Skip to content

Notes

Short, practical pieces on getting a team to use AI.

  1. Who should own AI at a company too small to hire for it Once the tools are bought, someone has to own the prompts, the connections, the approvals and the training. At 20 to 200 people that's rarely a new hire, so here's what the job covers and who usually fits it.
  2. How to check AI research before you act on it A wrong research answer from an AI tool reads the same as a right one. I have a second tool check the first, and I open any number I'll act on at its source.
  3. Claude, ChatGPT or Copilot for a 20-person company At 20 people the seat prices are nearly identical, so price won't decide it. I compare each plan as of October 2026 and explain why Copilot is my distant third.
  4. Why software built with AI breaks after the demo AI coding tools make a working demo cheap. Here's what breaks it a few months later, and four questions to ask whoever builds your next tool.