From the practice.

Writing on AI readiness, operational change, and the small moves that compound. New posts twice a week.

• 16 June 2026

  When the review comes back with fourteen comments

  An approval problem is usually an undocumented workflow in disguise. Define the flow and the tools have something to slot into.

• 15 June 2026

  Readiness is not having the latest model

  Anthropic put its most capable model into public hands on Tuesday and pulled it by Friday. For us it changed nothing, and that is the point.

• 11 June 2026

  Start with the brief you write every week

  The most practical way into AI adoption is the creative brief, because the hard part was never the tool.

• 9 June 2026

  Governance is more practical than you expect

  A policy sitting in a shared drive is not governance. The working version answers the questions your team is already asking.

• 4 June 2026

  Capable islands, no bridges

  Tools get adopted long before the workflow around them does. The result is a lot of capable islands and no bridges between them.

• 2 June 2026

  The model is the smaller decision

  Pick a model and move on. The environment around it is what actually decides whether AI works.

• 29 May 2026

  What to decide before you let an AI agent loose

  Gartner says 40 percent of agents will be pulled by 2027. The failures are predictable, which means they're preventable.

• 27 May 2026

  Where to start building context for an AI system

  Four categories worth getting in order. Three to prepare in advance, one to build as you go.

• 22 May 2026

  Which way are we tipping?

  Strategy ambition has run ahead of operational capability. The gap is the opportunity.

• 20 May 2026

  Most AI adoption is going better in meetings than in practice

  The quiet resistance to AI isn't sabotage. It's inertia, and the fix isn't more training.

• 14 May 2026

  Here's your plan to commission an AI rollout

  Eight steps to get right before you start. Most of them are unglamorous, which is the point.

• 13 May 2026

  When an AI programme drifts, the problem is often mandate

  Three common patterns of mandate failure, and the four levers that fix them.

• 7 May 2026

  What I'm seeing inside agencies and brands trying to move on AI

  Most AI adoption isn't a tooling problem. It's a readiness problem.