TL;DR
The thesis in one screen, one line per section. Two minutes here, twenty for the rest.
- The ArgumentA six-sentence claim, five conclusions, one bet: fund the orchestration layer, not the interface.
- 1The Market$600B+ a year, 82.9M homes, poorly measured. A 12-row table of which services price before a visit today and which don’t. The routine tiers of the hard trades are the wedge.
- 2Twenty YearsDiscovery, then coordination, then digitized operations. Marketplaces proved demand and hit the thin-layer limit (Homejoy). ServiceTitan moved the data to the pro side.
- 3Six FrictionsLead-gen conflict, pricing uncertainty, weak trust signals, expensive labor, uneven adoption, no home context. And how they feed each other.
- 4Why NowConversational AI, digitized contractor ops, agentic payments, costly skilled time, consumer expectation. The mandate replaces the checkout: the credential enforces the money limit, orchestration enforces the scope limit. Sensors come later.
- 5Where Value AccruesFive durable assets. Memory compounds into judgment: two scenarios (an AC tune-up, a water heater) show where the record changes the decision. Operator’s evidence: the routine tier works end to end; the AC-extension miss marks the complex tier’s limit.
- 6The MapNot a stack, a switchboard: modular components, coexisting routes. Uber as precedent and its limit (no underwriting, so Homejoy). Urban Company as proof abroad (54,000 pros, EBITDA-positive, IPO ~103× subscribed). Incumbents, honestly: Angi/Thumbtack, Yelp, ServiceTitan, warranties, PE roll-ups, new entrants. National orchestration leaders are possible; fulfillment stays local.
- 7The Next DecadeA bottom-up market model toward ~$1 trillion by 2036, the 2030 base case, and the 2036 direction: robots change who holds the screwdriver, not who holds the judgment. Five dated, scoreable bets (e.g., by 2032 an AI-native coordinator passes $1B in completed-job GMV with 60%+ repeat demand), and the five places it breaks.
- 8Diligence LensSix questions that separate an operating model from a polished interface.
- What If We’re Right?The hedges come off: a 2031 repair with zero human coordination, and a 2036 refrigerator the house repaired itself because the record knew the relay was at year five. Homes, finally, with memory.
The Argument
Software spent twenty years making it easier to find a plumber. It barely changed what happens next: scoping, pricing, scheduling, paying, and remembering. AI can now handle much of that coordination at near-zero marginal cost.
The interface will commoditize. Durable value sits behind it: pricing confidence, reliable supply, verified outcomes, and the home's memory. The winner will be the system that learns from every completed job.
Five conclusions
1Discovery loses value as coordination gets cheap. The front door commoditizes. The work behind it compounds.
2AI expands pre-visit scoping and pricing, but never to 100%. A quote before inspection is an underwriting decision. The winner is whoever knows the variance best.
3The best supply needs platforms least. A network survives only if it raises what a top professional earns per working hour. Otherwise the take rate is a tax, and pros will route around it.
4Homes have no memory. Service history is scattered across owners, contractors, manufacturers, insurers, and warranties. Whoever gives the home a memory owns its repeat demand, and the judgment that builds on it.
5. Winners pair software coordination with an operating model that survives regulation, service variance, and local density.
The investable bet
The layer to own is orchestration: understand the request, price it, schedule it, dispatch the right professional, pay, and record the outcome.
Start where work is standardized and pricing is predictable, then expand into harder categories as data accumulates. The moat is memory: a system that knows this house makes better decisions for this house.
Start narrow, build local supply density, then expand. Fund the layer, not the interface.