AI-derived client profile turned LinkedIn outreach into signed engagement

Project Description:

- Achievements: Built and validated a replicable methodology for AI-collaborative professional writing — evidence-before-framework sequencing, one-question-at-a-time interviewing, a disqualifier list carried as standing context across sessions, treating AI output as roughly 85% complete rather than finished, and a seven-rule client-confidentiality framework — then tested it against my own live consulting practice before ever writing it up. The test produced a real result: an AI-derived ideal-client-profile and disqualifier-list process fed a targeted LinkedIn outreach campaign that turned into a signed client engagement about three months later. The book itself — three of twelve chapters now fully rewritten using this exact interview method, applied to my own 24-year career rather than generic advice — is where the methodology gets documented and packaged for other consultants to actually use.

- Tools: Claude (the one-question-at-a-time chapter interviews and rewrites — the core production method), ChatGPT (parallel research workflows), Perplexity (citation-first fact verification), Wispr Flow (dictation for the raw interview material).

- Methodology: I wanted to write a guide for other consultants on using AI effectively, so I had AI generate a market-validated outline and a first draft. It was competent and well-organized — and it could have been written by and for any consultant anywhere, because it was generic. That's not a book worth publishing under my name. So instead of lightly editing that draft, I'm running each chapter through a structured interview with Claude, one question at a time — not a list of ten questions at once, which people reliably abandon — that interrogates the generic content against what actually happened in my career: an engineering degree, an MBA, a term in the Arizona State Senate as co-chair of the finance committee, a regional-partner role at a national public affairs firm, and founding my own practice. Each chapter gets rewritten from the ground up using only real material — a specific committee vote that came down 4-4, a real outreach sequence that turned a cold connection into a signed engagement, not generic claims. Two disciplines make this repeatable rather than a one-off: I supply raw evidence about my actual business before introducing any framework, and I introduce that framework as a question — "does this fit, and say so plainly if it doesn't" — rather than an instruction to apply it. That sequencing is what let Claude correctly reject a marketing framework I'd planned to adopt, because the evidence showed my actual client acquisition runs on relationships and referrals, not a content funnel. Every specific factual claim — names, dates, figures — gets treated as needing verification before it reaches a reader, which caught real instances of the AI overstating a credential or conflating two different elected bodies in my own biography.

- Learnings: One question at a time is the whole difference between a discovery exercise that gets finished and one that gets abandoned in a browser tab — batching questions feels efficient and kills completion rates. The bigger lesson is what evidence-before-framework actually buys you: an AI that's been given the real facts first will tell you a framework doesn't fit, which is a more useful outcome than an AI that applies whatever framework you hand it and makes it sound like it fits.

- Relevant URLs: Manuscript in progress, not yet published; planned for Amazon KDP.

Attachments: Chapter 1, "The Hours Nobody Buys" (~2,780 words) — the book's opening chapter, as a real sample of the personalized, revised writing: the honest reconstruction of where the hours actually went, the Arizona Senate story, and the "access, not speed" thesis that anchors the rest of the book.