Verified against ChatGPT · 2026-08-12
Write a coaching note from one transcript moment that an agent will actually act on, not a list of everything they did wrong
Focuses coaching feedback on one specific moment in a transcript, connects it to prior feedback so patterns are visible, and gives a concrete alternative line, instead of an exhaustive critique that overwhelms rather than improves.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Write a coaching note for an agent based on one specific moment in this transcript, not a full critique of the whole interaction. TRANSCRIPT EXCERPT Customer: 'This is the third time I've had to explain this.' Agent: 'I understand, let me just pull up your account real quick.' [proceeds to ask the customer to re-explain the issue anyway] AGENT NAME OR ROLE Marcus, Tier 1 support, 4 months in role. SPECIFIC MOMENT FLAGGED The point where the agent said they understood the customer's frustration about repeating themselves, then asked the customer to repeat themselves anyway. COACHING STYLE Responds well to direct, specific feedback; gets less out of a heavily cushioned 'sandwich' delivery and tends to tune it out. PRIOR FEEDBACK GIVEN Was coached two weeks ago on reading ticket history before asking the customer to restate their issue. Focus entirely on the flagged moment — do not turn this into a review of the whole transcript or list unrelated things that also could have gone better, since a coaching note trying to fix everything at once tends to fix nothing because there's too much to act on before the next call. State plainly what happened at that moment, why it mattered to the outcome (not just "it wasn't ideal" but the actual consequence it likely had), and one concrete alternative line or action the agent could use next time in a similar moment — a coaching note that only says what not to do without a specific replacement leaves the agent knowing less than before about what to actually do instead. If prior feedback was given on something related, name that connection directly rather than treating this as a first-time observation — a pattern across multiple coaching sessions is a different and more important signal than an isolated moment, and an agent who has heard similar feedback before deserves to know that explicitly rather than have it framed as a fresh, disconnected note each time. Match the delivery to the stated coaching style, but do not let a gentler style dilute the actual substance of what needs to change — soften how it's said, never soften what is being said. OUTPUT FORMAT 1. The moment, quoted. 2. What it likely cost (the actual consequence, not a vague judgment). 3. The concrete alternative for next time. 4. Connection to prior feedback, if any, stated directly.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Restricting the note to one flagged moment rather than a full-transcript critique directly counters GPT-5.1's tendency, when given a whole transcript and asked for coaching feedback, to be thorough by default and surface every possible improvement it can find — which produces a note that reads as comprehensive but is actually less useful, because an agent handed five things to fix before their next call will likely act on none of them well, while one specific, well-explained moment is something they can actually carry into the next interaction. Requiring the actual consequence of the moment, not just a judgment that it "wasn't ideal," matters because vague feedback like "could have been more empathetic" gives the agent no way to connect the critique to a concrete outcome, whereas "the customer had already flagged repeating themselves as the specific frustration, and was then asked to do exactly that" makes the stakes legible and memorable. Naming the connection to prior feedback explicitly, when it exists, addresses a structural weakness of one-off transcript reviews: without being told to check, a model treats every transcript as a fresh, isolated data point and won't surface that this is the second time in two weeks the same underlying habit (not reading history before asking the customer to repeat themselves) has shown up, which is the more important coaching signal than either instance alone — a repeated pattern needs to be named as a pattern, not re-delivered as if it were new each time.
What you get back
Moment: agent said 'I understand' to the customer's frustration about repeating themselves, then asked them to restate the issue anyway. Cost: this likely deepened the exact frustration the customer named, since the words of empathy were immediately contradicted by the action taken. Alternative: pull up ticket history before responding, and if it's genuinely not available, say so plainly ('I don't have the earlier notes pulled up yet — bear with me for one more recap') rather than implying understanding you then don't act on. Connection to prior feedback: this is the same underlying habit flagged two weeks ago about reading history before asking customers to repeat themselves — worth naming as a pattern in this session rather than a new note.
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — 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
