Ventures
What the method actually produced.
Each of these began as a structural diagnosis rather than a product idea. In every case the architecture is what made the business possible — and in one case it is the reason no production backend was ever paid for.
FohBoh.ai
The problem
A restaurant group runs point-of-sale, inventory, labor, payroll and accounting on separate systems, each producing its own version of “sales” and “cost.” Nobody can say which number is authoritative — and AI trained on those numbers inherits the contradiction rather than resolving it.
What was architected
The Metrics Governance Engine — a deterministic certification layer between operational systems and every system of record. Data is sealed on arrival, normalized, reconciled across authoritative sources, scored against trust gates, and issued as a certified operational fact rather than a raw number.
What changed
Certification happens before data reaches the system of record, so neither the source system nor the destination can influence the outcome. Sentry™ and Cortex™ are the first applications built on that foundation — proof that the engine works, not the reason it exists.
fohboh.ai ↗BailiwickQuikFix
The problem
On-demand home and commercial repair is a large, fragmented market with no dependable promise attached to it. Response times are estimates, dispatch is manual, and margin leaks everywhere between the customer and the tradesperson.
What was architected
Three applications, a licensed governance engine, and a 60-minute service commitment set by drive-time geofence rather than optimism — combined with voice and AI into an intelligence stack no competitor has assembled.
What changed
No production backend was ever paid for. The venture was architected, modeled, governed and packaged — fifteen documents from pitch deck to engineering handoff, and three viable exit paths — before the expensive part began. That sequencing is the whole argument.
StarBar · SnapCount™
The problem
Periodic inventory is still done on a clipboard and typed in afterward. It is slow, error-prone, and because it is unpleasant it gets done late or approximated — which corrupts every cost metric downstream of it.
What was architected
Per-shift, voice-driven counting against a locked watchlist — one item at a time, spoken, with tenthing for partial bottles. Mid-shift breakage and transfers are captured with a photo, so the variance math stays clean and custody is never inferred.
What changed
The count stops being an estimate. Certification is deterministic — named opening count, named closing count, both inside the shift window, variance within threshold, trust score at or above 85, and no open exception. All six, or it is not certified.
starbar.ai ↗What we take on
Five conditions. All of them, or we decline.
- The problem is structural. Not an irritation, and not a feature somebody's incumbent forgot to ship. Something the current arrangement of an industry cannot fix without changing shape.
- Somebody is already paying for it. In money, in labor, or in losses they have stopped noticing. A problem nobody is currently funding is a hobby.
- Architecture is what changes the value. If the outcome turns on capital, timing or salesmanship rather than on how the thing is designed, we are the wrong partner and will say so.
- The decision-maker is available. One person who can decide, in the room. Not a committee tour, and not a proxy relaying answers.
- The evidence can be gathered. There is a reachable population of people with the problem who will tell us the truth by what they do.
Investment, ownership and portfolio-level relationships sit with the parent company rather than with the Studio.
See the full Bailiwick portfolio →