Accelerating generative AI delivery with reusable assets

AWS's Generative AI Innovation Center, now part of AWS Frontier AI Engineering & Services, has published a catalog of 23 open source, production-ready generative AI assets that the company says can cut development time from months to weeks. AWS cites customer results including 50x increases in testing capacity and 90 percent reductions in engineering hours, with DoorDash, Lyft and Ricoh named as companies that built solutions on the assets. The assets cover RAG pipelines, document ingestion modules, agent scaffolding templates, evaluation frameworks and prompt management patterns, and AWS attributes redundant rebuilding to scattered repositories, no quality signal and an informal reuse culture.

Artificial Intelligence https://aws.amazon.com/blogs/machine-learning/ Accelerating generative AI delivery with reusable assets If you’re building generative AI solutions, reusable assets can help you stop rebuilding components that already exist: Retrieval Augmented Generation RAG pipelines, document processors, agent scaffolding, and so on. This has a compounding cost: time not spent solving the customer’s actual problem. This post shows you how to access a curated catalog of 23 production-ready assets from the AWS Generative AI Innovation Center, now part of AWS Frontier AI Engineering & Services. These assets can reduce your development time from months to weeks, with evidence from companies like DoorDash https://aws.amazon.com/solutions/case-studies/doordash-bedrock-case-study/ , Lyft http://lyft.com/blog/posts/aws-and-lyft-bringing-agentic-ai-to-life-for-riders-and-drivers , and Ricoh https://aws.amazon.com/blogs/machine-learning/how-ricoh-built-a-scalable-intelligent-document-processing-solution-on-aws/ . Since 2023, the AWS Generative AI Innovation Center has published open source reusable assets. These production-grade building blocks give you a validated foundation to start from instead of a blank file. These assets have powered customer solutions delivering measurable results: 50x increases in testing capacity, 90 percent reductions in engineering hours, and deployment timelines compressed from months to weeks. In this post, you will learn why teams keep rebuilding generative AI components, what makes an asset truly reusable, how companies like DoorDash and Ricoh achieved measurable results by using these assets, and how to access and deploy 23 open source assets from the catalog. For a complete list of all available assets, see the asset catalog at the end of this post. The problem: Why teams keep rebuilding Three root causes drive redundant development across generative AI teams: 1. No single source of truth – Assets scatter across repositories, wikis, and drives. Builders can’t find what exists, so they rebuild from scratch. 2. No quality signal – Without verification of production readiness, security reviews, or maintenance status, builders have no way to trust existing solutions and default to writing their own code. 3. No reuse culture – Sharing remains informal and ungoverned. What one team builds stays with that team, and problems get solved repeatedly. The cost of rebuilding compounds. Each redundant build consumes engineering hours, delays customer delivery, and creates maintenance burden. Even AI-generated code requires production hardening, security reviews, testing at scale, and ongoing maintenance. These assets arrive already validated, security-reviewed, and production-grade, avoiding that verification burden. When teams finally reach production, they’ve spent weeks on undifferentiated heavy lifting, work that should have been reusable from day one. The AWS Frontier AI Engineering & Services resolves this challenge through reusable assets. What a reusable asset actually is A reusable asset isn’t a code snippet or a sample. It’s a production-grade, composable building block that implements a discrete generative AI capability. Reusable assets are validated for quality and security before publication. They include clear deployment guides for discoverability and are instrumented to track adoption across customer engagements. These assets are maintained across their lifecycle. Examples from the catalog include RAG pipeline components, document ingestion modules, agent scaffolding templates, evaluation frameworks, and prompt management patterns. The assets are tested in production environments, reviewed for security compliance, and refined based on real customer feedback. This is the difference between a one-off artifact and a building block you can trust. Real results: Customer success stories The value of reusable assets isn’t theoretical. Here’s what happens when teams build on proven foundations instead of starting