Field notes on making agents work in the real world.
I'm starting to publish what I've learned building agentic systems inside mission-critical airline IT — on reliability, security, and the economics of going agent-native. Here's what's on the way. Follow along on LinkedIn while these land.
Beyond the demo: what production-grade agents actually require
Evals, knowledge graphs, and grounding — the unglamorous work that decides whether an agent ships or stays a slide.
Securing agents that act with real permissions
As agents gain tool access, the security surface changes. A Certified Ethical Hacker's view on governing autonomous systems.
The economics of an agent-native SDLC
TCO modeling and the business case for moving an engineering org off a manual lifecycle — without hand-waving.
The agent governance gap: who is accountable when an autonomous system acts?
Audit trails, human-in-the-loop, and the accountability model regulated enterprises need before they let an agent touch production.
Why regulated enterprises stall on AI — and the pattern that unsticks them
Procurement, risk, and compliance kill most enterprise AI before it ships. The adoption path that gets past the pilot without cutting corners.
Grounding agents in enterprise knowledge: why retrieval alone isn't enough
Knowledge graphs, governed context, and the retrieval architecture that keeps agents accurate against a company's real, messy source of truth.
In the meantime, I post shorter notes on LinkedIn ↗ — follow to catch these essays as they publish.