What 142,318 homes say when they can talk.
The owner demo shows one house. This is the other side: the fleet view a national builder would run warranty, trades and reserves from. Every figure below is generated from a seeded model, not real company data.
Communities with a brain dot size = homes · color = predicted claims / 1,000 homes, next 90 days
Live fleet signals last 24h
Fleet health distribution score 0–100, all homes
Health score blends open alerts, maintenance adherence, sensor anomalies and age-adjusted component risk. Homes below 70 get a proactive service outreach.
What the fleet learned this quarter patterns surfaced automatically
Predicted claims, next 90 days by component · with est. cost
Prediction = component age × fleet failure curve × this home's install conditions (trade, weather exposure, inspection photos). Each bar links to the homes behind it in the full product.
Claims: predicted vs. actual trailing 12 months, fleet-wide
The gap between the lines is the model's error. It has narrowed every quarter as more homes report back.
Highest predicted claim load communities · next 90 days
Trade partner scorecards click a column to sort · 24-month window
Score = defect rate, claim cost, cycle-time and pre-drywall photo QA, normalised against the same trade in the same market. In the full product this feeds bid selection and payment holdbacks.
How a home gets its brain one capture during construction · no new trades, no new hardware
The whole capture rides on work a construction manager already does: the pre-drywall walk, the selections sheet, the closing package. The only new act is pointing a phone at the walls for a few extra minutes.
Capture coverage by community share of homes with a complete digital record
What capture costs the field assumptions, stated plainly
- Extra time at pre-drywall walk≈ 4–6 min / home
- New hardware on siteNone (phone camera)
- New trades or vendorsNone
- Data already in builder systemsPlans, selections, serials, trades
- AI indexing per home< $2 compute, automatic
- Owner onboardingAt closing, with the keys
- Marginal cost per home, all-in≈ $15–25
These are our estimates, not measured figures. The first pilot community would replace every number on this card with a real one.
Warranty accrual forecast monthly, next 24 months · baseline vs. with HomeBrain
Baseline is the reported FY2025 warranty accrual spread monthly with normal seasonality. The HomeBrain line applies the same reduction assumption used in the model (15% at full ramp, reached over two years). Change the assumption there and this chart follows.
Where the avoided cost comes from share of modelled savings
Why this is believable mechanisms, not magic
- Pre-expiration inspections catch claimable items while the trade is still on the hookTrade-paid
- Photo-triaged claims skip the truck roll for ~1 in 3 ticketsLabor
- Maintenance reminders cut avoidable HVAC and water-heater failuresFewer claims
- Trade scorecards shift volume to lower-defect partnersCompounding