Next phase
Platform
Where the engineering answer to the brief lives — in the same place as the product it describes, rather than in a deck nobody opens twice. Today it holds one thing: what this prototype actually covers of what you asked for.
Brief coverage, overall
60%
Averaged across 13 sections of your document
Sections shown in depth
8 of 13
Key idea working on screen, not described
Customer-facing modules
85%
3.6 · 3.⅘ · 3.3½ — where the shop's work reaches the owner of the vehicle
Screens in the walkthrough
17
One repair order, seven points of view
Demonstrated in depth8
The key idea in the brief is visible and interactive, not described.
3.6Customer transparency85%
The status a customer sees is published by a signature, not by a field being filled in.
You can watch this work
- Live status timeline in the customer's hands, no login
- Quality-control and road-test gate that physically holds the notification
- Unearned steps are not rendered at all, so no false expectation is set
- “We have not forgotten your vehicle” on a job running into its third week
- Plain-language explanation of a finding, sent only after a named person approves it
- Inspection report with photographs on every graded item, including the green ones
Not in the prototype
- One-click alternative wording — the prototype offers free editing instead
- Live video of the repair, which the brief itself defers
- Real SMS and push delivery — registration alone runs four to six weeks
3.3½Warranty & comebacks85%
The log is not the point. The classification is, because each cause has a different owner.
You can watch this work
- Comeback register against original repair orders, with days elapsed and cost absorbed
- Cause assigned by a person: part failure, workmanship, misdiagnosis, unrelated
- The owner of the fix stated next to every cause
- Detection from a phone call in which the customer never says the word “comeback”
- Unclassified items counted separately — they cannot be managed until sorted
Not in the prototype
- Supplier warranty claim workflow and vendor scorecard
- Technician-level trend over time, which needs real history to mean anything
§4Tiering & modularity80%
Multi-tenant from the first commit, because retrofitting it later is a rewrite.
You can watch this work
- Three shops on one deployment, each on its own plan
- Labour rate, tax rate and bay count per tenant
- Feature matrix mapped to the three tiers in the brief
- Accounting shown as separable from the shop product entirely
Not in the prototype
- Billing, entitlement enforcement and plan upgrades
- Tenant-level data isolation, which is an architecture question rather than a screen
3.⅘AI Aid for technicians85%
Four words from the shop floor become four sentences for the owner, and there is no path from generated to sent.
You can watch this work
- Technician's shorthand turned into a customer-ready explanation
- Reasoning included, so an honest recommendation stops reading like an upsell
- Approval as the only transition out of review — the Send button physically refuses
- Written where the technician already writes, not on a screen of its own
- Prompts for the inspection detail a senior technician would have asked for
- The reviewer's name attached to the wording that was sent
Not in the prototype
- Photo processing — lighting correction and highlighting the part being discussed
- Learning from this shop's own approved wording over time
3.2Accounting reports & analytics75%
Totals, dates and averages on the report itself — and it prints.
You can watch this work
- Monthly financials with totals and averages, in the metric names already in use
- Sales tax by period, sourced to the shop address rather than the customer's
- Utah taxing repair labour handled as a per-line flag, not a rule on the line type
- Which reports print natively today against which print here
- Owner dashboard: car count, ARO, close ratio, effective labour rate, gross profit per hour
- Completed against Posted, and why the two never agree
Not in the prototype
- Projections of expected receipts — the brief asks for these by name
- Report configuration stored per user rather than per browser
- Role-based access enforced rather than illustrated by the role switcher
3.1Statement reconciliation70%
You already capture the invoice number when parts come in. Nobody joins it to the monthly statement. That join is the module.
You can watch this work
- Statement read as a document, deliberately not as an integration
- Three-way cross-reference: part number, price, purchase order or ticket
- All three agree and it is matched; anything else lands in Need to Fix
- Price variance separated from an orphan line billed to no ticket at all
- The join key shown as data the current system already collects at receiving
Not in the prototype
- Single invoices handled separately from a consolidated monthly statement
- Card and online purchases matched against an open PO or ticket
- Document archive with a configurable retention period
- Real optical character recognition — the reading step here is simulated
3.8Real-time inventory80%
Stock is committed when the customer approves, not when the ticket is finally paid.
You can watch this work
- Committed column moving off a customer approval on another screen
- On hand, committed and available as three separate numbers
- Reorder list driven by available stock rather than stock on the shelf
- Receiving as an act that captures the invoice number the reconciliation later joins against
- The drift stated in figures: committed on approval here against nothing deducted today
Not in the prototype
- Barcode scanning and physical stock counts
- Raising a purchase order — receiving one is built, raising one is not
- Whether the second system in use today holds parts or tooling — still an open question
3.5Estimates engine70%
The labour guide starts at 1984. For the work this shop is known for, the recipe comes from its own completed jobs.
You can watch this work
- Estimate assembled from a recipe rather than from memory
- Every line carrying the reason it belongs, in the customer's language
- The writer may remove a line and never has to assemble one from nothing
- Hours drawn from this shop's own history against the quoted figure
- A flagged finding turned into a whole job — parts, machining and hours — on one press
- Every derived line naming the finding it came from
- Draft jobs hidden from the customer and excluded from close ratio
Not in the prototype
- An administrator editing the templates themselves
- Supplier and price selection, with OEM enforced where a job requires it
- Live catalogue integrations — PartsTech, WorldPac, O'Reilly and the rest
- Duplicate suppression against work already done on the vehicle
Partly demonstrated3
The central mechanism works; the surrounding scope is named rather than shown.
