Agentic AI is becoming a common sight in workplaces everywhere, as more of us get to grips with how the technology can revolutionize workflows across the board.
Amazon Web Services (AWS) has long been a key player in the development of agents and the AI tools behind them, and at the recent AWS Summit New York 2025, announced Amazon Bedrock AgentCore, its new platform to support the next generation of AI.
But just how major a step can agentic AI offer for your business?
The next big thing – that’s already here
“Agentic AI could be as transformational as data networks – it changes how we live our lives, and how we do our work…(and) it’s a world that’s going to get real very quickly,” Deepak Singh, VP, Developer Agents and Experience, AWS, told me at the event.
AWS and a wide number of tech heavyweights are all backing agentic AI – with Swami Sivasubramanian, company VP for Agentic AI, calling the technology “the most impactful change we’ve seen since the dawn of the internet”, but as the underlying foundational models evolve, and frameworks and tooling evolve, building and running AI agents at scale is challenging, especially for companies keen to ensure they don’t fall behind the competition in a fast-paced market.
The new Amazon Bedrock AgentCore platform looks to give developers everything they need to create and deploy advanced AI agents.
Singh highlights that AgentCore is a “big step in realising that vision”, noting, “our goal is to make sure that we have the right services in place that people can connect all of this together, and build and run”
“The goal is making the new agentic AI era a reality requires building all these components,” he says, “our goal at AWS is to be the best place for people to run their agents – just like the way we have helped them run services and applications for the last 20 years.”
“We have 20 years experience of building infrastructure services…and that experience directly comes to AgentCore…we understand what the needs of high-scale production level agents are.”
The next (agentic) generation
Singh notes customers are already building agents on top of Bedrock, so the aim of the new services is to give them extra capabilities to make these agents smarter, safer and more intuitive than ever before.
“The way people build their tools and services is going to get agentic,” he notes, highlighting Slack AI as a good example of building an initial agent internally on top of Bedrock, “the next generation of applications is going to be agentic.”
Given agents have been around for some time now, I ask Singh how AWS makes the technology exciting and relevant for customers who are already up to speed when it comes to all things AI.
“Agents in one way or another have been around for about a year and a half,” he notes, “but the last six to nine months, I’ve seen big changes in how agents work.”
“There are two things that customers want to do,” he adds, “they are looking to make everyone in their organization more effective, allowing them to accomplish more by solving mundane tasks, unblocking things that they couldn’t do before – and then there’s a set of agents that are taking care of things on your behalf, you are just providing high-level instructions and they act as they’re augmenting what you do, or they’re part of your team…(customers are) starting their journey now.”
“The way you have built agents has changed,” Singh adds, with the evolution of foundational models, allowing agents to reason and think, playing a major role in this.
He highlights AWS’ own Strand SDK as a good example, as it drives an agentic loop which allows the model to do a lot of heavy lifting while it provides the instructions, letting the user get to the answer they need with minimal stress.
Connecting agents from third parties is also crucial, he notes, with MCP and A2A protocols still only a few months old, and already showing their potential, allowing people to build entirely new and sophisticated kinds of agents already.
“The most valuable agents are those which span multiple organizations,” Singh had said in a previous session, so I ask him how easy that can be.
“This is part of the reason you need (something like) AgentCore,” he says, “before, a lot of this was a lot of work…the barriers to building, to finding the right tools and agents just fell significantly – which is why we have the belief that this will accelerate the AI universe”
And as innovation in the AI space continues at a rapid pace, Singh notes AgentCore (and AWS) will play a big role going forward.
“AI is changing rapidly – every two weeks there’s a new model”, he laughs, “we know these protocols, frameworks, and models are going to evolve (so) what AgentCore is setting down is a set of core foundational services that will grow along with all the evolution, so the end users can still keep building.”
“We want AWS to be the place where everyone runs enterprise AI agents, and for AgentCore to be the way they do it” Singh concludes, “there is going to be a world that is going to be agentic – it is going to happen within the next few year…and AgentCore is going to be an accelerant to making this happen.”
