What the AI features in Freveo send to a model, where that model runs, what is kept afterwards, and what none of it is used for.
Last updated 3 October 2026
Freveo has AI features that draft text for you — invoice and project wording, profile copy, document clauses — and an assistant that can read your workspace and carry out actions in it when you ask. Every one of them is optional: nothing sends anything to a model unless you invoke it.
AI output is a suggestion, not a decision. Read it before you rely on it, and never send a client something you have not checked.
A drafting feature sends your instruction plus the record it is drafting from — for example a project's title and description, or an invoice's line items. The assistant sends your message plus whatever your question requires it to look up in your own workspace.
That means workspace data, including personal data about your clients, can be sent to the model when a feature you invoked needs it. Nothing is sent from an account that never uses these features.
The model that produces these suggestions runs on infrastructure Freveo operates. Your instruction and the record it is drafting from are processed there and go no further: no third-party model provider receives your workspace data, and none is a subprocessor of ours for it.
If that ever changes — if we move inference to a hosted provider — we will name that provider and where it runs, here and in the privacy policy, and we will tell you in the product before the change takes effect. Until this section names one, there is none.
Freveo does not train models on your workspace data, and does not give it to anyone else to train on.
Because inference runs on our own infrastructure, there is no provider holding what you send, reviewing it, or retaining it beyond the moment it takes to answer. Nothing you send to a model is kept as training data, sample data, or an evaluation set.
AI conversations are not stored as part of your workspace: the assistant has no history you can scroll back through, and we do not keep one either. What we do store is how many AI credits each account has used, so the product knows what is left.
If you use the report button on an assistant reply, that turn — your message, the reply and the trace of what the assistant did — is attached to a support ticket so we can look at it. That is the only case where a conversation is written down.
The client and project records in your workspace are yours, and the AI features do not change that. We do not use them to train anything, we do not sell them, and we do not contact your clients. If you would rather no client data reached a model at all, do not use the AI features on records that contain it.
AI use is metered in credits. Your plan includes a monthly allowance that resets each month, and credit packs add a balance that does not expire. When you are out of credits the AI features stop until the allowance resets or you buy a pack — nothing else about the product is affected.
AI features can be confidently wrong. They can misread a record, invent a detail, or word a clause in a way that does not mean what you want it to mean. Freveo is a tool, not an accountant or a law firm: what you send to a client is your responsibility, whether you or a model drafted it.
If this policy changes in a way that matters — a new provider above all — we will say so in the product before the change takes effect. The date at the top always reflects the latest revision.