Chris Thomas on operational agents, the Claude Agent SDK, security before governance, and what mid-market AI transformation actually looks like when it has to clear procurement. Hour-long episodes, not pull-quotes.
Each of these is a long-form sit-down — usually 45 to 90 minutes. The shows and hosts varied. The thesis didn't.
Where the work gets done, not where it gets demoed.
Chris breaks down the difference between an LLM feature and an operational agent that owns a workflow end-to-end. Why most enterprise AI pilots stall before production, what the Claude Agent SDK changes about that calculus, and the rule of one HITL gate before any outbound side effect.
Ship the rules the agent follows, not just the agent.
Every operational agent at YNDR ships with two artifacts: the handbook the agent reads, and the ops manual the human team reads to supervise it. Chris walks through the template, why it survives model upgrades, and how it becomes the asset the client actually owns after handoff.
EU AI Act and NIST RMF matter — after you defeat prompt injection.
A conversation on why most AI governance programs are theater until the agent itself is hard to compromise. The lethal-trifecta defense pattern, RAG access control, zero-click prompt injection, and the order of operations: harden the agent first, then map to the frameworks the board will ask about.
One operator who advises the board and ships the agent.
Chris on why mid-market PE portfolios are converging on a single fractional Chief AI Officer rather than scattering vendor pitches across every portco. The 90-day plan, the board cadence, and how to write a CAIO mandate that survives a change of operating partner.
Build the platform agents and humans share, not the chatbot.
A long conversation on the architecture YNDR ships into clients: Next.js + Supabase + Claude Agent SDK, with HITL views, audit trail, and the Agent Hub at clientapp.com/hub. Why the agent is a teammate, not a feature, and what that does to the org chart underneath it.
What actually breaks when AI hits a 200-person company.
The unglamorous middle of an AI transformation: the procurement team that has never bought an LLM, the compliance officer who needs a SOC 2 letter that doesn't exist yet, the team lead who's quietly worried the agent replaces them. Chris on how to land transformation in a company that runs on real margins.
If you're booking a guest on any of the below, this is what I will say something useful about — backed by what we shipped this quarter, not what was true in 2023.
I'll come prepared, talk through what's actually in production, and skip the AI-as-magic narrative. Operators, builders, and PE operating partners get priority. Send a show link and your usual format.
