Amazon announced the release of agents for Amazon Bedrock, a new feature that allows developers to quickly create fully managed agents. By performing API calls to enterprise systems, agents for Amazon Bedrock speed up the release of generative AI applications that can manage and carry out activities. Agents go beyond Foundation Models to comprehend user requests, decompose difficult activities into manageable steps, carry on a discussion to gather further details, and take appropriate actions to complete the request.
Agents for Amazon Bedrock streamline the prompt engineering and orchestration of user-requested tasks. Once set up, these agents autonomously construct prompts and enhance them securely with company-specific data, delivering natural language responses to users. These advanced agents possess the ability to deduce the necessary actions for processing user requests automatically. They break down tasks into multiple steps, orchestrate sequences of API calls and data lookups, and retain memory to effectively fulfill actions on behalf of the user.
In Amazon Bedrock, fully managed agents eliminate the requirement for infrastructure management or provisioning. Without writing any new code, users may easily take advantage of complete monitoring, encryption, user rights, and control of API invocation support. The Bedrock console or SDK makes it simple for developers to upload the API schema. The agent then successfully coordinates tasks with FMs and uses AWS Lambda functions to efficiently carry out API calls.
Generative AI has captured our imaginations for its ability to create images and videos and even generate code, and I believe it will transform every application, industry and business, said Swami Sivasubramanian, VP of databases, analytics and ML at AWS.
Amazon has developed the AI models powering agents internally, but they have also added models from third-party companies like Anthropic, Stability AI, and Cohere to the Bedrock library. Agents securely connect a foundation model to the right data source through a simple API. The agent can also be used to help automatically convert data into machine-readable format.
Several other tech companies are actively developing agents as well. Meta expressed the potential of bringing AI agents to billions of people. Additionally, OpenAI discussed the topic of AI agents and pondered the best approach to their creation.