← Home
Building

I don't just write about AI orchestration. I ship it.

Here is the through line. For 20 years the data steward wrote the rules by hand. AI doesn't retire the steward. It promotes one. The steward becomes the orchestrator: declare the outcome, set the rules the agent runs against, then verify the result. That is the whole bet behind AI for data governance, and writing it down is the easy part. Each app below is me trying to break it in public.

FixOrCall is governance as guardrails. I encode a steward's safety policy, the agent triages a real repair, and an outcome log keeps score on whether it obeyed. That score is trust, measured. Quorum is the memory layer generic chatbots skip. The agents run on a board built from your real business, remember what you committed to, and grade the advice against what actually happened. The AI Call Assistant is policy at the edge. It runs a live customer call against my pricing and territory rules with no human in the loop, in real time, with money on the line. Same steward thesis, three different ways to stress it.

FixOrCall

Live

A repair-triage agent for your home · getfixorcall.com

Snap a photo of a broken appliance, home system, or power tool, describe what's wrong, and get a straight answer to the question that matters first: fix it yourself or call a pro? Every answer spells out the time, cost, and risk. When it's a DIY, you get model-specific steps, the parts you'll likely need, and the single best video for the job. When it's a pro job, it says so plainly. The steps aren't the hard part; LLMs and YouTube already list those. What's missing is the honest, risk-weighted call that errs toward safety, so gas, sealed refrigerant, 240V panels, and in-wall plumbing all route straight to "call a pro."

Built with Claude API (vision + triage) Next.js React Tailwind Supabase Stripe YouTube + Places APIs Vercel

Quorum

Validating

Your standing board of advisors that actually remembers

An affordable, always-on personal board of advisors for solopreneurs and small-business owners who feel decision isolation, can't justify $18K a year for a Vistage seat, and won't build the AI version themselves. Generic GPTs give you a stateless persona-chat. The wedge here is the part they skip: a done-for-you board built from your real business, a memory loop that tracks the commitments you make and scores them next session, and a coach that watches patterns across sessions and names the avoidance. I'm validating demand now with a short survey before taking it further.

Built with Claude API Tally Supabase

AI Call Assistant

Private

Speed-to-lead for operators whose hands are full

An AI voice assistant for owner-operators who lose jobs because they can't get to the phone. Their hands are on the equipment, and most callers won't leave a voicemail. It answers the inbound calls and texts they miss, qualifies the lead, gives a rough "starting at" range, and texts the lead straight back to the operator. Whoever answers first wins the job; this makes sure that's always you. It's built and running privately rather than publicly listed. If it fits a business you run, reach out.

Built with Claude API Twilio ElevenLabs
And in parallel, studying for the AIGP (AI Governance Professional) certification, staying close to where the rules for AI are actually being written.

Building something in this space?

If you're working on AI for the unglamorous data and operations work, or you just want to compare notes, I read everything.