Writing
Notes on building AI that holds up.
Practical writing on assistants, retrieval, agents, and evaluation: what works once real users, real data, and real budgets are involved.
Building AI systems
- AI phone agents that callers don't hateShort answers, fast handoffs, and real tools: what actually gets callers helped.
- Why most RAG systems fail in production (and how to tell early)The failure modes I check first, and the cheap signals that reveal them early.
- Designing AI agents that know when to stopTools, permissions, and loop control so an agent stops, escalates, or retries on purpose.
- A golden set in a week: evaluating LLM features without a research teamA small, honest evaluation set you can build in five days and run on every change.
Strategy & first steps
- Build, buy, or wait: deciding on an AI featureA practical way to decide where AI is worth building, buying, or not yet.
- How to choose your first AI project (without wasting a quarter)Pick a first project that ends in a confident decision, not a stalled pilot.
- AI governance that doesn't slow the team downAn inventory, risk tiers, a one-page policy, and controls built into the product.