Support · Triage
Ticket triage that hands you a sorted queue and drafted replies
Reading forty messages to find the six that matter is the most expensive hour in a support team's day.
What the worker does
Ticket triage in Polaris means an AI worker with the Gmail connection reads incoming support mail, classifies each message against the severity definitions in its skill file, and delivers a sorted queue as a comment: what is urgent, what is a duplicate of a known issue, what has a drafted reply ready. Support agents send the replies and handle the ones that need a human.
- Connections
- Gmail · Slack
- Cadence
- Per batch, several times a day
- Output
- Sorted queue plus drafted replies
Triage is a classification problem wearing a costume
Nearly every message in a support queue is one of a small number of things: a question already answered in the docs, a bug already known, a billing problem, an angry customer who needs a person, or something genuinely new. A human doing triage is running that classification forty times before lunch.
The classification is mechanical once the definitions exist. The definitions are the part only your team can write, because urgent means something different in a payments product than in a design tool.
How a batch is processed
- 1
Messages arrive as work
The Gmail connection gives the worker access to the support address; Slack signals arrive in the Inbox as prefilled task suggestions rather than silent tasks.
- 2
The worker classifies
Each message is matched against the severity definitions and the known-issue list in the skill file. Anything it cannot classify is marked as needing a person, not guessed.
- 3
Repeatable ones get drafts
Where the answer exists in your help center, the worker drafts a reply that cites the article rather than paraphrasing product behaviour from memory.
- 4
One delivery, ordered by severity
The comment leads with what is urgent. Duplicates are grouped. Every draft sits with the original message so the agent can check it in one read.
- 5
Agents send
A person approves each reply and closes the task. The rating on the delivery is how the worker's classification gets corrected over time.
What makes triage output trustworthy
It shows its reasoning
Each classification names the rule it matched, so an agent can see immediately when a rule is wrong rather than when a customer complains.
It flags uncertainty
A message the worker could not classify arrives marked as such. Confident misclassification is the failure mode that costs you a customer.
It never invents behaviour
Drafts cite the help article. If no article covers it, the draft says the answer needs a person.
It groups duplicates
Fifteen reports of the same outage arrive as one item with fifteen customers attached, which is also how you find out an outage started.
Questions people ask
+Does the worker have access to our whole inbox?
It uses the Gmail connection you authorise, once, for the organisation. Credentials are verified live and stored server-side, so workers use them and browsers cannot read them back.
+How does it learn our severity definitions?
You write them during hiring and they become a SKILL.md file on the worker. Refining triage means editing that document, which is also the audit trail for why a message was classified the way it was.
+Can it close tickets?
No. Polaris workers deliver as comments and humans close tasks. In support that boundary is the safety mechanism, not an inconvenience.
+What does triage cost per day?
It is billed in human-equivalent hours at about $2 an hour, computed from observable effort per session and clamped between five minutes and eight hours. Busy days cost more than quiet ones, and every session is a line on the work log.
Related
Support work, with an AI teammate on the queue
The queue is a conveyor belt of small decisions. Sorting them is mechanical; making them is not.
A template library that stays in your team's voice
Templates go stale the same way documentation does, except a stale template gets sent to a customer.
Escalations that do not go quiet after the handoff
The customer's question is not what the bug is. It is whether anyone is still looking at it.
Hire an AI support specialist
Hand it the backlog on Friday and read eighteen drafted replies, sorted by severity, before you send a single one.
Connect Gmail to Polaris
The honest version: Gmail is in the catalog, the Google sign-in flow has not shipped, and the product says pending rather than pretending.
Human-in-the-loop
The system cannot complete the loop without a person, by design.
Skill file
The capability an agent has, written down where a person can read and edit it.