Tender uses AI models in seven features:
| Feature | What it does |
|---|---|
| Tender chat | A conversation with tools that query your tenders |
| Tender document analysis | Thematic analysis of the documents of a tender |
| Document extraction | Reading CVs and company documents to fill in profiles and teams |
| Bid support | Drafts, structure and fit of a bid |
| Match descriptions | Automatically written explanations of why a tender fits your profile |
| Semantic matching | Comparing published tenders against the company profile |
| Reading scanned documents | Transcribing tender documents that arrive as images |
Tender document analysis is a long job: the model is fixed when it starts and doesn't change midway.
Three of these features aren't decided by whoever is in front of them, because nobody is. Match descriptions and semantic matching run on their own every time new tenders appear, so their model is fixed by the organisation's administrator. Reading scanned documents is fixed by the platform: it needs a model that can read images, and it cannot be changed.
Where the model is chosen
None of this is configured in Tender: what your organisation chooses, it chooses once, under Organisation → AI models in your account, and it applies to every application at the same time. There you can see, feature by feature, which models the platform authorises and which one is in use.
Who decides what, what the degrees of control mean and what happens when a model is retired are covered in AI models.