SaaStr 876: Shipping Enterprise AI Agents with the CPOs of Rubrik, Glean, and Harvey Building AI agents is one thing. Shipping them to enterprise customers who will hold you responsible for every mistake they make is something else entirely. Jason Lemkin sits down with the CPOs of three of the most ambitious enterprise AI companies - Rubrik, Glean, and Harvey - for an unfiltered conversation on what it actually takes to go agentic in production. Not demos, not roadmaps, actual working agents at scale, inside security, legal, and enterprise search. Rubrik's CPO breaks down how they're building Ruby, their agentic interface for cyber recovery - and why determinism matters when your agent is making decisions about someone's core infrastructure. When an agent makes a mistake at Rubrik, the blast radius isn't a bad email. It's a broken recovery workflow. Emrecan Dogan (Glean) shares how Glean has evolved from enterprise search to a full system of intelligence - powering agents that run on behalf of entire departments, not just individual users. Plus: why MCP alone isn't enough, and what offline-processed context gives you that runtime fetch never will. Anique Drumright (Harvey) explains how Harvey deploys legal engineers - ex-lawyers with 8 to 10 years of practice - with every enterprise customer, and why that human expertise in the deployment pod is what makes agentic workflows actually stick in law firms. The group also digs into the question nobody wants to answer: when your agent takes an action and something goes wrong, who's responsible? And what does it mean to build a product when customers are going to use it in ways you never anticipated? If you're building agents, buying agents, or trying to figure out how to monetize AI in a market that's moving faster than any product roadmap can keep up with, this one is worth your time.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.