Amazon Bedrock
The easiest and most secure way to build a generative AI app
GenAI, devtools
Background
Amazon Bedrock is developer platform for building generative AI apps. It provides access to foundation models (FMs) from AI startups and Amazon for building apps on AWS. Agents are a core feature in Bedrock which serve as the foundation for building genAI chatbots. Agents enable chatbots to reason, access tools, and perform complex, multi-step tasks based on natural language instruction - resulting in highly capable and personalized chatbots for companies across domains: healthcare, banking, logistics, and customer service.
What we did
We redesigned Bedrock's Overview and Examples page, which enabled improved discovery of foundation models (FMs), and easier onboarding via templates for agents (AGs) and knowledgebases (KBs). Our redesign of Agents involved simplifying the create UX (reducing from 2 minutes to 5 seconds) and enhancing the performance and relevance of Agent responses with Long-Term Memory and Code Interpreter configurations. We designed novel chat UI patterns, such as code generation analysis and chat to build, which were then implemented across many other AWS services that wanted to integrate GenAI.
Results
Our redesign of Agents boosted engagement by 19% in three months, driving a 52% increase in Agents created on the Bedrock platform. Enhancements to the Bedrock overview page showcased new anchor providers, Anthropic and Meta. The introduction of model onboarding task cards made it easier for customers to grant model access, resulting in a 48% increase in model engagement.
