Verified against ChatGPT · 2026-08-08
Draft the customer-facing apology and the internal postmortem for a missed SLA in one pass without mixing the two audiences up
Produces a customer-facing SLA breach notice and a separate internal incident note from the same facts, so the external message stays accountable without over-explaining, and the internal note stays honest without becoming a legal liability if it leaked.
The prompt
Ready to copy — highlighted parts are example details you can swap.
A specific SLA commitment was missed on a specific customer's ticket. I need two documents from the same facts: a message that will actually be sent to the customer, and a short internal note for the account team — written for different audiences, so don't write one and just relabel it for the other. SLA TERMS THAT WERE BREACHED P1 tickets guaranteed a first response within 2 business hours under the Enterprise support tier. WHAT HAPPENED First response took 11 hours because the ticket was misrouted to the wrong queue and sat unassigned overnight. CUSTOMER CONTEXT Enterprise account renewing in six weeks; this is the second SLA miss on their tickets in the past quarter. ROOT CAUSE STATUS Routing misconfiguration confirmed by the on-call engineer; not yet confirmed whether it affected other tickets in the same window. COMPENSATION POLICY Enterprise SLA policy authorizes a service credit equal to 5% of monthly fees per confirmed P1 breach, no further discretion without VP approval. For the customer-facing message: open by stating plainly which specific commitment was missed and by how much, in the customer's terms (hours, not internal ticket IDs) — do not bury the acknowledgment under a paragraph of context first. State what is known about the cause only to the extent it is actually confirmed; if the cause isn't fully confirmed yet, say that directly rather than offering a guess dressed up as an explanation, since a customer who was told a specific cause and then finds out it was wrong loses more trust than one who was told the truth was still being confirmed. Apply the compensation policy exactly as given — do not invent a credit, discount, or gesture beyond what the policy authorizes, and do not under-deliver relative to it either. Close with a concrete next step and a named point of contact, not "we're looking into it." For the internal note: state the same facts plainly, including anything that would be awkward for the customer to see, since this note's job is accurate internal record-keeping, not diplomacy. Note whether this SLA has been breached for this customer before, if that context was given, since a repeat breach usually changes what the account team should do next beyond this one incident. Flag anything in the customer-facing draft that internal readers should know was deliberately simplified or left out, and why, so nobody on the account team accidentally contradicts the external message by assuming it contains the full picture. OUTPUT FORMAT 1. Customer-facing message, ready to send. 2. Internal note, under 150 words, for the account team. 3. One line flagging anything simplified in the customer message and why.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Splitting the two documents from one fact set, rather than asking for a single writeup, matters because GPT-5.1 defaults to a single register when given one prompt and one set of facts — left unconstrained it will write something that reads like an internal incident note wearing a customer-facing greeting, which either over-explains internal process to the customer or under-states the facts to the internal audience. Explicitly telling the model to state the cause only to the extent confirmed addresses a specific failure mode: language models asked to explain an incident tend to complete the explanation, filling a not-yet-confirmed cause with a definite-sounding one because an unresolved sentence reads as less satisfying to finish than a resolved one — the instruction gives it explicit permission to leave a gap stated as a gap. Locking compensation to exactly what the stated policy authorizes closes off a common and costly failure where a model, trying to sound appropriately contrite, offers a bigger gesture than the person running this prompt is actually authorized to give, which becomes a real commitment once sent. The internal note's instruction to flag what was deliberately simplified in the external draft solves a coordination problem that has nothing to do with model behavior and everything to do with organizational reality: whoever reads the internal note later needs to know the external message wasn't the full story, so they don't repeat a detail to the customer that was intentionally left out of the first message.
What you get back
Customer message: "Your P1 ticket submitted at 9:14 AM did not receive a first response within our 2-hour Enterprise commitment — it was answered at 8:20 PM, roughly 9 hours late. This was caused by a routing error on our side that we are still confirming didn't affect other tickets in that window. Per your Enterprise SLA, a 5% service credit has been applied to this month's invoice. Your account manager, Priya, will follow up directly by Thursday with final confirmation of the cause." Internal note: routing misconfig, second breach this quarter, renewal in 6 weeks — recommend account review before renewal call, not just a credit.
Verified against
ChatGPT GPT-5.1 · 2026-08-08
Changelog
- 2026-08-08 — 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
