Level 2 · Core skills · Practitioner
How to Use AI for Discovery Questions
Use AI to prepare discovery, not to run it. Build a hypothesis, test it with sharp questions, plan a follow up for every likely answer and know which qualification gaps you must close.
AI is best at discovery before the call: it helps you form a hypothesis, write questions that test it and plan where each answer leads. During the call, you put AI away and listen.
In 30 seconds
- Start with a hypothesis, not a question list.
- Cover five categories: situation, pain, impact, decision process and timeline.
- Plan two follow ups for each likely answer, so you go deeper instead of moving on.
- Map what you still don't know to MEDDICC, BANT or SPICED.
- Practice with AI roleplay, then close the laptop. Never read AI questions off a screen mid call.
Why a hypothesis beats a question list
Weak discovery calls fail the same way: generic questions, generic answers, nothing learned. Ask AI for "good discovery questions" and you get that list, only faster. Instead, ask it to help you form a hypothesis: a specific, testable guess about the buyer's problem, based on your research. It gives every question a purpose, and it makes being wrong useful: a buyer who says "no, that's not it" usually tells you what is.
Key idea
A question list asks the buyer to do your thinking. A hypothesis shows them you already started.
The five question categories
Scroll sideways to see the whole table.
| Category | What you learn | Weak | Strong |
|---|---|---|---|
| Situation | How things work today | "Tell me about your team." | "When a new rep joins, who runs their first two weeks?" |
| Pain | What's broken, in their words | "What are your challenges?" | "Where do new reps get stuck first?" |
| Impact | What it costs the business and the person | "How big a problem is that?" | "What does a slow ramp mean for the number this year, and for you?" |
| Decision process | Who decides, how, with what criteria | "Are you the decision maker?" | "Last time you bought a tool like this, who got involved and what slowed it down?" |
| Timeline | Why now, and which date matters | "When are you looking to buy?" | "Is there a date this needs to work by, like a hiring class?" |
Strong questions are specific, open and anchored to your research or the buyer's words. If research can answer it, don't ask it.
Build a discovery plan in 5 steps
Run this in your company approved AI workspace with your seller context file loaded. It takes about 15 minutes once research is done. No file yet? Build one with the template.
Start from a verified account brief
Build one with How to Use AI for Account Research. Every fact you mention must come from a source you opened.
Draft your hypothesis
with Prompt 1. Lead with one and keep an alternative.
Generate the plan
with Prompt 2: eight questions, each tied to the hypothesis.
Branch and map
Plan follow ups (Prompt 3) and list your qualification gaps (Prompt 4).
Cut it to one card and practice
with Prompt 5. Bring the card, not a script.
For follow ups, demos and meetings with several stakeholders, use the 10 minute meeting prep routine.
Prompt
Prompt 1: Build your hypothesis
After research, before planning
Help me form a discovery hypothesis for my first call with {name}, {title} at {company}. My seller context file covers what we sell and our personas. The verified account brief is below.
1. Write my top hypothesis about the problem {name} most likely owns right now. One sentence, specific enough that the buyer could say "yes, that's it" or "no, it's actually this."
2. List the supporting evidence from the brief, with sources.
3. Write one alternative hypothesis.
Do not invent facts about {company}. If the brief is too thin, say so and tell me what to research next.
<account_brief>
{paste verified brief}
</account_brief>Prompt
Prompt 2: The discovery plan
Once you have a hypothesis
Using my hypothesis and the account brief, build a discovery plan for my meeting with {name}, {title} at {company}:
1. How I'll introduce the hypothesis in plain language in the first 2 minutes
2. 8 open questions across situation (max 1), pain, impact, decision process and timeline, each labeled and each testing the hypothesis
3. The 3 questions I must ask even if time runs short
4. Questions to skip because the brief already answers them
No yes or no questions. No leading questions. Each under 25 words, written the way a person says it out loud.Follow ups for every likely answer
The best discovery happens on the second and third question about one topic. Reps lose that depth when they're thinking about their next scripted question. Planning branches in advance frees you to listen.
Prompt
Prompt 3: Answer branches
After the discovery plan
For each of my 3 must ask questions:
1. List the 3 most likely answers from a {title} at a company like {company}
2. For each answer, write 2 follow ups: one about cause, one about consequence
3. Flag any answer that would disqualify the deal or point to a different buyer
Each follow up under 20 words and open ended. Format as a tree I can scan in 10 seconds.You
Branch this: "Where do new reps get stuck first?"
AI assistant
If "product knowledge": "What makes the product hard to learn?" / "How does that show up on first calls?" If "getting meetings": "What's different between new reps' calls and your best rep's?" / "When do new reps start booking consistently?" If "it depends on the manager": "What do your strongest coaches do differently?" / "Who notices when a rep is struggling, and when?"
Map the gaps to MEDDICC
Every discovery call should tell you more about whether this is a real deal. AI shows you, before the call, which qualification fields are still empty.
Scroll sideways to see the whole table.
| MEDDICC field | A question that usually closes the gap |
|---|---|
| Metrics | "How will you measure whether this worked a year from now?" |
| Economic buyer | "Whose budget would this come from, and what will they want to see?" |
| Decision criteria | "What would make you choose one option over another?" |
| Decision process | "Walk me through the steps from here to a signed agreement." |
| Identify pain | "What happens if nothing changes in the next six months?" |
| Champion | "Who else wants this fixed as badly as you do?" |
| Competition | "What else are you considering, including building it yourselves or doing nothing?" |
Prefer BANT (budget, authority, need, timeline) or SPICED (situation, pain, impact, critical event, decision)? Name it in the prompt.
