Every module feeds data into a central AI brain. The brain reweights overnight, and pushes sharper intelligence back to the floor. Across 5 live rooftops, that loop drives $21K+ per store per month in net-attributed gross uplift ($180K–$274K/year — methodology at /proof).
Data flows in from every module on the floor. The brain processes it overnight. Updated intelligence flows back out — improving routing, coaching, pricing, and predictions before the first customer walks in.
Six processing stages run automatically every night. By the time the first salesperson logs in, the routing model has already incorporated every deal from the day before.
sync-sims-to-closed-loop.js runs. lender_weight_cache updated. matchLenders() now routes against the last 90-day funded outcome window.Each module is a node in the learning network. It contributes real outcome data. The brain processes it. The module receives sharper intelligence. The loop compounds.
lender_weight_cache rebalance — matchLenders() routes against live outcome history, not static tiers
Routing Model Accuracy vs. Network Size
Each rooftop added to the network contributes real funded deal outcomes — actual approvals, actual declines, actual stip patterns — from its local lender relationships and market conditions. That data feeds the nightly reweight cycle. Every store in the network benefits from every deal closed anywhere in it.
A single-rooftop dealer routing through LouieAuto gets lender weights calibrated against that store's deal history. A dealer in a 10-store group gets weights calibrated against the entire group's outcome data — plus every other network participant in the same lender tier.
Vendors who sell routing software as a one-time configuration can't replicate this. Their weights are set at setup and adjusted manually. Ours compound nightly, automatically, with every deal.
Every dollar of uplift documented below flows from a closed learning loop — the brain ingesting outcomes, reweighting models overnight, and pushing sharper intelligence back to the floor. These aren't projections. They're measured across 5 live rooftops and 1,167+ AI-routed deals.
| Loop | What the Brain Learns | Monthly Dollar Impact |
|---|---|---|
| Lender Routing | Reweights 42 lenders nightly from real funded outcomes — FICO cliffs, stip frequency, funding velocity, buy rate spread — over a rolling 90-day window | +$8,460/mo reserve +$180/deal avg × 47 deals · 91.4% first-look accuracy · fund rate 68% → 83% |
| BDC Scripts | Promotes winning word tracks weekly based on contact-to-show outcomes; optimizes follow-up timing per lead segment; surfaces dead pile by reactivation probability score | +$8,841–$14,735/mo recovered 3–5 dead-pile deals/mo × $2,947 PVR · +31% contact-to-show · 41 min → 52 sec speed-to-lead |
| Service Equity | Cross-references every RO against customer equity positions at time of service visit; fires triggered equity alerts with vehicle + deal recommendations | +$12,400/mo per store Equity alerts on every service RO · service-to-sales conversion lifted to 100% coverage |
| Inventory Pricing | Tightens depreciation curves daily using listed price vs. market median, lead volume by price tier, and days-on-lot patterns across the network | ~$17,920/mo floor plan savings 8-day avg turn reduction × $28/day × 80 units · 47% reduction in aged units across 5 stores |
| F&I Coaching | Surfaces low-penetration F&I managers by deal type and objection pattern; identifies which product pitches underperform for specific customer profiles | +$15,980/mo F&I gross 42% → 57% VSC attach (+15pp) · +$340/deal × 47 deals/mo |
| All Loops Combined | The brain compounds — each month of deal outcomes makes the next month's routing, coaching, and pricing more accurate. No manual re-configuration. No vendor touch required. | $180K–$274K/yr net-attributed per store Gross module sum $63K–$69K/mo before overlap & attribution (50–80%); net-attributed figure from before/after study · methodology at /proof · 5 rooftops live |
The compounding principle: A dealer running LouieAuto for 6 months has a routing model calibrated against their own lender relationships, their own market, their own buyer profiles. Month 7 is more accurate than Month 1 because the brain has been learning from real outcomes the entire time. That is not a feature of the software — it is the engine behind every dollar in the table above.
Walk a real deal through the routing model, watch the lender weights run, and see a morning briefing built from live data. The numbers above came from stores exactly like yours.
Enter the Demo →