from scratch. Contact Center GenAI Agent: DoorDash and Lyft According to the AWS case study, DoorDash https://aws.amazon.com/solutions/case-studies/doordash-bedrock-case-study/ built a fully voice-operated self-service generative AI contact center solution using the Contact Center GenAI Agent http://github.com/aws-samples/contact-center-genai-agent asset. As reported, the solution handles over 100,000 calls per day with 2.5-second response latency, achieving a 50x increase in testing capacity. Using Amazon Bedrock https://aws.amazon.com/bedrock/ reduced development time by 50 percent. The team reached live testing readiness in 2 months. Lyft http://lyft.com/blog/posts/aws-and-lyft-bringing-agentic-ai-to-life-for-riders-and-drivers used the same Contact Center GenAI Agent http://github.com/aws-samples/contact-center-genai-agent asset to bring agentic AI to life for riders and drivers, demonstrating the asset’s versatility across different customer engagement models. The Contact Center GenAI Agent provides automated question answering capabilities for both text and voice channels. It employs a RAG architecture to facilitate multi-turn conversations while maintaining context. The solution integrates with Amazon Connect https://docs.aws.amazon.com/connect/latest/adminguide/what-is-amazon-connect.html and uses Amazon Bedrock Knowledge Bases https://aws.amazon.com/bedrock/knowledge-bases/ , a fully managed RAG capability, for hybrid search and metadata filtering. Learn more: GenAI Intelligent Document Processing: Ricoh Ricoh serves major health insurance payers, managed care organizations, and healthcare providers. These customers process hundreds of thousands of critical documents each month, including insurance claims, grievances, appeals, and clinical records. As Ricoh scaled to serve more healthcare customers, they needed a standardized, multi-tenant document processing approach. That approach had to onboard new implementations without requiring unique development and tuning for each one. Using the Accelerated Intelligent Document Processing on AWS http://github.com/aws-solutions-library-samples/accelerated-intelligent-document-processing-on-aws asset as a foundation, Ricoh built a standardized, multi-tenant solution for automated document classification and extraction. This transformed their document processing from a custom-engineering bottleneck into a scalable, repeatable service. Results: As documented in the published case study, Ricoh https://aws.amazon.com/blogs/machine-learning/how-ricoh-built-a-scalable-intelligent-document-processing-solution-on-aws/ achieved a 7-fold throughput increase, from 10,000–70,000 documents monthly. Engineering hours per deployment dropped by over 90 percent 80 hours to under 5 . Onboarding time fell from 4–6 weeks to 2–3 days. The system now processes 1,000 documents in minutes during traffic spikes, delivering over 1,900 potential annual person-hours in savings. The GenAI IDP Accelerator, introduced in this 2025 blog post https://aws.amazon.com/blogs/machine-learning/accelerate-intelligent-document-processing-with-generative-ai-on-aws/ , is a scalable, serverless solution for automated document processing and information extraction using AWS services. This system combines optical character recognition OCR with generative AI to convert unstructured documents into structured data at scale. The AWS Frontier AI Engineering & Services continuously maintains the assets as well. Learn more: - Blog post: AWS GenAI Intelligent Document Processing IDP Accelerator https://aws.amazon.com/blogs/machine-learning/accelerate-intelligent-document-processing-with-generative-ai-on-aws/ - Blog post: How Ricoh built a scalable intelligent document processing solution on AWS https://aws.amazon.com/blogs/machine-learning/how-ricoh-built-a-scalable-intelligent-document-processing-solution-on-aws/ . - Source Code: accelerated-intelligent-document-processing-on-aws https://github.com/aws-solutions-library-samples/accelerated-intelligent-document-processing-on-aws New assets: Building the future of agentic AI Reusable assets become more valuable as teams shift focus from infrastructure to business logic. Two recently launched assets demonstrate how reusable building blocks can accelerate the development of next-generation agentic applications. Fullstack AgentCore Solution Template FAST The Fullstack AgentCore Solution Template FAST http://github.com/awslabs/fullstack-solution-template-for-agentcore is a starter project repository. With FAST, you can quickly deploy a secured React frontend connected to Amazon Bedrock AgentCore, a platform to build, connect, and optimize agents at scale with any framework or model. In this implementation, FAST uses a React-based frontend with