Prompt
Prompt 4: Qualification gap map
Before and after every discovery call
Here is everything I know about the {company} opportunity: account brief, earlier notes and my discovery plan.
Map it to MEDDICC. For each field: what I know (with the source), what I'm assuming, what's missing, and the one question that closes the gap. Then rank the 3 gaps that matter most for deciding whether this deal is real.Run it again after the call. The difference between the two maps is what you learned. Then turn the transcript into CRM notes with AI for Sales Call Notes.
Weak vs strong question sets
Both sets below target the same fictional VP of Sales at a company hiring a large class of new reps.
Before
(prompt: "Give me good discovery questions for a VP of Sales")
- What are your top priorities this year?
- What are your biggest challenges?
- What tools do you use today?
- What's your budget and timeline?
Weak because every question fits any VP anywhere, and question 3 is answered by their job posts.
After
(hypothesis: "With a large class coming, onboarding depends on one leader's time, and slow ramp puts the hiring plan at risk")
- "I saw you're hiring a big class of new reps. Who carries onboarding today?"
- "When a class starts all at once, what breaks first?"
- "How long until a new rep is fully productive, and how do you know?"
- "What does that cost you, in pipeline and in your own week?"
- "Is there a date this needs to work by?"
Strong because it opens with researched context, follows one thread from situation to impact and leaves room for "actually, the real problem is something else."
Worked example: Calder Freight Systems
Fictional example. Calder Freight Systems, Ramplight and every person here are invented for the Wicked Growth Club course
Seller: Ramplight, a platform that helps sales teams onboard new reps faster with AI coaching. Buyer: Priya Raman, VP of Sales at Calder Freight Systems, a freight software company.
Verified research: a Series B in March with plans to double the sales team over 18 months; 14 open prospecting roles; an enablement posting that says new reps take 6 to 7 months to reach full productivity; a CRO podcast comment that the constraint "isn't leads," it's rep ramp.
Hypothesis: "Calder is about to start large classes of new reps, onboarding runs through Priya personally and a 6 to 7 month ramp puts the hiring plan at risk." Alternative: "The real constraint is uneven manager coaching."
Must ask questions: 1. "Your enablement posting mentions a 6 to 7 month ramp. What drives most of that time?" 2. "When the next class starts, who carries onboarding, and what gives in their week?" 3. "Your CRO has talked about ramp as the constraint. What would he need to see?"
Gaps going in: economic buyer assumed (the CRO) but unconfirmed; decision process unknown; unknown whether security reviews tools that touch call recordings; competition unknown, including "build it ourselves."
What the rep won't do: guess Calder's budget or cite an unapproved customer result.
Practice with AI roleplay
AI gives you unlimited discovery practice. Hide the buyer's real situation so you have to earn it with good questions. The full method is in AI Sales Roleplay.
Prompt
Prompt 5: Discovery roleplay
The day before the call
Play {name}, {title} at {company}. Stay in character until I type END.
Hidden situation (reveal only when I ask a good, specific question): {the real pain, which may differ from my hypothesis}.
Answer briefly. Give vague answers to vague questions. If I ask two questions at once, answer only the second. If I pitch too early, get politely impatient.
After END, score me 1 to 5 on hypothesis framing, depth of follow ups, quantifying impact, uncovering the decision process and talk ratio. List the MEDDICC fields I still don't know and quote the moment I missed the best follow up.Mistakes to avoid
Quick check
Q: AI drafts "What tools do you use today?" but the buyer's job posts already list them. Keep it? A: Cut it, or go deeper: "I saw you use a call recorder. How do new reps use those recordings?"
Q: The buyer says your hypothesis is wrong. Did discovery fail? A: No. A corrected hypothesis is one of the best outcomes of a first call. Follow the new thread.
Try it now
Try it now: 15 minutes, one real meeting
- Pick your next discovery call and open its verified account brief.
- Run Prompts 1 and 2. Choose your hypothesis and three must ask questions.
- Run Prompts 3 and 4. Write your top three gaps on a card.
- Do one 5 minute roleplay with Prompt 5. After the real call, compare it to your plan.
Frequently asked questions
Can AI write good discovery questions?
Yes, if you give it a hypothesis, your research and the buyer's role. Without that context, it produces the generic questions buyers hear on every call.
Should I use AI during a discovery call?
Only an approved note taker, with consent. Don't read AI suggestions while the buyer talks. Prepare with AI before the call and analyze with it after.
How many discovery questions should I prepare?
About eight, planning to ask three to five well. Two good follow ups on one real pain teach you more than ten surface questions.
What's the difference between MEDDICC, BANT and SPICED?
All three are qualification frameworks. BANT is the simplest. MEDDICC suits complex deals with several stakeholders. SPICED centers impact and a critical event. Use your team's framework and name it in the prompt.
AI for Sales Certification
Practice this on your own accounts
10 self paced sessions, an optional weekly live class and a graded capstone, all practiced on your own accounts. Membership is $50 a month. Add the $250 certification when you're ready.
Not sure where to start? Take the 2 minute assessment