Customer Support & Ops

Verified against ChatGPT · 2026-08-13

Score ticket priority with visible reasoning instead of a black-box urgent/not-urgent label

Assigns a priority level to each ticket in a queue along with the specific factors that drove the score, so an agent can quickly sanity-check the ranking instead of trusting an unexplained label.

ChatGPT (GPT-5.1)4 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

You are scoring priority for a batch of support tickets so an agent can triage the queue quickly. Every score needs a visible, checkable reason — never an unexplained label.

TICKET BATCH
22 open tickets from the last 4 hours.

PRIORITY FACTORS THAT MATTER FOR THIS BUSINESS
1) Number of users affected (most weight), 2) whether a workaround exists, 3) whether it's billing-related (real money impact), 4) account tier. Tone/wording is explicitly not a factor.

PRIORITY LEVELS AND WHAT EACH MEANS
P1 (respond within 1 hr): active outage or data loss risk. P2 (within 4 hrs): feature broken with no workaround. P3 (within 24 hrs): everything else, including questions and minor bugs with a workaround.

CUSTOMER CONTEXT AVAILABLE
3 of the 22 tickets are from Enterprise-tier accounts; one customer already escalated this same issue once last week.

HOW TO SCORE
For each ticket, weigh the factors in 1) Number of users affected (most weight), 2) whether a workaround exists, 3) whether it's billing-related (real money impact), 4) account tier. Tone/wording is explicitly not a factor. against the actual ticket content — a ticket that sounds urgent in tone (lots of exclamation points, "URGENT" in the subject) is not automatically higher priority than a calmly worded ticket describing an active outage affecting many users; score based on actual business impact described, not emotional tone of the writing. If 3 of the 22 tickets are from Enterprise-tier accounts; one customer already escalated this same issue once last week. indicates this is a high-value account or a customer already escalated once before, factor that in explicitly and say so, rather than scoring purely on the ticket text in isolation. Give a one-line reason for every score that references the specific factor that drove it — "High: affects billing for an entire team account, not just one user" not just "High priority." If two tickets in the batch would tie at the same priority level, note which should genuinely go first based on any tiebreaker in 1) Number of users affected (most weight), 2) whether a workaround exists, 3) whether it's billing-related (real money impact), 4) account tier. Tone/wording is explicitly not a factor., since a flat tie doesn't help an agent decide what to open next.

WHAT NOT TO DO
Do not inflate priority based on the customer's tone or capitalization alone — this is one of the most common mistakes and it trains customers to write in all caps to jump the queue. Do not assign the same priority level to every ticket in the batch as a way of avoiding a judgment call; if the batch genuinely has a distribution of urgency, the scores should reflect that distribution.

OUTPUT FORMAT
A table: ticket ID/excerpt, assigned priority level, one-line reason referencing the specific factor. Sorted with highest priority first.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

Language models are demonstrably sensitive to surface intensity markers — capitalization, exclamation points, words like "urgent" or "critical" — and without an explicit instruction to ignore tone as a factor, GPT-5.1 will let a loudly written but low-impact ticket outscore a calmly worded but genuinely severe one, which is the exact dynamic that trains a customer base to escalate through tone rather than through actual severity, a well-documented failure mode in any triage system, human or automated. Requiring a one-line reason tied to a specific named factor rather than a bare label forces the model to actually apply the weighting scheme given rather than pattern-match to a plausible-sounding priority level, and it gives the human agent reviewing the queue something to disagree with — an unexplained "P1" either has to be trusted blindly or independently re-verified from scratch, while "P1: affects billing for an entire team account" can be checked against the ticket in seconds. Explicitly factoring in account tier and prior escalation history only when instructed to, rather than assumed, matters because those signals aren't in the ticket text itself — GPT-5.1 has no way to know a customer already escalated once unless told, and treating every ticket as a fresh, context-free instance systematically under-prioritizes genuine repeat-escalation cases that businesses generally want surfaced faster. The instruction against flattening the whole batch to one priority level addresses a specific hedging tendency: when a model is uncertain about relative severity across many tickets at once, defaulting everything to a middle priority avoids being wrong about any single one, but it defeats the entire purpose of triage, which exists specifically to create a distribution an agent can act on in order.

What you get back

| Ticket | Priority | Reason | |---|---|---| | #204 'checkout page returns 500 error for all EU customers' | P1 | Active outage affecting all users in a region, no workaround, matches P1 definition directly | | #211 'billing charged twice, Enterprise account, second time reporting this' | P1 | Billing impact plus repeat escalation on Enterprise tier — prioritize above #204 as tiebreaker given prior unresolved escalation | | #198 'URGENT!! why is the export button gray??' | P3 | Cosmetic/UI question with likely workaround (refresh or permissions check); capitalization not weighted as urgency |

Verified against

ChatGPT GPT-5.1 · 2026-08-13

Changelog

  • 2026-08-13 Initial publish, verified against ChatGPT GPT-5.1.

Need this built into your business?

If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.

EXPLORE WHAT SCULT BUILDS
All Customer Support & Ops prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY