AI and agents were the focus AWS re:Invent 2025as the cloud giant introduces a slew of new services and platforms that expand capabilities and drive wider usage.
But how realistic can these lofty goals be, given so much hype and speculation surrounding the technology? I've talked to some of the leading minds on Amazon and AWS to learn more.
Empowering with AI
“We want to give every application developer the opportunity to become an agent developer who creates reliable, trustworthy, secure agents that are very easy to build,” boldly states Swami Sivasubramanian, VP of Agent-Based AI at AWS.
“Developers are no longer limited by programming languages and syntax, so the speed of development increases from years to months, from months to weeks.”
Sivasubramanian notes that “every job discipline is basically becoming saturated with AI agents” – and that includes the company's new “edge agents,” which are designed to handle monotonous or time-consuming tasks so a human employee can focus on the most fun work, like writing code, and Kiro, an agent-based vibration coding IDE that Sivasubramanian says has received “an amazing response” from hundreds thousands of users have already registered before its now open general availability.
“The good news is that this is just the beginning,” he says, “in many ways the world of software development will fundamentally change,” noting how the role of the full-stack developer will shift from someone who simply wears many hats to one who can work with fellow agent teammates to create the future—what the company calls “Renaissance developers.”
The internal perspective is another key motivator and source of inspiration, he adds, as the scale of Amazon itself, spanning everything from e-commerce to cloud computing to rockets, means its internal workflows could be a good initial use case.
“The needs are different, but the fundamental problems are the same,” he says. “I really think this will fundamentally change everything,” he concludes. “AgentCore will be the fundamental building block for everyone—what AWS did for the IT industry, Agentcore will do for every agent.”
success of Kiro, released in July 2025is a pleasant surprise for Sivasubramanian, as he notes that both startups and large companies are indeed participating in the AI transformation, and it also bodes well for future AWS product development.
He notes that these new versions are not only intended for today's workers, but also for the next generation, which will likely approach work with artificial intelligence and an AI-first approach.
“We envision a world at AWS where we have a foundation for billions of agents,” he says, “and to me that’s an understatement in itself: we want every employee to have a partner agent that empowers them and solves undifferentiated, heavy-duty, defined problems.”
AI with trust
Thus, the critical question for many users is how to ensure human presence without placing the responsibility solely on the person. Artificial Intelligence Tools.
“Agents will not be able to deliver on their fundamental promise if they are not trustworthy,” says Sivasubramanian, noting that the solution provides a foundation for trusted, verifiable sources rather than simply making automated decisions without proper consideration, including an AWS Agentcore interception tool that can alert a human user to the need for verification and verification.
But customization should still be a focus, he notes, since users will always want different things and operate at different levels of expertise. Technology like new AWS Nova Forgewhich allows organizations to enjoy previously unheard of levels of customization can be a game changer without having to spend big bucks.
Also at AWS re:Invent 2025, I spoke with Rohit Prasad, senior vice president and chief scientist of Amazon's artificial intelligence division, who outlined the “persistent industry challenge” of core models and the “harsh reality” of large-scale applications trying to run them.
“This forces everyone to strive for a model that is an expert in its field, its business, its sources of knowledge, and so on,” says Prasad. “Intelligence is not monolithic… the bulk of the benefit comes in specialized areas – every organization is trying to differentiate their business… and that's where AWS has just demonstrated world class by making technology very accessible.”
And Sivasubramanian agrees, noting that “this concept of open learning is accessible—if you look back at the history of technology, all the major technologies that started out being used only by experts then became simpler and simpler and expanded so that everyone could use them.”
So, looking to the future, it's clear that AWS sees agents as the next big thing in the world of work, with CEO Matt Garman saying the technology will be “bigger than the Internet or the cloud.”
I ask Sivasubramanian what he thinks about this and whether it is a viable prospect.
“Agents are going to fundamentally change the way we live and work, and I truly believe that,” he laughs, “but that means that to achieve that future, you also need to build all the right underlying infrastructure and edge agents—and that’s exactly what we’re doing at AWS.”
“A lot of tasks will be automated, but the core work and what we do and what we're trying to achieve won't disappear,” he says, “we'll be able to do them faster and better, and we'll move on to other tasks.”
“The era of agents will take time,” Prasad says. “I believe that we will live in an Internet of agents who will perform tasks on your behalf, communicate on your behalf… but in order for agents to actually work for you, you must be able to trust them, and trust comes from reliability.”
“If you look back at the history of major changes like the Industrial Revolution or the beginning of the Internet or the cloud… some things did change, but what people had to do didn't change,” Sivasubramanian concludes.
“You'll see unprecedented productivity in what you do, and we'll no longer be limited by our ability to create complex things, we'll only be limited by our creativity—and that's what makes this truly exciting.”
