Leading design through AI

01A team that looked idle, in the wrong climate

After a reorg, my team’s pipeline was thin. Projects had been reverted or deprioritised, while the wider design org was pivoting hard to AI and MCP-server patterns as the new platform priority.

A team that looks like it has nothing to do, in that climate, is a team headed for a headcount conversation. I also had a new skip-level who knew me only by the artefacts he saw, and a team watching how I’d handle AI in their own craft.

02Three problems, one window

• Output

  Produce a week’s version of what usually takes three or four: a state-of-MCP report, a Q4 direction doc, a pattern-guidance audit, DACIs, posters, kick-offs.

• Positioning

  Turn that output into a visible, credible contribution to the org’s top-priority workstream — before the “nothing to do” story set in.

• Credibility

  Lead the team’s AI adoption for real, when my own track record was leading it from a distance — just as the org shifted to a code-first design expectation I had genuine gaps in.

03Stop prompting. Start managing.

Brief → Draft → Critique → Revise

I stopped writing long docs from scratch and started treating AI as a sparring partner — context, goal, and an explicit ask to push back, rather than prompt-and-wait.

The mechanical layer went to the model: first-draft structure, competitive scanning, audit scaffolding. I kept the evaluative layer for myself — that’s where the actual judgement lives. I briefed it the way I’d brief a direct report. It drafted, I critiqued, it revised.

Direct AI like a direct report. Evaluate like a senior.

04A week of output, shaped like a month’s

• State-of-MCP report

  Mapped the MCP-server workstream across the design org — who owned what, where the gaps were, and where navigation expertise actually had leverage. The anchor artefact of the bundle.

• Q4 direction proposal

  A team direction framed around universal AI navigation patterns — scoped against the org’s top platform priority.

• Pattern-guidance audit

  A full audit of navigation pattern guidance, scaffolded by the model, judged line-by-line by me.

• DACIs & kick-offs

  The decision and kick-off scaffolding for a new Navigation Guidance Uplift stream.

• Two reports shipping

  Visible, scoped work on a named project — so the team had output landing, not just plans.

• Positioning bundle

  The package that went to my skip-level: “here’s where we’re contributing,” with the evidence attached.

The metric gap, named honestly. This is still a qualitative case. The number I owe it is a clean before-and-after: pages produced, time-to-first-draft, artefacts per week. I’d rather flag the gap than make a figure up.

05Artefacts, not asks

I took the bundle to my skip-level as “here’s where we’re contributing” — which lands very differently from “please find us something to do.” Out of it came a named workstream with two reports shipping, scoped against the org’s top AI priority.

And I refused to split “build my portfolio site” from “learn the AI tooling.” This site is the receipt: prototyped in AI-native tools on purpose, to log real hands-on hours in the same class of tools the team was being asked to adopt.

The rule I set myself — lead the team’s AI adoption from experience, not from the sidelines.

06What changed

From “looks idle” to visible contributor in a week

The team went from a thin post-reorg pipeline to named, scoped work on the org’s top AI workstream — two reports shipping instead of drifting.

• A different stance with leadership

  The next skip-level conversation opened with artefacts instead of asks — defensive turned directional.

• A manager producing faster than the team had seen

  The pace shift was visible to the people watching how I’d handle AI in their craft.

• A frame I can teach, not a personal hack

  Direct AI like a direct report, evaluate like a senior — now the model I use to lead designers through AI rather than around it.

• Positioning that reads as legitimate

  The work was real and the expertise genuinely fit. What I manufactured was the framing and the package — and it’s worth saying that plainly.

07What I’d do differently

Two things. First, close the loop faster: I shipped the positioning bundle but didn’t confirm my skip-level had actually read it, so a real win sat unbooked for most of the week.

Second, prove the hands-on claim instead of just making it. Leading adoption “from inside the tools” only counts if the thing ships. My own test: if the portfolio is still a plan and not a partly-shipped thing weeks later, the stance was observation in a practice costume. You’re reading the test result.