data centers,” says René Weseler, senior executive manager and founder at Buildsimple.

To automate their core processes, customers use Buildsimple to extract relevant information from unstructured documents. The solution offers four levels of automation. These range from the implementation of AI and inbound-mail solutions with transaction-based processing to the use of corporate knowledge and comprehensive agent-based solutions. Also, Buildsimple can supplement document analysis with additional AI modules, such as anonymization, summarization, content and context analysis, and image fraud detection.

After attending an AWS Summit, Buildsimple decided to build on AWS, using serverless technologies alongside a multifaceted AI approach. The solution relies on AWS Lambda, which empowers organizations to focus solely on code while it handles all infrastructure management, facilitating faster development, improved performance, enhanced security, and cost efficiency. To coordinate multiple AWS Lambda functions into workflows, the company uses AWS Step Functions, a serverless-orchestration service.

The built-in governance, encryption, and access controls of AWS help the company innovate at scale while maintaining security under the AWS Shared Responsibility Model. “The security surrounding AWS services was absolutely major for us at the beginning,” says Weseler. “This gave us a solid foundation on which we were able to easily implement security and compliance.”

Using its own AI logic and architecture, the company runs models seamlessly on Amazon SageMaker AI. This way, Buildsimple can build, train, and deploy ML models for virtually any use case with fully managed infrastructure, tools, and workflows. “The use of Amazon SageMaker AI delivered the operational excellence that we needed from day one,” says Weseler.

For identifying data at scale, the solution uses Amazon Textract, an ML service that automatically extracts printed text, handwriting, layout elements, and data from virtually any document. When a use case requires large language models (LLMs), Buildsimple incorporates Amazon Bedrock, which provides the core capabilities for building production AI in one solution. For example, the company accesses Anthropic’s Claude Sonnet through Amazon Bedrock for document classification. In short, the flexibility of the AI infrastructure gives customers the freedom to choose the right approach for each use case, whether classic ML, deep learning, LLMs, or even agentic AI.

Buildsimple also uses Amazon Bedrock AgentCore, an agentic solution for building, deploying, and operating effective agents securely at scale. By developing AI agents, the company can support automated decision-making that is based on document information, process context, and corporate knowledge.

A confidence score measures the accuracy of each document-related AI decision, and results below a certain threshold receive additional human review. Buildsimple has improved its average data-extraction accuracy from 75 percent to more than 90 percent, with some customers reaching up to 99 percent.

Processing up to 2 million documents per day at peak, the company has scaled to host over 3,500 AI models for its customers. Business stakeholders can create AI solutions in 2 hours and securely train models on real-world documents, such as loan applications and insurance claims, without technical assistance.

“In today’s age of AI, software developers need to implement new technologies quickly and reliably to stay in business in the long run,” says Weseler. “Customers’ expectations are high. Using AWS, we unlocked our biggest success factor: the ability to be fast, reliable, and secure in the market.”