Verified against ChatGPT · 2026-08-10
Draft a first sales call script structured around branch points, not a straight-through monologue
Builds a first-meeting call script as a decision tree keyed to how the prospect actually responds at each stage, instead of a single scripted path that falls apart the moment the conversation deviates from it.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Write me a first sales call script — but structured as a branching flow keyed to how the prospect responds, not a single straight-through script I'd read top to bottom regardless of what they say. MEETING CONTEXT 30-minute intro call booked from an inbound demo request form, prospect has seen our website but not a live demo. MY OPENING GOAL Confirm what specifically triggered the demo request and set expectations for what the next 30 minutes will cover. MOST LIKELY POSITIVE RESPONSE AND MOST LIKELY SKEPTICAL RESPONSE Positive: they engage and describe their current manual process unprompted. Skeptical: they say they're "just looking" and not evaluating anything seriously yet. WHAT I MUST LEARN BEFORE THE CALL ENDS Their current tool (if any), rough team size using it, and who else would need to sign off on a purchase. CALL LENGTH 30 minutes. Structure this as: an opening (stated, not scripted word-for-word — a natural framing of why I'm calling and what I hope to get out of the next N minutes), then a branch point after the opening with two paths: what to say and ask next if they respond positively/engaged, and what to say and ask next if they respond skeptically or flatly. Within each branch, include one more branch point for the most likely follow-on response, so the script has real depth on the two or three ways this conversation is actually likely to go, rather than one happy-path script that assumes perfect cooperation. At every branch, make sure the "must learn" items still get covered somewhere in the flow — flag explicitly if a branch risks ending the call before I've learned something I need. Write it as talking points and questions I'd actually say out loud, not as a formal transcript — should sound like a real person on a call, contractions included, not a written memo read aloud. Keep total script length appropriate for the stated call length; do not write a script that would take three times as long to actually deliver. Format as a labeled tree: Opening, then Branch A (positive) with its own sub-branches, then Branch B (skeptical) with its own sub-branches. End with a checklist confirming where each "must learn" item gets covered.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A linear script is the default output shape a model reaches for when simply asked for a "call script," and it fails the moment a real prospect deviates from the assumed happy path — which is most of the time — leaving the rep either improvising with no support or awkwardly forcing the conversation back onto a script that no longer fits what was just said. Structuring the request as an explicit branch tree forces the model to reason about the two or three realistic ways the conversation actually forks, rather than writing one confident monologue and calling it done; this is the same reason real sales enablement teams build call guides as decision trees rather than scripts, and stating that structure explicitly in the prompt is what gets GPT-5.1 to produce it instead of defaulting to prose. Requiring every "must learn" item to be traceable to a specific point in every branch — with an explicit flag if a branch risks ending the call before it's covered — catches the common failure where a skeptical-response branch politely wraps up the call without ever getting to the qualifying questions, which in a real call means walking away without the information needed to know if the prospect is even worth pursuing further. Insisting on spoken, contraction-included phrasing rather than formal written prose matters because a script that reads like an internal memo produces exactly that stilted cadence when a rep tries to deliver it live, while phrasing modeled on actual speech is something a rep can internalize and adapt in real time rather than read verbatim.
What you get back
Opening: "Thanks for grabbing time — before I show anything, I'd love to hear what made you fill out that demo request. What's going on right now that made this worth looking into?" Branch A (engaged, describes pain): ask what they're using today and what's breaking about it, then move toward the must-learn items on team size and sign-off. Branch B ("just looking"): acknowledge that directly, ask what would need to be true for this to become a real priority, then pivot to a lighter version of the must-learn questions framed as "so I don't waste your time later."
Verified against
ChatGPT GPT-5.1 · 2026-08-10
Changelog
- 2026-08-10 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — Google Ads management, built and maintained for you — that's Scult's day job.
EXPLORE GOOGLE ADS MANAGEMENT
