What I built. What changed. What the evidence says.
Case notes, system decisions, operating doctrine, and honest post-build analysis for teams hiring an AI builder, commissioning a system, or deciding how to run the AI layer.
Three audiences. One evidence standard.
How I build AI systems that survive the handoff
The six-part operating model I use to turn an AI demo into a system a real team can understand, control, verify, and keep running.
October 10, 2026
4 min read
The archive
Client proof
What a live client proof record should show—and what it should not claim
A dated, privacy-safe look at The Business Lyfe tenant and the difference between proving a workflow and pretending an early sample is scale.
Agentic discovery
CoverIntent: building for discovery when the buyer is an agent
Why agentic discovery needs evidence-backed capability records, explainable intent matching, and a human path—not a new layer of SEO slogans.
Clinic systems
Building MedSpañol / Medbits around clinic time—not AI theater
How a bilingual clinic-system build connects guidance, inventory, store operations, staff workflows, and human clinical boundaries without claiming outcomes before measurement.
AI office manager
What an AI office manager should actually do for a contractor
A practical evaluation framework for separating a chatbot from a governed AI front office with clear authority, escalation, and proof.
Missed calls
The missed-call recovery loop every home service company needs
A practical playbook for turning missed calls into owner-visible follow-up, booked work, and proof without pretending every call is recovered revenue.
Money truth
Why recovered revenue needs payment proof, not optimistic dashboards
How to separate found money, in-play work, booked work, and truly recovered cash when AI is helping collect or recover revenue.
