Case Study · AUTOMOTIVE AI · DENMARK

AutoIQ: An AI product for mechanics, not an AI feature

A bilingual diagnostic decision-support workspace from licence plate to repair, built for professionals.

The story

What AutoIQ needed, and what we built.

A mechanic diagnosing a fault juggles vehicle registries, OBD codes, forums, videos, service bulletins, and a billing system, switching tabs the whole way from symptom to fix. AutoIQ set out to replace that with one focused workspace, a decision-support layer for professionals rather than a replacement for the mechanic's judgement or their diagnostic equipment. The hard part was never “add AI”. A single confident answer from a model is worse than useless to a workshop. So KUMO built the system around the model instead: reliable vehicle context, structured and probabilistic diagnosis, a refinement loop that asks the right follow-up, repair guidance specific enough to act on, and the subscription, billing, and admin layer needed to run it as a real product. We designed and built Version 1 end to end, in Danish and English, for Danish workshops, and it is live in production.

What we delivered

Six areas of production work.

Structured, uncertainty-aware diagnosis

The engine returns schema-constrained results: ranked probable causes, each with a probability, a plain explanation, an urgency signal, and an overall confidence, instead of one overconfident answer. When confidence is low, it asks rather than guesses.

A refinement loop that gets sharper

Targeted follow-up questions narrow the diagnosis, skip causes the mechanic has ruled out, avoid repeating themselves, and keep working even after every initial cause is eliminated, so the tool stays useful on the hard cases.

Vehicle context that stays locked

Danish licence-plate lookup, VIN, or manual search pins the exact vehicle, make, model, year, engine, variant, drivetrain, and carries it through the whole case, so the AI never reasons from symptoms in isolation. AutoIQ's published coverage spans more than 110,000 vehicle configurations across every powertrain, including electric and hybrid, in a browser-based product that runs on computer, tablet, and mobile.

Repair guidance and evidence, compliant by design

Action-level repair steps, checks, and cautions specific to the vehicle, plus technical service bulletins summarised rather than copied and tied to the relevant cause, with the source a click away.

A vehicle-aware AI chat

Mechanics keep investigating in a chat that holds the diagnostic context and answers broader questions about the confirmed vehicle, with streamed responses and sticky controls handled properly.

The commercial layer to run it as a SaaS

Subscriptions, trials, usage quotas, and Stripe billing, with Denmark-ready tax and invoicing, plus admin analytics and support tooling, all bilingual.

Highlights

Project highlights.

Live In production
DA · EN Bilingual product
Plate → repair One workspace
Ranked causes With confidence signals

Tell us what you're solving for.

We'll listen first, ask the right questions, and follow up with a clear proposal.