Getting Ready for a Discovery Call With Claude or ChatGPT

Ask AI to help you prep for a sales call and the default output leans toward talking points: what to say about the product, how to handle the obvious objections, a confident opening line. It's not wrong, and it's also the exact instinct Neil Rackham's research spent years documenting as a weak predictor of success in complex sales: reps who talk about features close less often than reps who ask the right questions in the right order.

The more useful prep isn't a script of things to say. It's a set of questions built to surface what the prospect actually needs before you say anything about what you sell, which is a genuinely different kind of preparation than most reps default to when a call is an hour away and the instinct is to polish the pitch one more time.

What does generic AI sales-call prep actually produce?

Ask an AI to help prepare for a discovery call and you'll typically get a structure built around your product: key features to mention, common objections and rebuttals, a suggested opening line. It reads like a well-organized pitch deck condensed into talking points.

The problem isn't that the advice is wrong. It's that leading with the product is the specific pattern Neil Rackham's research team found correlates with worse outcomes once a sale gets complex and multi-stakeholder, and a generic prompt has no reason to steer away from it unless asked to, since a request for "call prep" reads naturally as a request for talking points about your offering.

What can AI actually do well before a call, and where should you be careful?

Unaided, it's genuinely strong at generating a first pass of open discovery questions once you describe the prospect's industry and role, and at helping you rehearse handling an objection through role-play before you're on the actual call.

It's weak at anything requiring real information about this specific prospect: their internal budget politics, who else is involved in the decision, what they've already tried and rejected. Feed it what you actually know; it can't invent real intelligence about a company it has no access to, and a confident-sounding guess dressed up as insight is worse than an honest gap you go into the call aware of.

What's worth pasting into Claude before this call?

These four work in Claude, ChatGPT, or Gemini as written.

1. The situation map. "I have a discovery call with [role] at a [industry] company. Generate five open situation questions to understand their current setup before I pitch anything." Why it works: situation questions establish the baseline SPIN's research treats as foundational, and generating them before the call keeps the conversation from opening on your product instead of their context.

2. The implication ladder. "Given this likely problem: [describe it]. Write three implication questions that help the prospect surface the cost of not solving it, without me stating the cost for them." Why it works: Rackham's research specifically found that implication questions, ones that get the prospect to articulate consequences themselves, correlated with success far more than a rep asserting the cost directly.

3. The objection library. "List the five most likely objections a [role] at a company this size would raise about [category of solution], ranked by likelihood, with a question-based response to each rather than a rebuttal." Why it works: a question-based response keeps the prospect talking and surfacing more information, rather than shutting the objection down with a defensive answer that ends the thread.

4. The role-play rehearsal. "Play a skeptical [role] hearing my pitch for the first time. Push back realistically, and afterward tell me which of my questions actually got you talking versus which ones I could have skipped." Why it works: rehearsing under mild simulated pushback surfaces which of your planned questions are actually doing work and which are filler, before you find out live on the call, when the cost of finding out is a real prospect losing patience and moving on to a competitor who asked better questions first.

What's left once you've run through all four?

This sequence will sharpen the questions you bring into this specific call, more than most reps manage without prep. It won't give you a discovery habit that survives every call this quarter, one that resists sliding back toward pitching the product the moment an actual live prospect pushes back, or a bad week makes talking about features feel like the safer, easier option. One well-prepped call isn't the same as a questioning discipline that holds up under pressure, call after call.

How does the SPIN framework actually change the prep?

Neil Rackham's research team recorded and analyzed thousands of real sales calls before publishing SPIN Selling, and the finding that reshaped enterprise sales training was specific: situation questions establish facts, problem questions surface difficulties, implication questions get the prospect to articulate the cost of those difficulties themselves, and need-payoff questions get the prospect to state the value of solving it, all before the rep pitches anything.

Before, generic prompting: asked to prep for a call with an operations director at a mid-size logistics company, a generic AI response lists product benefits relevant to logistics and a few likely objections about switching costs, ready to be delivered as a pitch.

After, Rackham's framework applied: the framework refuses to generate pitch content until the four question types exist in sequence. It first asks what situation questions would establish the director's current process (not assumed, asked), then problem questions targeting where that process likely breaks down, then, critically, implication questions that get the director to state out loud what a broken process costs them in specific terms, delays, overtime, lost contracts, rather than the rep asserting it. Only after a need-payoff question gets the director to articulate why solving this matters does the framework allow any product content into the conversation at all.

A generic call-prep prompt has no built-in reason to enforce that sequencing rule, no pitching before the prospect has named their own problem and its cost, in their own words. The framework does, as a hard rule rather than a collection of somewhat better talking points.

How do you get this framework into Claude, ChatGPT, or Gemini?

Claude installs it as a Skill: Settings → Skills → Add skill, upload the .zip. It auto-invokes when your question matches the skill's description, and typing /neil-rackham-framework ahead of a message forces it regardless of that judgment call. In Claude Code, the package lives in ~/.claude/skills for every project or a project's own .claude/skills folder if scoped to one.

ChatGPT has no Skills equivalent, so the .md content becomes a Custom GPT's Instructions: Create a GPT → Configure → Instructions, pasted directly in. The default Custom Instructions fields cap at 1,500 characters, far short of a full framework; a Custom GPT's Instructions field holds roughly 8,000.

Gemini has no dedicated upload flow either, so the content goes into the system prompt if your interface has one, or opens the first message instead: "Here is a thinking framework to apply throughout this conversation. Read it carefully, then answer using this framework's approach," with the framework pasted underneath.

Where does this lead next?

For the call already on your calendar, Discovery Call Prep is a $79 tool built to run this exact sequence before you dial in. Rackham's documented method is worth building into an ongoing discovery discipline rather than a once-off, and it sits, along with Jill Konrath's and Matthew Dixon's, in the Sales Leader category, each an .md file usable with Claude, ChatGPT, or Gemini.