Insights

Notes from the desk.

Short-form analysis on what we're seeing in mid-market AI adoption right now. Updated weekly.

Apr 22, 2026

Most "AI strategies" we read are vendor lists in disguise

If your strategy doc reads like a procurement plan, you don't have a strategy. You have a shopping list. Three questions that separate the two.

Apr 15, 2026

Eval suites are the new test pyramid

You wouldn't ship a backend without tests. We're seeing teams ship LLM systems without evals — then act surprised when behavior drifts. A starter pattern.

Apr 8, 2026

The two-pizza pilot

Most stuck pilots have one fixable thing in common: too many people in the loop. Why we cap engagements at four humans, and what we cut to get there.

Apr 1, 2026

Cost guardrails are a feature, not a follow-up

The runaway-spend stories all rhyme. Three guardrails we ship on day one of every LLM build, and the bill that shows what they prevented.