Have you ever walked into a conference session thinking you were done for the week, only to realize that the announcements were coming so quickly that you almost needed your own agent to keep up? That was the atmosphere in Las Vegas, and it was the backdrop for my conversation with Madhu Parthasarathy, the general manager of Agent Core at AWS.
He spent the week at the center of AWS’s wave of agentic AI news, working on the ideas that are already moving from keynotes and demos into the hands of real enterprise teams. Sitting down with him offered a rare moment of clarity amid the noise, and his calm perspective on what really matters helped bring the bigger picture into focus.
Madhu discussed the thinking behind Agent Core and why he believes 2026 will be the year when companies finally start moving from prototypes to production-scale agents. He walked me through the two areas that customers keep coming back to: trust and performance, and why the new policy framework and agent evaluations could remove long-standing barriers to deployment.
His examples were based on real behaviors he sees at large companies, whether it’s internal support workloads, developer productivity, meeting preparation, or customer-facing flows designed to reduce friction between intent and outcome.
We also explored the deeper shift introduced by Nova Forge, including the idea of ​​combining enterprise data with model checkpoints to create domain-specific agents that can operate with greater accuracy and context. Madhu explained why there will never be a one size fits all model and how choice remains central to AWS’s approach to agentic AI.
My guest also reflected on how infrastructure changes, such as Trainium three ultra servers and expanded Nova model families, determine the pace at which companies can experiment, evaluate, and adopt emerging capabilities.
During our conversation, trust came up again and again. Madhu was clear that non-deterministic systems also come with concerns, which is why action limits and guardrails become as important as model quality. He described the excitement he sees among customers who now feel they have workable ways to give agents responsibility without completely handing over the keys.
As he put it, this is the moment when trust starts to grow as the guardrails finally meet the expectations of business leaders.
We concluded with the topic that many people were whispering about all week: modernization. Madhu reflected on AWS Transform, the drive to help organizations move away from legacy architectures much faster than before, and the impact agentic systems will have as they support full stack migrations between Windows environments and custom languages.
Madhu cuts through the noise with a grounded vision of trustworthy autonomy, multi-agent orchestration, policy-driven security, and the shift to agents as true collaborators.
The question now is where you see the greatest opportunities. How can these agent-based systems change your workflows, and what would it take to entrust them with the tasks you never seem to have time for? I would like to hear your opinion.
Useful links
Subscribe to the Tech Talks daily podcast
#AWS #reInvent #revealed #acceleration #Agentic