3.3AI call intake60%
The story starts on the phone. A summary is drafted, a person confirms it, and only then does a repair order exist.
You can watch this work
- Call summarised into structured fields that carry into the write-up verbatim
- Classified as a repair request or as a general question with no ticket
- “Can I speak to Dave” explicitly not turned into a ticket
- Prompts for what the writer has not collected yet, before the call ends
- A repair order that exists only because a named person confirmed it
- Recording disclosure stated on the screen rather than assumed
Not in the prototype
- Real speech-to-text, and training on this shop's own recorded calls
- Live prompting mid-call rather than after the transcript
- Out-of-hours capture and callback queue
3.9Marketing automation40%
The circle closes at pickup: the invoice, the review request, and the reminder that brings the vehicle back.
You can watch this work
- Follow-up sequence composed at pickup rather than remembered later
- Review link on a branded domain — the reason links get flagged today
- Service reminder scheduled against mileage, not against a calendar guess
- Warranty window opened at pickup, which is what makes a comeback measurable
Not in the prototype
- Deposits and diagnostic fees taken at intake
- Campaigns, segments and the marketing side of the funnel
- Actual delivery, which waits on A2P 10DLC registration
§5Technical foundations35%
Web, multi-tenant, role-aware and English throughout. The rest is architecture, not a screen.
You can watch this work
- One dataset projected through seven roles rather than seven mock-ups
- Responsive from a workshop tablet to a phone
- Interface and documentation in English, per the brief
- Cloud deployment with no local installation
Not in the prototype
- Native mobile applications for staff and for customers
- Every external integration named in the brief
- Authentication, permissions, audit trail and data protection
- Historical retention with a configurable period
Deliberately not built2
Named with a reason. A prototype that claims everything proves nothing.
3.4AI diagnostics for mechanics10%
Not built, and the reason is a licence rather than an algorithm.
You can watch this work
- The seed of it: a recipe that learns from this shop's own completed jobs
Not in the prototype
- Search by diagnostic trouble code against probable causes
- ProDemand and Identifix data, whose licences prohibit derivative works
- A knowledge base built to replace the forums
3.7Customer app & shop marketplace5%
Deliberately out. It is a second product, and it needs the first one to exist.
You can watch this work
- The customer-facing screens work on a phone and need no login
Not in the prototype
- Native iOS and Android application
- Map of participating shops, ratings and availability
- Service history across a household's vehicles
- Mileage-based maintenance recommendations and two-tap booking
- The answer model for customer questions — AI first, or partner shops
Your staging, against where the prototype went
Section 7 of your brief proposes an order and then says the exact sequence is subject to agreement and should be proposed by the company, drawing on its expertise. We took that literally. Below is your order with what the prototype actually reached — the argument for changing it comes with the estimate.
| Your stage | Content | Brief § | In the prototype | Our reading |
|---|---|---|---|---|
| 1 | Parts templates for common jobs | 3.5 | Built | Kept as stage one. The recipe is the cheapest thing that changes a writer's day. |
| 2 | Reconciliation and basic reporting | 3.1 · 3.2 | Built | Also the module that can be sold on its own, which makes it worth building early. |
| 3 | Customer transparency and statuses | 3.6 | Built | We would move this earlier than the brief does — it is the module the customer feels. |
| 4 | Real-time inventory | 3.8 | Built | One step earlier in the deduction is the whole fix, so it costs less than its position suggests. |
| 5 | AI modules — calls, diagnostics, inspections | 3.3 · 3.4 · 3.⅘ | Partly | Calls and inspection aid are here. Diagnostics is a licensing conversation, not a sprint. |
| 6 | Customer app, marketplace, analytics dashboard | 3.7 · 3.2 | Dashboard only | The dashboard is here. The marketplace is a second product and should be scoped as one. |
How the walkthrough is ordered
The product navigation is grouped by role, which is the right axis once you know the system and the wrong one the first time you open it. The walkthrough follows cause instead: nothing appears on a screen before the thing that produces it has happened.
Act 1
Something breaks
A truck that has been in one family since 1998 starts knocking under load. Nothing exists in the system yet — not a customer record for this visit, not a repair order, nothing.
Act 2
What is wrong, and what it costs
The vehicle is on the lift. Everything the shop finds has to reach the owner in language he can act on, and be priced before anyone touches a bolt.
Act 3
Three weeks in the bay
An engine overhaul with machining is not a morning's work. This is the stretch where a customer goes quiet and starts wondering, and where the shop's own numbers drift.
Act 4
Money, and the circle closes
The vehicle goes home, the money is accounted for, and the same phone line that started the story is what catches it if the work comes back.
What lands here next
Solution architecture
Multi-tenant model, data isolation, the integration surface with your current system, the diagnostic databases and the parts vendors — and which of those is a hard dependency rather than a preference.
MVP composition
Which modules make release one, which are deliberately deferred, and the reasoning behind a sequence that differs from the one in the brief.
Work breakdown & estimate
Epics down to stories, effort per module, team composition, and cost ranges tied to scope rather than to a single number.
Roadmap & critical path
A calendar with the third-party review queues marked out, because those are the longest lead times in this project and none of them is ours.