Skip to main content

Support ticket triage

Route, prioritise and classify every ticket the moment it arrives.

The problem

Triage rules break on the first ticket they did not anticipate, and an LLM classifier costs more than the ticket is worth at volume. Both give you a label with no idea how confident it is.

What Needle replaces

The classification prompt that tags tickets on intake.

What you do with the answer

  • Auto-resolve when the action is clear and the probability is high; queue for a human when it is not.
  • Send urgent tickets to a faster lane using the Noul, with the threshold set by your team, not the model.
  • Report the score distribution over time to see how the mix of severity is shifting.
POST /v1/decisions needle-1
{
  "model": "needle-1",
  "state": "Customer was charged twice for order 4417
            and asks for the duplicate to be refunded.",
  "questions": {
    "action":   {"type": "choice",
                 "criteria": {"refund": null, "deny": null}},
    "urgent":   {"type": "noul",
                 "instructions": "Is this urgent?"},
    "severity": {"type": "score",
                 "criteria": ["low", "medium", "high"]}
  }
}

200 answers 1 forward pass

action choice → refund

refund 0.97
deny 0.03

urgent noul → 0.22

P(true) 0.22

severity score → 0.61

low 0.48
medium 0.43
high 0.09

0 output tokens Illustrative response