AI use cases

These scenarios combine the AI connector, Dolibarr, and often Email or HTTP. Sample data: Acme Solar, contacts @example.com.

🤖 Incoming email summary

Goal: a rep receives a long thread; Knot outputs three actionable bullets.

1. Webhook trigger from your messaging bridge (or manual input).
2. AI node: system prompt “You are a sales assistant; reply in English; max 3 bullets”.
3. Email to the Dolibarr record owner.

Simulate with fake HTML body before activation.

🤖 Quote draft

Goal: speed up a proposal from free-form notes.

1. Object event or Manual on proposal creation.
2. Get third party + order history (Dolibarr).
3. AI: suggest line items (JSON).
4. Logic If: human approval before Update (mandatory guardrail).

Never let AI alone validate a legal amount — always require human review.

🤖 Structured extraction

Goal: turn PDF text or a note field into Dolibarr fields.

1. HTTP or business upload → raw text in context.
2. AI with “Reply only as JSON: { “siret”, “amount”, “due_date” }”.
3. Logic: validate schema (required keys).
4. Dolibarr Update on third party or invoice.

Add a limited retry if JSON is malformed.

🤖 Cost and quality

Lever	Effect
Compact model	Cheaper; enough for classification
Short prompt	Fewer tokens
Logic pre-filter	Fewer AI calls
Simulate	Zero provider cost

Next steps

• Configure a provider
• Permissions