Using Intelligence: SLA Radar and Triage Assist

The Intelligence section

The Intelligence tab brings together the two tools that help you work proactively instead of reactively: the SLA radar, which tells you which tickets are about to slip, and Triage assist, which reads an inbound message and recommends how to handle it. Both are grounded in your real tickets — nothing here acts on its own.

The SLA radar

Every ticket has a service-level target based on its priority:

The radar measures each open ticket against its target and assigns a breach risk:

• Healthy — comfortably within target.
• Watch — half the target window or less remains. Keep it in active review.
• High — a quarter of the window or less remains. Assign an owner and respond before the window closes.
• Breached — the target window has passed. Escalate now and post an internal owner update.
• Met — the ticket is Resolved, Closed, or Archived; its timer has stopped.

The radar sorts the most at-risk work first: Breached, then High, then Watch, then Healthy. It also gives you a summary line — total open tickets, how many are breached, how many are High risk, how many are on Watch, and how many need escalation (breached plus High). Each ticket shows its age, its time to breach (which goes negative once breached), its computed due time, and a plain-language recommendation.

How the timer is calculated

The clock starts at the ticket's creation time (or an explicit due date if one is set) and runs until the target window closes. It stops the moment the ticket reaches Resolved, Closed, or Archived. A ticket with an explicit due date is measured against that date instead of the priority default — useful for commitments you've made to a specific deadline.

Triage assist (AI)

Triage assist reads a support message and returns a structured recommendation:

• Category — what kind of request it is.
• Priority — low, normal, high, or urgent.
• Sentiment — positive, neutral, or negative.
• Summary — a one-line distillation of the message.
• Suggested reply — a drafted response you can edit and send.

To use it:

1. Open Intelligence and find Triage assist.
2. Paste the customer's message.
3. Run the triage.
4. Review the recommendation. Apply the suggested category and priority to the ticket if you agree, and use the drafted reply as a starting point.

Important: Triage assist is a recommendation engine, not an autopilot. It never changes a ticket on its own — you review and apply. Persisted outputs are audited, auto-assignment and priority scoring run in dry-run only, and there is no autonomous escalation.

Availability

Triage assist is powered by the suite's Qwen intelligence layer and is enabled per workspace. When it isn't enabled or the model isn't ready, the tool safely reports that it's unavailable and returns no recommendation — it never blocks your work or invents a result. If you paste an empty message, it returns nothing rather than guessing.

A realistic example

At 4:55 p.m. you open the SLA radar and see one ticket in Breached ("Payouts delayed", 6 hours old against a 4-hour urgent target) and two in High. You escalate the breached one immediately and post an internal owner note. For a brand-new ticket with a long, emotional message, you open Triage assist, paste it, and get back: category: billing, priority: high, sentiment: negative, summary: "Customer charged twice and wants a refund today," plus a drafted apology-and-next-steps reply. You set the ticket to High, route it to billing, and send the drafted reply after adding the refund reference.

Tips

• Check the SLA radar at the start and end of every shift — the escalate count tells you exactly how many tickets need attention right now.
• Set an explicit due date on tickets tied to a promised deadline so the radar measures against your real commitment.
• Treat Triage assist's priority as a second opinion; you own the final call, and the SLA target follows whatever you set.
• Always personalize the suggested reply before sending — it's a draft, not a finished answer.

Troubleshooting

Triage assist says it's unavailable. The AI layer is enabled per workspace and must be ready. When it's off or the model isn't reachable, the tool fails safe and returns no recommendation — this is expected, not an error you need to fix.

A ticket shows as Breached but I already replied. The timer measures against the response target from creation (or due date). Replying doesn't stop the clock — only moving to Resolved, Closed, or Archived does.

The radar is empty. The radar only lists open tickets. If everything is Resolved or Closed, there's nothing at risk — a good place to be.