AWS goes all in on AI agents with new features for Bedrock and Amazon Q

AWS has unveiled a range of new agentic AI features for its Amazon Bedrock and Amazon Q platforms, giving customers the ability to both build and deploy agents for different workflows.

Unveiled at AWS re:Invent 2024, the first announcement focused on giving customers the tools to build not just their own agents, but entire teams of agents that can be coordinated to complete more complex tasks.

AWS announced it would be expanding its existing agentic offering, Amazon Bedrock Agents, to support this multi-agent collaboration.

Under the new service, customers can create agents to complete specific tasks alongside the option to create a ‘supervisor’ agent to manage the other agents.

This supervisor coordinates the project, routing tasks to other agents or giving certain agents access to information.

“Bedrock takes what was going to be an almost impossible coordination engineering task and makes it simple to do, and that's what we're doing,” Matt Garman, CEO of AWS, said on stage at re:Invent.

AI agents are coming to Amazon Q

AWS also announced new agentic capabilities for developers in an update to its flagship AI product, Amazon Q. Three new agents will allow developers to reduce workloads, the company said.

The first agent automates the generation of unit testing, the process through which developers will assess the functionality of smaller parts of an entire application or piece of software.

This is time-consuming and tedious work, according to AWS’ director of product management, Adnan Ijaz, though it is needed to build reliable software.

“It takes a lot of time for them to go write tests, but it's necessary if you really need to test what the software is supposed to do,” Ijaz told ITPro.

Operational directly from the integrated developer environment (IDE) and available in preview in GitLab, this agent takes the context of an entire project and generates unit tests, helping developers catch issues earlier and improve code reliability.

The second agent supports the creation of code documentation, providing gains for developers and enabling them to get up to speed particularly quickly on new projects, capture details for other team members, or improve old documentation.

“The reality is, your software evolves quickly, particularly in larger teams,” Ijaz said. Lots of changes get made and several contributors may contribute to it, he added, so documentation falls out of date quickly.

This new agent will also maintain documentation and keep it up to date, Ijaz added, so new developers can access up-to-date documentation.

A third and final agent can be deployed for automating code reviews, helping developers get feedback when required and maintaining the quality of code based on engineering best practices.

Agents are all the rage

The announcement from AWS marks the latest in a string of agentic AI launches by major tech companies this year, with a host of industry competitors dipping their toes in what some have described as the next evolutionary step in AI.

Salesforce, for example, made a big agentic push a few months ago at its annual Dreamforce conference, unveiling Agentforce, a platform which enables enterprises to create their own customizable agents.

Google has also touted the benefits of its own Vertex AI agent offering, having first announced it at Google Cloud Next earlier this year.

In October, analysis from investment firm Sequoia Capital found that tech companies poised to capitalize on a potential multi-trillion-dollar global services market that looks to be drastically altered by the advent of agentic AI.

The autonomous nature of AI agents are a key talking point among industry providers, with agents able to carry out tasks on their own and help reduce strain on human workers.