> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fyvel.io/llms.txt
> Use this file to discover all available pages before exploring further.

# AI suggestions

> Machine learning for automatic category suggestions

Fyvel uses machine learning to learn from your existing mappings and suggest categories for bookings that are not mapped yet. The more bookings you map, the better the suggestions become.

### How suggestions work

The AI model analyses your existing mappings — which bookings you assigned to which categories. From that it recognises patterns and suggests matching categories for new, unmapped bookings.

For each booking with a suggestion you see:

* The **suggested category**
* A **confidence in percent** — how sure the model is about this suggestion

<Frame caption="AI suggestions with category and confidence">
  <img src="https://mintcdn.com/fynancegmbh/SYDvdNqQmboyTd2Z/images/working-with-fyvel/AI-Suggestions.png?fit=max&auto=format&n=SYDvdNqQmboyTd2Z&q=85&s=74557854813b83d99d825f47a696745d" alt="Mapping in Fyvel with AI suggestions and the confidence display" width="2374" height="1554" data-path="images/working-with-fyvel/AI-Suggestions.png" />
</Frame>

### Accept or decline suggestions

* Click the **green check** to accept a suggestion. The booking is assigned the suggested category.
* Click the **red X** to decline a suggestion.
* With a bulk selection (several bookings selected) you can accept or decline several suggestions at once.

### Filter suggestions

Turn on the **Suggestions** filter to see bookings that have a suggestion. You can also set a **minimum confidence** — for example, show only suggestions above 80% confidence.

That is useful when you want to clear the most certain suggestions first and finish many mappings quickly.

### Retrain the AI

Click "Train AI" to update the model with your current mappings. New suggestions will be available shortly.

Do this after you have made many new mappings. The model learns from the new data and improves its suggestions.

### Tips

* The more bookings you map manually, the better the suggestions become.
* Retrain the model regularly after larger mapping rounds.
* Start with high-confidence suggestions (>80%). Those are usually reliable.
* Review low-confidence suggestions individually.

***


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.