Amazon Web Services launched Amazon Bedrock Agent Core, a set of tools designed to deploy and operate AI agents at scale. Agent Core includes a secure serverless runtime, access to tools and support for open-source frameworks.

In the big picture, AWS is aiming to be the best place to build and run AI agents that can carry out tasks with minimal human involvement. AWS is also looking to give enterprise customers tools that can give them stability in a rapidly changing AI environment.

During a keynote at AWS Summit New York, Swami Sivasubramanian, AWS VP of Agentic AI, laid out the cloud provider's approach to agentic AI. The four pillars to AWS' agentic AI strategy revolve around embracing agility, ensuring security and trust, reliability and scalability and observability.

Those pillars will be critical for enterprises given that AI agents are software systems that feature foundational models, complete tasks, take actions, plan, remember context and learn with minimal oversight. AWS' argument is that the fundamentals of building AI systems are as critical as the near weekly advances in model capabilities.

According to Sivasubramanian, the fundamental frameworks and approaches will matter even more as AI agents scale. There will be billions of AI agents working alongside humans in multiple settings and that scale will bring excitement, complexity and a bevy of concerns.

He said:

"We are focused on making our agentic AI data set accessible to every organization by combining rapidly innovation, with a strong foundation of security, reliability, and operational excellence. Our approach accelerates progress by building on proven principles."

In the end, Sivasubramanian's talk in New York had a lot to do with the balance of innovation and fundamentals as well as foundational approaches that can change models and underlying technologies. To AWS, a strong foundation and approach enable and accelerate innovation rather than constrain it.

There's also a reality check behind AWS' rather practical approach: Enterprise adoption of AI agents will trail the technology advances and vendor marketing speak.

AWS is setting itself up for AI agent production systems where stability matters and models for most use cases are good enough to last a while. As agentic AI becomes more enterprise ready, basics such as identity, authentication and stability matter.

The goal for AWS is to leverage Amazon Bedrock Agent Core and its partner ecosystem to enable enterprises to go from experiments to production with AI agents designed to run mission critical business processes. That progression is what has enterprises nervous, Sivasubramanian said.

Here's a look at Amazon Bedrock Agent Core:

  • Agent Core features a secure serverless runtime with session isolation. The agent runtime provides dedicated compute environments for AI agents with session, service and memory isolation leveraging AWS' Nitro abstraction layer.
  • Access to tools and capabilities so AI agents can execute workflows with the right permissions, context and controls.
  • The use of any model or open source framework.
  • Identity services to manage permissions of an AI agent and authenticate them.
  • Built-in checkpointing and recovery for interruptions.
  • Observability that's built in for internal and third party AI agents.
  • Agent Code Gateway for integration with other agents and various systems.

Early customers in private beta for Amazon Bedrock Agent Core include Autodesk, Cisco and Workday.

Other items from AWS Summit New York include:

Customization for Amazon's Nova models. AWS announced the ability to customize its Nova models for enterprise use cases in SageMaker AI. AWS will provide optimization recipes, model distillation and customization to balance cost and performance. Nova has launched eight models in 6 months.

"Over 10,000 customers are already using Amazon Nova. What really matters is that these models have real world impacts," said Rohit Prasad, SVP and Head Scientist for AGI at Amazon.

Nova will also get customization on-demand pricing for inference.

AI agent availability on AWS Marketplace. Customers will be able to buy AI agents and tools within AWS Marketplace. These agents can be acquired with standardized central billing and license management via AWS.

Sivasubramanian said the aim is to make it easy to deploy agents easily. “Now you can test and run AI agent solutions from a range of vendors, then quickly push the production and scale,” he said.

Updates to Amazon Connect and AWS Transform. Both will get specialized AI agents.

Amazon Kiro, an AI powered developer environment designed to speed up agentic systems from concept to production. Kiro takes specifications and designs and turn it into code.

Recent AWS:

Business Research Themes

Data to Decisions, Future of Work, Next-Generation Customer Experience

Roles

Chief Information Officer

Tags

Amazon, Artificial Intelligence (AI), Large Language Models (LLMs)