Agentic AI
,
Artificial Intelligence & Machine Learning
,
Events

AWS’ Sivasubramanian Says Enterprises Need Focused AI Agent Investments

Anna Delaney (annamadeline) •
September 24, 2026    

Swami Sivasubramanian, vice president, agentic AI, AWS

Enterprises need to move beyond endless artificial intelligence experimentation and focus investments on a handful of measurable projects to successfully deploy autonomous agents at scale, said Swami Sivasubramanian, vice president of agentic AI at AWS.

See Also: How Postman Embeds Wiz Risk Data Into Dev Workflows

Organizations often embrace a strategic approach of letting “a thousand flowers bloom,” encouraging teams to experiment broadly with AI without first defining what success looks like. That can leave companies stuck with prototypes and demonstrations that fail to produce measurable returns. Sivasubramanian said business and technology leaders should instead agree on metrics tied to customer benefits, sales, revenue or productivity, then concentrate resources on a few projects to build shared infrastructure that other teams can reuse.

“Once you define it, then pick two to three projects that the entire leadership commits behind, and then start investing in what I call ‘the boring infrastructure’ part,” Sivasubramanian said. “So everybody doesn’t have to build infrastructure, but then they end up actually effectively becoming compounding momentum.”

In this video interview at the HumanX Conference in Amsterdam, Sivasubramanian also discussed:

Building trust boundaries that help organizations gradually give AI agents greater autonomy;
Keeping humans involved in high-stakes or irreversible decisions made by AI agents;
Creating “compounding momentum” as agents gain context, improve accuracy and become more deeply integrated with enterprise workflows.

Sivasubramanian, vice president of agentic AI at AWS, leads AWS’ agentic AI portfolio, overseeing the development of AI agents that can autonomously investigate, reason and take action to solve complex business problems. His team has launched numerous agentic capabilities that are reshaping the AI landscape, including Amazon Quick Suite, Kiro, Amazon Bedrock AgentCore, DevOps Agent, Security Agent and Amazon Transform.