AI Features Overview

Ticket0 uses large language models at several points in the support workflow. Each feature is an independent toggle you can enable or disable under AI Agent → Features. See AI feature toggles for defaults and descriptions.

Auto-triage

When a ticket arrives, Ticket0 asks the AI to classify it and attach signals that drive the inbox, routing, and reporting:

• Topic / category (e.g. billing, technical, account, general)
• Urgency score
• Sentiment of the latest customer message (positive, neutral, negative, frustrated)
• Detected language (used for translation when that feature is on)
• Escalation flag when rule-based checks fire on phrases like explicit requests for a human, legal threats, GDPR/data-deletion, emotional distress combined with high urgency, or large financial disputes
• Churn-risk score for customers whose signals suggest they're about to leave

Controlled by the autoTriage toggle (default on). With auto-triage off, tickets still land in the inbox but arrive without AI tags, so you lose filtering and routing-by-category.

Draft replies

With draftReplies on (default), Ticket0 generates a suggested reply during the processing window and pre-fills it in the composer. An operator opens the ticket, reviews the draft, and either sends it as-is, edits it, or writes from scratch. Draft text is grounded in your knowledge base and your team's historical replies.

Drafts are produced regardless of the confidence score — the score only controls what happens after the draft exists (see Auto-reply below).

Auto-reply

With autoReply on (default off), Ticket0 is allowed to send drafts to customers without operator review — but only when the draft's confidence score meets the workspace threshold (default 85 out of 100). Below threshold, the draft is still created and waits for an operator in the composer. Below the hardcoded 60 floor, no draft is shown at all.

This is the only feature that lets AI send messages unsupervised. It's off by default on purpose; admins opt in once they've reviewed enough drafts to trust the quality. See Confidence thresholds for the full two-threshold model and tuning guidance.

Knowledge retrieval

All AI drafts are grounded in your knowledge base. When a ticket needs a reply, Ticket0 semantically searches your KB articles and passes the top matches into the draft prompt as context. Controlled by the knowledgeRetrieval toggle (default on). KB quality is the single biggest lever on draft quality — keep articles specific, current, and de-duplicated.

Similar tickets

On inbound, Ticket0 searches past resolved tickets by semantic similarity and shows matches in the right sidebar of each ticket. Useful for pattern matching and as a signal into draft confidence. Controlled by the similarTickets toggle (default on).

Translation

When translation is on (default off), inbound messages are detected and translated so operators can work in their own language, and outbound drafts can be written in the operator's language and translated back to the customer's. Off by default — turn on when you handle multi-language support.

Spend caps

Each workspace has a configurable monthly AI spend cap (USD). When the cap is reached, AI features pause for the rest of the billing month — tickets continue to arrive, operators can still reply manually, and the cap resets automatically on the 1st of the next month. An alert notification goes out at the warning threshold (default 80%) and again when the cap is reached. Configure under AI Agent → Spend.

Related

• AI feature toggles — defaults and descriptions for each toggle
• Setting up your OpenRouter key — required before any AI feature runs
• Confidence thresholds — tuning when auto-reply is allowed to fire