From AI Hype to AI Control: Managing Agents in the Enterprise скачать в хорошем качестве

From AI Hype to AI Control: Managing Agents in the Enterprise

AI agents are moving fast—from demos to hundreds of deployments inside real enterprises. But as adoption accelerates, a new set of problems is emerging: trust, security, and runaway cost. In this Davos conversation, Pavan Krishna (S2) speaks with Krishna Gandi, Founder and CEO of Fiddler AI, about what actually breaks when AI agents hit production. Drawing parallels to earlier platform shifts—from data centers to cloud—they explore why enterprises now need an observability and control layer for AI, just as they once did for infrastructure. The discussion covers hallucinations, prompt-injection attacks, silent failures, cost explosions from recursive LLM calls, and why “non-deterministic software” changes how leaders must think about governance. From insurance and banking to e-commerce and customer service, this conversation makes one thing clear: AI optimism must be paired with operational paranoia. A grounded, executive-level look at AI agents, observability, and what it takes to deploy AI responsibly at scale.