Ticket urgency

Rate how soon a message needs a response, and send unclear cases to a person.

Problem

First-in, first-out queues treat an outage the same as a request for an old invoice. Keyword rules miss urgency that is described rather than shouted.

Goal

Place each message on a written urgency rubric so code can choose a queue, and catch messages that don't say enough to be placed.

Input / state

The message, and optionally the team it was routed to.

message: The customer's message, unedited.
team: Optional: the team chosen by routing, for context.

Decision question

ScoreQuestion key urgency

How soon does this message need a response from the support team?

Rubric levels

0Can wait: a general question or feedback with no time pressure.
1This week: a problem with a workaround, or a request that isn't blocking.
2Today: a problem blocking the customer's work, or a stated same-day deadline.
3Immediately: an outage, a security concern, or a problem stopping the customer's business right now.

Allowed outcomes

Decided in your code from the answer and its confidence. Thresholds are yours to set and test.

Standard queueScore below 1.5 with confidence above your threshold.
Priority queueScore from 1.5 to below 2.5 with confidence above your threshold.
Page on-callScore of 2.5 or more with confidence above your threshold.
Unclear urgencyConfidence below your threshold: the message doesn't make its urgency clear.

Example scenarios

Predefined demonstrations. No model is called; the outcome shown is an illustrative judgment.

Example state

{
  "message": "Checkout has been failing for every customer since 09:10. We're losing sales every minute."
}

Outcome

Illustrative

Page on-call

Score of 2.5 or more with confidence above your threshold.

Why

An ongoing outage that stops the customer's business is the top level of the rubric.

Better together with

Complementary classifiers, and what each one adds here.

Support ticket routingChoice
Urgency picks the queue inside the team that routing chose.
Combined inSupport inbox triage
In-scope request checkNoul
Urgency on an off-topic message is noise; screen scope first.
Combined inSupport inbox triage

Used in chains

Support inbox triageStep 4 of 4

Sources

Primary documentation this classifier builds on. Scenarios on this page are illustrative, not taken from production use.

SourcedTypeSafe docs · Pattern
Routing by confidence
Checks confidence first and sends spread-out answers to a person, then routes the rest on the score.
SourcedTypeSafe docs · Documentation
System One API reference
The request schema (model, state, questions) and answer schema (choice, score, noul with probabilities and confidence) used by the request snippets here.
SourcedTypeSafe docs · Documentation
Quick start
Routes a support ticket to a department with a Choice, rates frustration with a Score and checks urgency with a Noul in one request.

Using it well

Check confidence before the score. A middle score with low confidence usually means the message is ambiguous, not moderately urgent.
Keep rubric levels concrete and observable, such as "stopping the customer's business", rather than "very urgent".
Paging a person is costly, so give that branch the strictest threshold.

Request definition

POST https://api.typesafe.ai/v1/systemone

{
  "model": "jev-latest",
  "state": {
    "message": "<message>",
    "team": "<team>"
  },
  "questions": {
    "urgency": {
      "type": "score",
      "instructions": "How soon does this message need a response from the support team?",
      "criteria": [
        "Can wait: a general question or feedback with no time pressure.",
        "This week: a problem with a workaround, or a request that isn't blocking.",
        "Today: a problem blocking the customer's work, or a stated same-day deadline.",
        "Immediately: an outage, a security concern, or a problem stopping the customer's business right now."
      ]
    }
  }
}

Request shape from the System One API reference; a Kev server accepts the same body. Replace the placeholder state with your data. Requests need your own API key or server.