What "AI builder" means on a Head of Design's résumé

My résumé says I’m a Head of Design and an AI builder. The second phrase is easy to write and easy to fake, so I think a recruiter or hiring manager should be able to check it. Here’s what I mean by it, what a working week looks like, and how I’d evaluate the claim if I were on the other side of the table.

Why the phrase is there

"AI builder" isn’t a claim that I’m an engineer, or that I lead an AI-native organization. It means something narrower and more useful: I build with AI tools every day, I ship real work through them, and I’ve redesigned my own design-ops workflow around them. I write and prototype in Cursor and Claude against real components. I direct agents through research, drafting, building and checking. And I hold the parts of the job that shouldn’t be automated.

The test for the phrase is simple. Can the person show you something that shipped, explain how it got made, and point to where a human made the decisive call? If yes, the phrase means something. If the answer is a list of tools, it doesn’t.

A design-ops workflow, not a stunt

The way I run portfolio and design-ops work today is a pipeline with clear stages and one owner per stage:

  1. Research and synthesis. Agents gather and cluster inputs: search data, notes, existing pages, competitive examples. I decide which signals are real.
  2. Drafting. Agents draft copy and structure inside written voice rules. I edit, cut and decide what ships.
  3. Building. Pages and components get built against the site’s existing parts and tokens, not invented from scratch.
  4. Checking. A separate review stage checks the exact version about to ship against written rules: width at standard viewports, metadata, links, banned terms. It fails closed.
  5. Go-live. Nothing publishes without my yes.

None of that is exotic. It’s the same structure a good design team runs, compressed. The difference is speed and the number of loops per day, which makes the rules and the gate more important, not less.

What a real week looks like

A representative week of design-ops work, without the theater:

  • Research. Pull the latest search performance for the portfolio, cluster what’s working and what isn’t, and turn it into a short list of decisions. The useful output isn’t a report; it’s "these three pages need clearer titles, and this topic has no page at all."
  • A page shipped. At Cognichip, where I’m Head of Design, a brand reel needed to go on my public case page. The path was the same as any release: a web-optimized version prepared, a preview built, the review stage checking it at every viewport, then my approval, then live. You can see the result on the Cognichip case study.
  • A gate doing its job. The review stage regularly sends work back. Most catches are small: content escaping the column on one viewport, a missing description, a link that points somewhere it shouldn’t. Each catch that recurs becomes a written rule, so the next build doesn’t make it.

The point isn’t volume. It’s that each stage has an owner, a standard, and a record.

What I delegate, and what I never do

I delegate almost all first drafts, most building, and nearly all checking. Agents are faster and more patient than I am at all three.

I never delegate:

  • The go-live yes. Anything public waits for me. An agent’s confidence is not approval.
  • Taste calls. Which of three good options is the right one, what to cut, when something is "fine" but not good.
  • Claims. Anything that states a fact about my work or a company I’ve worked with gets checked against what’s already public. No invented numbers.
  • Access. Which tool gets which permission, scoped to the minimum.

I wrote about the leadership side of this in The job is unblocking. The short version: delegate almost everything, and own the bottleneck.

What changes for a design team

The workflow above started as my own, but the lessons are team lessons. When a design team adopts agents, three things shift.

Writing becomes the core design artifact. The brief, the state list, the voice rules and the "when not to use" notes are now inputs that machines act on directly. Designers who write clearly get better output. Designers who rely on hallway context get confident nonsense. I coach writing as a design skill now, the same way I used to coach layout.

Review moves earlier and gets more frequent. When drafts are cheap, the scarce resource is attention. A team needs a rhythm for looking at first drafts quickly, killing most of them, and investing only in the survivors. Waiting for polished work before reviewing wastes the speed.

Quality has to be explicit. In a small team, taste spreads by osmosis. With agents in the loop, it has to be written down as rules and checked at a gate, or it erodes one acceptable compromise at a time.

None of that requires a reorganization. It requires a Head of Design who has done the work personally and can show the team where the human calls sit.

The receipts

If you’re evaluating a design leader’s AI claim, ask for receipts, not demos:

  • A decision log. Dated records of what was decided, by whom, and on what evidence.
  • A release history. Each change checked at the exact version that shipped, with a pass or a send-back.
  • Written rules. The voice guide, system rules and "when not to use" notes the agents work from. If those don’t exist, the agents are improvising.
  • Something live. A page, a feature, a flow you can open right now.

How to evaluate the claim in an interview

Questions I’d ask a Head of Design candidate who says they build with AI:

  1. "Walk me through the last thing you shipped with agents. Where did you intervene?" Listen for specific moments, not a tool tour.
  2. "What’s a rule you wrote because an agent kept getting something wrong?" Good answers are boringly concrete.
  3. "What do you never let an agent decide?" If the answer is "nothing," keep looking.
  4. "How do you know the output is good, not just done?" Listen for a quality bar, a review step, and subtraction.
  5. "How would you bring this workflow to a team that doesn’t use it yet?" This is the leadership question. Personal productivity doesn’t scale by itself.

The method underneath all of this, stage by stage, is in AI action stacks, and the research half is in How I research. "AI builder" is just the honest label for a Head of Design who still makes things and has rebuilt the making around agents.

Related When design is the bottleneck: fixing flow between design, product and engineeringKeeping a quality bar in an AI design workflowAI action stacks

See more workGet in touch