How do you use ChatGPT for sales discovery questions?

Quick Answer: You can use ChatGPT to generate targeted sales discovery questions by feeding it your ICP, deal context, and the specific stage of the conversation. The prompts in this guide are organised by discovery stage, buyer persona, and deal size so you can pull the right questions for any B2B SaaS sales situation.

Sales discovery is where deals are won or lost. Ask the wrong questions and you sound generic. Ask no questions and you lose control of the call. The problem is that writing sharp, context-specific discovery questions from scratch takes time most reps do not have.

ChatGPT solves this, but only if you prompt it correctly. A vague prompt produces vague questions. This guide gives you a structured system for using ChatGPT to build a discovery question library that fits your product, your buyers, and where they are in the funnel.

What Makes a Good ChatGPT Prompt for Sales Discovery Questions?

A strong discovery question prompt gives ChatGPT three things: context about who you sell to, context about the conversation stage, and a clear output format.

Without those three inputs, ChatGPT defaults to generic questions you could find in any sales textbook. With them, it produces questions specific enough to use on a real call tomorrow.

The base prompt structure to use every time:

You are a B2B SaaS sales expert. I sell [product] to [ICP job title] at 
[company size/type] companies. Generate [number] discovery questions for 
[stage of conversation]. Focus on [specific pain area or goal]. Format 
as a numbered list.

This structure is the foundation for every prompt in this guide.

Stage 1: Early Discovery (Opening the Conversation)

Early discovery questions do one job: get the prospect talking about their current situation without triggering defensiveness. These are not qualifying questions yet. They are context-gathering questions.

ChatGPT Prompt for Early Discovery

You are a B2B SaaS sales expert. I sell [product] to [VP of Operations / 
Head of Revenue / CTO] at mid-market SaaS companies (50-500 employees). 
Generate 8 open-ended discovery questions for the first 10 minutes of a 
discovery call. The goal is to understand their current process and 
surface where friction exists. Avoid questions that feel like an audit. 
Format as a numbered list.

What ChatGPT will return (example output for a revenue intelligence tool):

1. Walk me through how your team currently tracks pipeline health week to week.
2. Where does your forecasting data live right now, and who owns it?
3. What does a typical deal review look like for your team?
4. How do you currently identify which deals need attention before they go cold?
5. What tools are in your stack for managing rep activity and deal progression?

Why these work: They are open-ended, non-threatening, and they invite the prospect to describe their world. That description is where the real discovery happens.

Refining the Output

If the questions feel too broad, add this line to your prompt:

Make each question specific to a company that has recently scaled 
its sales team from 5 to 20 reps and is experiencing process breakdowns.

Specificity in the prompt produces specificity in the output.

Stage 2: Pain and Impact Discovery

Once you have context, the conversation shifts to uncovering pain and quantifying its impact. These questions are harder to write because they need to feel natural, not interrogative.

ChatGPT Prompt for Pain Discovery

You are a B2B SaaS sales expert. I sell [product] to [Head of Sales] at 
B2B SaaS companies with 20-100 reps. Generate 8 discovery questions 
designed to surface the business pain caused by poor pipeline visibility. 
Each question should help the prospect articulate the cost or consequence 
of the problem, not just describe it. Format as a numbered list.

Example output:

1. When forecasts miss, what does that typically mean for the business three months later?
2. How much time does your leadership team spend manually chasing deal updates each week?
3. What happens to rep morale when deals slip that nobody saw coming?
4. If you had to put a number on the revenue you lost last quarter to deals that stalled silently, what would it be?
5. What decisions are you unable to make right now because you do not have reliable pipeline data?

The key prompt variable here is "cost or consequence." That instruction shifts ChatGPT from describing problems to helping you quantify them. Quantified pain is what builds a business case, especially if your team is already thinking about broader B2B SaaS digital marketing agencies or revenue growth strategy.

Stage 3: Future State and Decision Discovery

Late discovery questions focus on what the prospect wants to achieve and how they make buying decisions. These are the questions that qualify deals and set up your pitch.

ChatGPT Prompt for Future State Questions

You are a B2B SaaS sales expert. I sell [product] to B2B SaaS companies. 
Generate 6 discovery questions that help me understand what success looks 
like for the prospect 12 months from now, and 4 questions that surface 
how they evaluate and buy new software. Format in two labelled sections: 
"Future State" and "Decision Process".

Example output:

Future State

1. If this problem were fully solved, what would your team be able to do that they cannot do today?
2. What does hitting your number look like for you personally this year?
3. What would need to be true for you to consider this initiative a success six months in?

Decision Process

1. Who else needs to be involved before a decision like this moves forward?
2. Have you evaluated tools in this space before? What did that process look like?
3. What would need to happen for this to become a priority in the next 30 days?

Categorised by Buyer Persona

The same product creates different problems for different people. A CFO cares about cost and risk. A VP of Sales cares about forecast accuracy and rep productivity. A CTO cares about integration and security.

ChatGPT lets you generate persona-specific question sets in minutes.

Prompt for CFO/Finance Persona

Generate 6 discovery questions for a CFO at a 200-person SaaS company. 
I sell [revenue forecasting software]. Focus on questions that surface 
their concerns about financial predictability, board reporting accuracy, 
and the cost of bad data. Avoid technical questions entirely.

Prompt for VP of Sales Persona

Generate 6 discovery questions for a VP of Sales at a B2B SaaS company 
with 30 reps. I sell [sales coaching software]. Focus on questions about 
rep ramp time, call quality, and what happens when top performers leave.

