“How you do anything is how you do everything.”

One candidate, several kinds of buyers

Twenty years covers competitive intelligence, measurement, analytics products and AI. A recruiter skimming for an insights lead and a hiring manager hiring a product marketer need different evidence from the same career. A generic résumé undersells the fit. An AI-written application oversells it, and hiring teams can tell.

That is a marketing problem I know well, so I treated it like one.

Segment, then personalize

I defined four segments, each with its own résumé, homepage order and lead proof. Then I scored every role on skill fit, culture and pace, pay and location, and set red flags for always-on, frontline performance roles.

The highest-priority employers get a page of their own, shared only with them: the same facts, ordered around what that team needs.

Facts are locked. The prose is free.

Every claim in every application traces to my own records. Every packet passes through the same five stages.

1. Facts

   355 confirmed rows from my own records. No row, no claim.

2. Rules

   One rulebook. A letter runs 250 to 310 words, tells two stories at most and never ends on a gap.

3. Agent

   Claude finds and scores roles, assembles each résumé from my approved library and drafts the rest for me to review and rewrite.

4. Checks

   A claims check and a packet audit must pass. A second AI reviewer reads major changes cold.

5. Me

   I approve every role and send everything myself. The agent never contacts anyone.

“Be sure you’re right, then go ahead.”

My dad’s rule, and the reason the system works. The facts and checks are the first half. Speed is the second: once a role is approved, the packet comes together fast, which matters because my rule is to apply within 72 hours of a posting.

Client data stays behind the door

My best work belongs partly to my clients, and the roles I want involve customer data and AI. So this site is built the way I’d want a client’s data handled.

• Two tiers. Public pages carry no client results. The full case studies sit behind a one-time email code, for people evaluating me for a role.
• Access by relationship. When I’m in process with an employer, its email domain is approved, and I remove it when the process ends.
• Kept out of search and third parties. Case pages are excluded from search engines, fonts are served from this site so no visitor data goes elsewhere, and people from client work appear by title only.

The same judgment I’d bring to your team

• AI does the format and the logistics. Every fact, story and decision is mine.
• The guardrails are the product. An agent is only as trustworthy as the rules and checks around it. It’s the pattern I used for The Brain at Innocean: governed data, written rules and people making the call.
• I learned what I needed: agent instructions, a static site with access control and a serverless form on Cloudflare, typography and design systems. I directed the build; the agents wrote most of the code.

Tools: Claude (agent, code, design), a second model for independent review and imagery, Cloudflare Pages, Access and Workers.