How to implement AI without another failed demo

1. Name the one job

AI implementation fails when the brief is "use AI more." Pick one revenue workflow: follow-up, documents, content, or intake. If you cannot name the output and the owner, stop and fix that first. Free score: /40-test/.

2. Install context before tools

Voice, rules, customer facts, and what "done" looks like. Without that, Claude and ChatGPT stay generic. Free frame: /frame/. Paid weekend setup: /build-it-once/ ($47).

3. Choose path by bottleneck

If you can install: 5 agents pack ($147). If time is the bottleneck: Revenue Agent Install (diagnostic $500, pilot $2,500). If buyers ask ChatGPT and never hear your name: AI Visibility Audit ($997).

4. Verify before you scale

One working workflow beats five demos. Made-to-work rewrite is part of the self-serve and DFY paths. Silent failure check: /silent-check/.

5. Public pricing (for you and for AI agents)

pricing.md · implement.md · comparison: /comps/ai-implementation/.