the managed infrastructure of AgentCore to accelerate building full-stack agentic applications from weeks to days. FAST deploys in under 30 minutes using infrastructure as code IaC , with built-in integrations for AgentCore memory, a capability of Amazon Bedrock AgentCore, along with Gateway, Code Interpreter, and Observability. Authentication uses Amazon Cognito, configured with enterprise security best practices, and the template is designed for AI-assisted development with extensive documentation and steering docs. With FAST, you can access a complete reference architecture that shows you how the pieces integrate, with a working full stack application that you can learn from and build upon. With the modular design, you can swap out major components identity providers, frontend frameworks, hosting solutions, or agent frameworks as your requirements evolve. Learn more: - FullStack AgentCore Solution Template http://github.com/awslabs/fullstack-solution-template-for-agentcore - Accelerate agentic application development with a full-stack starter template for Amazon Bedrock AgentCore https://aws.amazon.com/blogs/machine-learning/accelerate-agentic-application-development-with-a-full-stack-starter-template-for-amazon-bedrock-agentcore/ Agentic Chatbot Accelerator The Agentic Chatbot Accelerator http://github.com/aws-samples/sample-agentic-chatbot-accelerator is a web application deployment solution that uses the AWS Cloud Development Kit AWS CDK for infrastructure as code so you can create agentic chatbots. Built on Strands Agents and Amazon Bedrock AgentCore, this solution streamlines the development of agentic chatbot applications. The accelerator provides rapid deployment using AWS CDK, with scalable agent runtime management through Strands Agents and Amazon Bedrock AgentCore. It is configurable and extensible for diverse chatbot use cases, with pre-configured patterns for authentication, hosting, and scaling, validated across customer deployments. Learn more: Walkthrough: Using an asset from the library Let’s walk through a concrete example of selecting and using the Contact Center GenAI Agent http://github.com/aws-samples/contact-center-genai-agent asset from the open source catalog. Prerequisites To follow this walkthrough, you need: - An AWS account with permissions to create AWS CloudFormation stacks, Amazon Connect instances, Amazon Bedrock resources, and Amazon Quick Sight dashboards. - AWS Command Line Interface AWS CLI installed and configured with appropriate credentials. - An Amazon Connect instance or permissions to create one . - Access to Anthropic Claude Sonnet and Anthropic Claude Haiku on Amazon Bedrock. - A Jupyter notebook environment for validation testing. The scenario You’re building a customer support solution that needs to handle both voice and text interactions, answer questions using your company’s knowledge base, and escalate complex issues to human agents. Finding the asset Navigate to Contact Center GenAI Agent http://github.com/aws-samples/contact-center-genai-agent on GitHub. The README provides a clear overview of capabilities, architecture, and deployment steps. The following figure shows the solution architecture for the Contact Center GenAI Agent. Evaluating the asset The asset includes security-reviewed code with AWS CDK infrastructure as code, integration with Amazon Connect for voice and chat, and Amazon Bedrock Knowledge Bases, a fully managed RAG capability, for question answering. An automated testing framework and conversation analytics dashboard are built in. Responsible-AI controls are integral to the architecture. Amazon Bedrock Guardrails enforces production-grade safety controls, including hallucination detection, content filtering, and topic-level boundaries, making sure agents operate within defined safety parameters from day one. The repository shows active maintenance, comprehensive documentation, and links to customer success stories DoorDash, Lyft that validate production readiness. Deploying the solution Deployment follows four steps: 1. Clone the repository and install dependencies. 2. Deploy two AWS CloudFormation templates: one for the Knowledge Base stack and one for the RAG solution stack. 3. Integrate with Amazon Connect using the pre-built contact flows included in the repository. 4. Validate the deployment using the included Jupyter notebooks. The full process takes under an hour. The solution also includes an Amazon Quick Sight dashboard for real-time conversation analytics. For detailed deployment instructions and prerequisites, see the deployment guide https://github.com/aws-samples/contact-center-genai-agent?tab=readme-ov-file deploy-and-test-the-solution . Clean up To avoid ongoing charges, delete the resources created