AI detection

Some situations cannot be captured with status fields or keywords: an unhappy customer, a hidden requirement, a team that is stuck. AI detection covers exactly these cases with a condition in plain language:

“A customer expresses dissatisfaction or threatens to escalate”

“Someone proposes a new requirement”

“The team is blocked and needs a decision”

The AI evaluates every new comment in scope and sends a notification on a match. AI detection is available as a trigger on ticket reminders (one ticket) and on filter alerts (all tickets in a filter).

📷 Image to add

Screenshot, about 1200 px wide: the reminder dialog with trigger “AI detects …” selected and an example condition in the text field, for example “A customer sounds unhappy”. English UI.

Writing good conditions

Describe the meaning, not a word list. The AI also understands indirect statements: “this is getting annoying” counts as dissatisfaction even though the word “complaint” never appears. For exact words, the keyword filter on the comment trigger is the better and cost-free choice.

Conditions can be up to 500 characters long.

Traceability

Every AI decision, match or no match, is written to the reminder’s audit log together with a short reasoning. The question “why did this trigger” or “why did this stay quiet” always has an answer.

The monthly AI quota

Each Jira site has a shared monthly pool of AI evaluations. The current usage is visible at the top of the HeadsUp overview page.

• The pool depends on the number of licensed users and the edition: 20 evaluations per user per month in Standard, 200 in Advanced.
• All users share the pool freely; there is no per-user limit.
• When the pool is used up, AI reminders pause until the first of the next month. Affected users receive a single notification; nothing pauses silently. All other reminders keep working normally.

📷 Image to add

Screenshot, about 1200 px wide: the AI quota card at the top of the overview page with the progress bar partly filled. English UI.

Good to know

• Evaluation runs on Atlassian-hosted AI models via the Forge platform. Comment texts never leave the Atlassian platform and are not shared with any third-party AI provider.
• AI is not a 100 percent guarantee. In unclear cases HeadsUp rather stays quiet than raises false alarms.
• The notification can arrive a few seconds later than with regular triggers, because evaluation runs asynchronously.
• Very short comments such as “ok” or “+1” are never evaluated; they cannot fulfill a condition and would only consume quota.