For Jeetu Patel, the next constraint on AI adoption is not simply access to computing infrastructure. It is whether people will trust an AI system enough to let it act on their behalf. In his interview with Y Combinator, Cisco’s president and chief product officer frames trusted delegation as a competitive issue, rather than a feature companies can add later.
Patel describes Cisco’s response as an agentic security platform built around AI Defense, a product he says began in the chat era and has been extended to agents. The capabilities he outlines include visibility into models, managing non-human identities, keeping an inventory of agents and testing models for vulnerabilities through algorithmic red teaming.
The distinction is between checking a model before deployment and controlling an agent while it is working. Patel says the platform can enforce guardrails at runtime and intercept an unwanted action. He illustrates the problem with an agent issuing a refund larger than the original purchase: a useful assistant becomes a business risk when its authority is not bounded.
His central argument is that delegating work to AI requires more than a capable model: it also requires knowing what an agent can do and retaining control when it acts.