AI diagnosis assistant: asking for help with a read

Where: Remote sessions › Diagnostics › Vehicle diagnostics › open a completed read › “AI diagnosis assistant”, or “Ask for help with …” next to each code

What you will see: a block with the notice that this is AI-generated advice, an optional question and, after asking, the summary, the probable causes, the step-by-step checks and a draft to explain it to the customer.

1. Asking for help

1. Open a completed read of your organization.
2. If you want, choose a code with “Ask for help with …” or write what it should look at (no customer details).
3. Press “Ask the assistant for help” and wait a few seconds.
4. Review the answer and mark “Useful” or “Not useful”.

2. What the answer contains

3. What the assistant never does

• It does not run procedures or write to the vehicle: it does not clear codes, program, code or run adaptations, routines or actuator tests. If the AI suggests something like that, it is removed from the answer.
• It does not help to disable emission systems (DPF, EGR, AdBlue, catalyst, lambda sensors), change the odometer, bypass the immobilizer or erase crash data: the question is refused.
• It does not invent data: a cause not supported by the read is removed, and whatever does not come from the library is marked as the assistant's reasoning.

4. Turning it on

• It is off by default. Three things are needed: the server allows it, the Hub turns it on and chooses the provider, and your organization opts in.
• Your organization's administration turns it on in Administration › Company › “AI diagnosis assistant”, with its daily and monthly spending caps (at most what the Hub allows). Saving asks for the second factor.
• The Hub configures it in Administration › Platform › Features and appearance: caps for the whole platform, requests per person and retention days, and in the “Model catalog” the models, their prices and what each organization pays.

5. Model, price and charges

• The Hub makes one or more models available to your organization (Anthropic Claude, OpenAI, xAI Grok, DeepSeek, Qwen, Kimi…). In Administration › Company › “AI diagnosis assistant” you see each one with its description, where it processes data and what it costs you, and you choose which one to use. Changing it asks for the second factor and is recorded in the audit log.
• The Hub sets the rate: free, per token, by time of use (the first request opens a block of one minute or one hour and the next ones inside it are not charged) or per request. Before asking for help you see the maximum price and that of a typical request, and afterwards what was charged.
• Paid requests are charged to the prepaid wallet: the most it can cost is held, what it really costs is charged and the rest goes back to the balance. Free ones need no balance. Without enough balance the AI is not called and you see the “Reduced report”: what the library says about each code and what to check, at no cost.
• If the Hub allows it, your organization can use its own provider key: you pay the provider for the use and the platform charges nothing or only the fee per request the Hub sets. The key is encrypted in your browser and only its last four characters are shown.

6. Data and retention

• The AI provider receives the codes with their status, the freeze frame, the live values and the tests of the read, the make, the year and the VIN (or only its manufacturer, the first three characters, if the Hub decides so, for example for providers outside the EU). Never customer details, names or the identification of the ECUs.
• Some providers process data outside the EU (United States or China): the screen warns about it, and their use relies on the standard contractual clauses. Neither we nor the providers train models with your data.
• Requests and answers are kept for 90 days (or what the Hub sets) and then deleted automatically. Each request is recorded in the audit log and its amount in the usage record for invoicing.

Important: This is AI-generated advice: check every cause and measurement against the manufacturer's documentation before replacing parts or telling the customer.

Tip: If you reach the request limit or the spending cap, the assistant says so and you can try again later.

Open the platform See it in the help centre