Artificial intelligence on tamam.be
Which parts are written by an AI model, which models those are, and what you should not rely on.
Where AI is used
In the places below, a language model writes the text you read. Each of those places also says so next to the text itself.
- The analysis of your request: category, urgency and a structured summary of your description.
- The assessment of photos you attach to your request.
- The do-it-yourself guide, including the steps, the cost estimate and the safety warnings.
- The help with drafting a quote, for professionals.
- The reading of a price list a professional uploads.
Who supplies the models
We do not run our own models. Everything goes through OpenRouter, an intermediary that gives us access to third-party language models. We are therefore a user of those models, not their maker.
- Understanding and summarising text: your description is structured into a request.
- Reading photos: what your photo shows is turned into words.
- Turning speech into text: what you say is written out.
Which model runs for that can change without anything changing for you. Who processes the data and on what basis is set out in our processing records.
What not to expect from it
A model writes what is probably right, not what is certainly right. Concretely:
- An indicative price is not a quote. Only the amount a professional sends you counts.
- A do-it-yourself guide replaces neither a professional nor an inspection. Gas, electricity, height or water: have it done.
- Nobody here reads the generated text before you see it. There is no human check in between.
- Models get things wrong, sometimes convincingly. If something is off, tell us at legal@tamam.be.
What is not AI
Assigning requests to professionals is not AI. It is a fixed calculation with four components and fixed weights, which does not learn and does not change based on outcomes. The weights are published: How we rank professionals
Your data
For every AI call we keep a hashed fingerprint of the request rather than the request itself, so we can investigate errors without building an archive of job descriptions. What else we process is described in the privacy policy