AWS offers 23 reusable assets for faster generative AI delivery

A catalog aimed at repeated development work

The catalog, now part of AWS Frontier AI Engineering & Services, includes reusable building blocks for Retrieval Augmented Generation pipelines, document processing, agent applications, evaluation and prompt management. AWS says the assets provide a validated starting point instead of requiring teams to begin with an empty file.

AWS identifies three causes of redundant development: assets are spread across repositories and other internal resources, teams lack reliable signals about quality and production readiness, and reuse is often informal. The company says even AI-generated code still requires production hardening, security reviews, testing at scale and ongoing maintenance.

According to AWS, the catalog's assets are validated for quality and security before publication. They include deployment guidance, adoption tracking and lifecycle maintenance. The company also says the building blocks are tested in production environments, reviewed for security compliance and refined using customer feedback.

DoorDash, Lyft and Ricoh use the assets

AWS cites customer deployments as evidence of the approach. DoorDash used the Contact Center GenAI Agent asset to build a voice-operated self-service generative AI contact center. AWS says the solution handles over 100,000 calls per day with 2.5-second response latency, increased testing capacity by 50x and reached live testing readiness in 2 months. Amazon Bedrock reduced development time by 50 percent, according to AWS.

Lyft used the same asset for agentic AI experiences for riders and drivers. The Contact Center GenAI Agent supports automated question answering through text and voice, using a Retrieval Augmented Generation architecture for multi-turn conversations and context. It integrates with Amazon Connect and Amazon Bedrock Knowledge Bases.

Ricoh used the Accelerated Intelligent Document Processing on AWS asset to create a standardized, multi-tenant system for document classification and extraction. AWS reports a 7-fold throughput increase, from 10,000, 70,000 documents monthly, while engineering hours per deployment fell from 80 hours to under 5. Onboarding declined from 4, 6 weeks to 2, 3 days, and the system can process 1,000 documents in minutes during traffic spikes.

New templates target agentic applications

AWS also highlights two recently launched assets for agentic software. The Fullstack AgentCore Solution Template, or FAST, is a starter repository that deploys a secured React frontend connected to Amazon Bedrock AgentCore. AWS says it can deploy in under 30 minutes using infrastructure as code and includes integrations for AgentCore memory, Gateway, Code Interpreter and Observability.

FAST uses Amazon Cognito for authentication and is designed for AI-assisted development. Its modular design allows teams to replace identity providers, frontend frameworks, hosting solutions or agent frameworks as requirements change.

The Agentic Chatbot Accelerator uses the AWS Cloud Development Kit for infrastructure as code. Built on Strands Agents and Amazon Bedrock AgentCore, it provides configurable patterns for authentication, hosting and scaling, with deployment approaches validated across customer deployments.

What the catalog means for builders

The assets are intended for teams that want to focus engineering effort on customer-specific business logic rather than rebuilding common infrastructure. AWS says the catalog can reduce development from months to weeks, while FAST is designed to shorten full-stack agentic application work from weeks to days.

Users can access the 23 open source assets through the asset catalog and deploy them using the documentation provided with each building block. The examples show that the same components can support different customer models, while the maintained and security-reviewed structure is intended to reduce the verification work required before production.