AI for projects with SLA

If you handle customer tickets as tasks of a project with service levels (SLA), the module AI Project SLA gives your team reply drafts, hand-over summaries, an SLA risk check and an optional automatic triage of new tickets. It needs the project SLA app and the knowledge sources module (see Knowledge sources).

Quick prompts on tickets

For project users, these quick prompts appear in the AI panel on a task (see The AI panel). They read the recent messages, the SLA lines of the task, similar solved tickets and your knowledge sources, and they change nothing:

  • Draft customer reply: a reply based on the conversation, the way similar tickets were solved and your sources, with numbered citations. It is never posted for you.

  • Summarise ticket: a hand-over summary including the SLA situation. It replaces the generic summary prompt on tasks.

  • Suggest priority and assignee: a suggestion with reasons, based on similar tickets, the SLA and the workload of the candidates.

  • SLA risk: for every SLA, what it measures, the deadline in your time zone, the hours left, whether it is paused, fulfilled, urgent or breached, and what to do now.

Note

The suggestions about similar tickets depend on the similar-tasks feature being on for the project and on an embedding model (see Knowledge sources). The prompts are offered on every task for project users; on a task without an SLA they say so.

Automatic triage of new tickets

The triage lets the AI classify a new ticket: its priority, its tags and ticket type, and its assignee when nobody is assigned yet. It is off by default and is set per project.

Set it up

  1. Open the project form and find the setting AI triage in the AI block. Choose Off, Propose for review or Apply automatically.

  2. Choose the Triage reviewer (by default the project manager). The reviewer must be an AI user, and the project cannot be saved with triage on otherwise.

  3. Optionally list the tags that stand for ticket types in Ticket types, for example Bug and Question.

  4. Ask an administrator to allow AI writes on tasks: the capability Write data must be on and an AI access rule must allow writing the priority, tags and assignees of tasks (see Who can use what and Access policy). Both are off by default.

What happens

  • A new task of the project is marked Waiting for AI triage. About every 10 minutes (and right after the task is created) the AI first embeds the new tasks so that similar tickets are known, and then triages each queued task in the name of the reviewer, with the reviewer’s rights.

  • The AI sees the ticket, its messages, the allowed priorities, ticket types and tags, the possible assignees (only when the task has none), similar solved tickets and the SLA. It can answer only with the offered values: a priority, tags and a ticket type from the lists and an assignee from the candidates. Values that equal the current ones are dropped. If nothing is left, the task shows Nothing to change.

  • With Propose for review, the proposal waits for the reviewer. On the task, the reviewer sees a banner with the reasoning and the buttons Apply and Discard, gets a to-do activity, and finds the proposal also under the AI write proposals. With Apply automatically the same proposal is applied at once, and if that fails it waits for review. Unreviewed proposals expire after 24 hours by default.

  • You can also triage a single ticket yourself with Triage with AI in the task header. The proposal is then addressed to you.

  • Applying a triage does not count as the team’s first reaction to the customer, so a “first activity” SLA is not marked as fulfilled by the AI’s classification.

Note

Nothing is changed on a ticket without the reviewer’s confirmation unless the project manager chose Apply automatically. Every applied change is recorded in the AI change log (see Monitoring). The AI reads only the names of the possible assignees, and the workload comes from the tasks themselves.