Prompt for Technical Buyer (CTO/VP Engineering)

Generate 6 discovery questions for a CTO evaluating [data integration 
middleware]. Focus on their current stack complexity, build vs buy 
decisions they have made recently, and what a failed integration has 
cost them in the past.

The persona variable is one of the highest-leverage inputs you can give ChatGPT. It shifts the question framing from product-centric to role-centric, which is how good discovery actually works. The same principle applies when choosing specialist support, whether that is B2B SaaS SEO experts, B2B SaaS PPC experts, or a broader go-to-market partner.

Categorised by Deal Size

SMB, mid-market, and enterprise deals require different discovery approaches. An SMB founder makes decisions in a week. An enterprise procurement process can run for six months with eight stakeholders. Your questions need to reflect that.

SMB Discovery Prompt (1-50 employees)

Generate 6 discovery questions for a founder or Head of Sales at a 
10-30 person SaaS startup. I sell [CRM software]. They are likely 
doing everything manually right now. Focus on questions about how 
they are managing growth, what is breaking, and what they wish they 
had time to fix.

Tone guidance to add: "Keep the questions conversational. This is a peer-to-peer conversation, not a formal sales call."

Mid-Market Discovery Prompt (50-500 employees)

Generate 8 discovery questions for a VP-level buyer at a 100-300 person 
SaaS company. I sell [sales enablement software]. They likely have a 
partial solution in place that is not scaling. Focus on questions about 
what is working, what is not, and where they are feeling the most pressure 
from leadership.

Enterprise Discovery Prompt (500+ employees)

Generate 8 discovery questions for a Director or VP at an enterprise 
software company with 1,000+ employees. I sell [compliance automation 
software]. There are likely multiple stakeholders and an existing vendor 
relationship. Focus on questions that surface dissatisfaction with the 
status quo, internal politics around the decision, and how previous 
initiatives like this have succeeded or failed.

For enterprise deals, add this line: "Include 2 questions specifically designed to uncover internal champions and blockers."

How to Build a Reusable Discovery Question Library

Running these prompts once is useful. Building a library you and your team can pull from is where the real value compounds.

Step 1: Run the prompts above for your top 3 ICPs. Generate question sets for each persona and each deal size. You will end up with 60-80 questions.

Step 2: Filter for the best 10-15 per scenario. Not every question ChatGPT produces will be a winner. Remove anything that sounds scripted or that duplicates another question.

Step 3: Organise by stage, persona, and deal size. A simple Notion database or Google Sheet works. Columns: Stage, Persona, Deal Size, Question, Notes.

Step 4: Test on real calls and iterate. After each call, note which questions opened the conversation and which fell flat. Feed those notes back into ChatGPT:

Here are 3 discovery questions that did not land well on recent calls: 
[paste questions]. Rewrite them to feel less formal and more like 
something a trusted advisor would ask.

Step 5: Share with your team and standardise. A shared library means every rep starts from a strong baseline instead of improvising. If you need outside help tightening positioning, messaging, or process, you can also compare vetted B2B SaaS inbound marketing agencies, B2B SaaS content marketing agencies, or explore the wider top agencies directory.

Common Mistakes When Using ChatGPT for Discovery Questions

Prompting without ICP context. "Generate discovery questions for a sales call" produces generic output. Always include job title, company size, and industry.

Accepting the first output. ChatGPT's first pass is a starting point. Use follow-up prompts to sharpen, reframe, or add specificity.

Ignoring tone. Discovery questions that sound like they came from a script kill rapport. Add tone instructions to every prompt: "Make these sound like questions a trusted advisor would ask, not a questionnaire."

Skipping the filtering step. Not every question in a ChatGPT list is worth asking. Some will be redundant, too blunt, or not relevant to your specific product. Treat the output as raw material, not a finished script.

FAQs

What is the best ChatGPT prompt for sales discovery questions?

The best prompt includes four elements: your product, your buyer's job title and company size, the stage of the discovery conversation, and the specific pain area you want to explore. A prompt like "Generate 8 discovery questions for a VP of Sales at a 100-person SaaS company, focused on surfacing the cost of poor forecast accuracy" will produce far better output than a generic request.

How do I use ChatGPT to prepare for a specific discovery call?

Before the call, feed ChatGPT the prospect's job title, company size, industry, and any context from their website or LinkedIn. Then ask it to generate 5-8 questions tailored to that specific situation. You can also ask it to predict the objections they are likely to raise and suggest questions that pre-empt them.

Can ChatGPT replace a sales discovery framework like MEDDIC or SPIN?

No, and it should not try to. ChatGPT is a question generation tool, not a methodology. Use your existing framework (MEDDIC, SPIN, BANT, or whatever your team runs) as the structure, and use ChatGPT to generate the specific questions that fit each component of that framework. The two work better together than either does alone.

How often should I update my discovery question library?

Review it quarterly. As your product evolves, your ICP shifts, or you move into new markets, your discovery questions need to reflect that. Set a recurring task to run fresh prompts and replace any questions that have stopped landing on calls.

Is there a risk that AI-generated discovery questions sound unnatural?

Yes, if you use them straight from the output without editing. The fix is to read every question out loud before adding it to your library. If you would not say it naturally in a conversation, rewrite it or ask ChatGPT to make it more conversational. The goal is questions that sound like you, not like a bot.