Working briefs on what I'm seeing across engagements. Written to be useful rather than complete.
How a small organisation gets its AI work to compound when everyone already uses a different tool. Covers the architecture in plain terms, the one condition on tool choice, the six habits that make it run, who has to own it, and the five failure modes.
Pieces on responsible AI checks for small teams, and on what actually changes when a foundation starts advising its portfolio on AI.