Skip to main content
Administration → AI controls everything that makes Copilot and the AI form-filling work: the consent that switches it on, the API key it spends, the model that answers, the voice key for the microphone, what it may be asked and a daily ceiling.
If you don’t see this in your navigation, your administrator can switch the module on under Administration → Modules, or it may not be included in your plan.
The AI settings screen with the consent switch, the provider choice and the voice key The AI settings screen with the consent switch, the provider choice and the voice key

Who can use this screen

Opening the screen takes View AI settings (Ai Configuration · view) and changing anything here takes Edit AI settings (Ai Configuration · edit) — Maintenance Administrator and Maintenance Manager among the starting-point roles, and always the System Administrator. See roles.

The header

Whether it is on

Every AI feature in this product sends the record it is working on — and nothing else — to the provider you configure below. It is off until somebody here turns it on, and turning it off stops it immediately.
Asking Copilot sends the record the question is about — a work order or an asset, its recent history and the lists your organisation has configured — to your chosen provider, outside your deployment.

Whose key it spends

This product has no API key of its own. Every request goes out on the key you put here, so the bill, the rate limit and the provider’s record of what was sent belong to your organisation. Without a key nothing AI works, and nothing is sent anywhere. Choose whose model to use. Two provider cards are offered, Claude from Anthropic and ChatGPT from OpenAI: Neither is chosen by default. Picking a card shows How to get a [provider] key — numbered steps for that provider’s developer console, including adding billing first (a key on an unfunded account authenticates and then refuses every request, which looks exactly like a broken key) and naming the key after this deployment so it can be revoked later. The steps fold away once a key is stored. Changing the provider replaces the key rather than carrying it across. Under the key, Last used shows when a question was last answered, and The provider refused this key on [date] shows the provider’s error — expired, rate limited, a billing hold — so a stuck key can be diagnosed here.

Speaking to it

A microphone appears on every create form, so somebody standing at an asset can say what it is instead of typing thirty fields. The recording is transcribed and thrown away; what is kept is the text, on the record of the question. Voice needs an OpenAI key whichever provider you chose above, because Anthropic has no speech-to-text service. On ChatGPT it is the same key, pasted again here — stored separately so removing one does not silently remove the other. On Claude it is a second key with a second company, so recordings go to OpenAI. Leave it empty and the microphone simply does not appear.

What it may be asked

Tick the capabilities your organisation allows. Nothing ticked means every one of these is allowed. Ticking any one of them allows only what is ticked, so tick Counts, charts and lists as well if people should still be able to ask how many. Anything not ticked is refused by the server, not merely hidden. Save stores the choice. See Copilot for where each one appears.

How much it may cost you in a day

Questions per day is a ceiling across the whole organisation, not per person. It exists because a loop asking the same question forever is invisible until the invoice. The figure N left today shows how much remains. Save stores it. What Copilot has actually been asked, and how often its suggestions were accepted, is on Administration → Copilot activity.