Launch sequence

Phase 0 — pick the niche. Use the decision filter directly: do you already have at least one warm relationship in it, through an existing client roster or your own operating history, and is it on the defensible list from Market and regulatory reality (financial services, legal, property management, healthcare revenue-cycle, freight/logistics, CPA firms — or, per The social and marketing angle, a specific marketing/social vertical you already understand)? If yes to both, proceed. If it's only on the defensible list with no warm relationship, deprioritize it until a warm entry exists — cold-starting in an unfamiliar vertical with no relationship is the slow path, covered in Failure modes.

Phase 1 (weeks 1–3) — build the first-offer proof asset. The recommended default first build, given a typical 1–4 week window: an AI intake or qualification agent — the narrowest, fastest, most universally valuable build across most shortlisted verticals (tenant maintenance-request triage and routing for property management; new-client intake and conflict/scope screening for legal or CPA work; carrier-onboarding document collection for freight; lead qualification and content-brief intake for a marketing/social client). Why this over a full RAG knowledge assistant or a voice receptionist as the first build: it has the shortest path to a measurable ROI number (hours saved on triage, response-time reduction), it doesn't require ingesting sensitive proprietary documents on day one (lower data-handling risk for a first build), and it reuses skills most operators with any web or automation background already have (form logic, structured data capture) — meaning less net-new tooling risk inside a short window. Stack: an LLM API with structured-output/tool-calling support, plus an orchestration platform (n8n or similar) for CRM or webhook wiring, plus a simple intake form or an existing phone/email channel. A voice receptionist or a RAG knowledge assistant is a natural expansion sale for phase 2 and beyond, once the first narrow build proves out and buys enough trust to touch more sensitive data or channels.

Phase 2 (weeks 2–5, overlapping phase 1) — warm outreach. Contact every existing client in the target niche; frame it as "I built [proof asset] — want to see it, and want me to build your version?" Begin narrow cold outreach (LinkedIn, niche-specific channels) in parallel, but treat it as secondary to the warm channel per Capital and timeline.

Phase 3 (weeks 4–8) — close and deliver the first paid pilot. Price at $5–15K fixed for the narrow build, scoped explicitly to one bottleneck, explicitly not "AI transformation."

Phase 4 (weeks 8–10) — instrument and document. Track the one metric that actually matters (time saved, response-rate lift, error-rate drop) with client-visible logging from day one — this is both the case-study asset and the audit-trail habit that protects against the overclaiming risk in Market and regulatory reality.

Phase 5 (month 3+) — expand or replicate. Either expand scope with the same client (a retainer conversion — path (c) from Three paths compared), or use the case study to close client #2 in the same niche, where the build is faster because the playbook now exists, and begin offering standalone audits (path (b)) to prospects not yet ready for a full build — a lower-friction top-of-funnel that still monetizes on its own.