A dealership command centre, built and independently reviewed in about 70 minutes
A 4x4 yard needed one place to see every car, what it really cost, what it is worth and which ones are sitting too long, usable from a phone on the lot.
The biggest number comes first, with its method underneath. Every figure is read from the build's own git log or a dated usage reading, and nothing here touched anyone's private data.
How fast
The first job started at 15:36 and the reviewed build was finished at 16:46. Discovery, ten builders and eight reviewers ran side by side, and a single lead merged the work with zero merge conflicts. The owner's real stock was pulled from the yard's public listings in the same window.
Fast is nice. What did it cost?
What it cost
No extra spend. The build ran inside flat-rate AI subscriptions that were already paid for. Usage was not metered separately for this build, so no per-build dollar figure is claimed. Each yard's data is kept separate from every other account.
What was built
- Every vehicle in one searchable list: make, VIN or stock number, with filters for status and age. On a phone, the tables turn into cards.
- A cost ledger for each car (purchase, repairs, detailing) with price history and true gross profit, shown as incomplete when a cost is missing instead of guessed.
- Aged-stock alerts at 30, 60 and 90 days and beyond, each with a suggested next step, plus acknowledge and snooze.
- A sale flow that cannot record the same sale twice, even with a double tap or two people at once.
- Reports on margin, turnover and stock age, with CSV import and export.
- A rule-based assistant: ask in plain words, it proposes the change, and nothing happens until the owner confirms.
- Individual logins with owner, manager and read-only roles, a tamper-evident audit log and a tested backup and restore.
Cheap and fast still has to work. How do we know it does?
How it was checked
598 automated tests pass. Then eight separate reviewers, each working from a fresh copy, checked security, the money and aging maths, race conditions, real browser use on desktop and phone, accessibility, the visual design and the acceptance rules. They found 1 critical and 9 high-priority issues, and all ten were fixed before the demo. The aging maths matched an independent calculation with zero mismatches.
Screens from the working app
Overview: active stock, stock cost and listed value at a glance 
Stock alerts: a car flagged at 150 days, with the next step 
Inventory as phone cards: cost, asking price, margin and age 
One vehicle: status, actions and photos 
Reports: sales by date range, aging and capital 
The assistant proposing a price change for the owner to confirm
Would it be okay if we talked about something like this for your business? A free consultation is the first step, and a no is welcome.