AI Foundations for Small Business · Lesson 10 of 11

How to Run a Discovery Conversation

Time: ~30 min reading + ~20 min worksheet Goal: Run a 60-minute conversation with an SMB owner that surfaces real opportunities, builds trust, and earns you the right to propose.

---

What Discovery Actually Is

Discovery is not a sales call. It's a structured conversation where you learn enough about the business to tell the owner something they don't already know about their own operation. When you do it right, the owner ends the meeting saying "I hadn't thought about it that way. What should we do?"

That's the moment that earns you the engagement. Not the pitch, not the deck — the quality of the questions you asked.

The Three Phases

A well-run 60-minute discovery has three phases:

Phase 1 — Ground Truth (15 min)

Understand the business at a basic level. Revenue rough size, employee count, what they do, how they make money, who the customers are. If you haven't done this homework before the meeting, do most of it before — save meeting time for what you can't google.

Phase 2 — Pain Inventory (30 min)

The bulk of the meeting. Walk the organization and surface real pain. This is where the opportunities hide.

Phase 3 — Reflection and Next Step (15 min)

Play back what you heard. Name the 2–3 biggest opportunities you're seeing. Set up the next meeting.

Phase 1: The Ground Truth Questions

Keep these tight. Don't burn 25 minutes re-establishing what's on their website.

  • "Can you give me the 90-second version of what you do and how you make money?"
  • "How many employees? Roughly split between [roles/departments]?"
  • "Who are your customers? Is it a handful of big ones or lots of small ones?"
  • "Revenue trajectory — growing, flat, cyclical?"
  • "What systems do you run the business on?" (ERP, CRM, accounting, comms, LOB)

The system question is underrated. It tells you the integration landscape, the platform fit (M365 vs. Google), and often reveals a legacy system that will be the biggest constraint.

Phase 2: The Pain Inventory

This is where the skill lives. You're going to walk the org chart in plain English.

The opening

> "I'd like to walk through the business with you and just hear where the friction is. Not 'what do you want to automate with AI' — just where the pain lives. I'll listen, and we'll figure out the AI angle later."

Owners love this framing. Most have been pitched AI already; they're tired of starting with the tech. Starting with the pain lets them talk about what actually matters.

The functional walkthrough

Walk each function in order. For each, ask four questions:

  • "Who in [function] spends their time on what?" — structured answer about the work
  • "What slows them down?" — pain inventory
  • "What's the one thing you wish they could do more of if they had time?" — the re-deployment target
  • "If that team was 20% more effective tomorrow, what would it look like?" — the value story

You're not asking about AI. You're asking about work. The opportunities surface automatically.

Functions to walk (adapt to the business):

  • Sales / business development
  • Marketing / content
  • Operations / production / fulfillment
  • Finance / accounting / AP
  • HR / people
  • Customer service / support
  • IT (if applicable)
  • Leadership's own time

Tip: Most SMB owners will forget to mention something. Prompt specifically: "How about [finance] — what's going on there?"

Deep-dive questions that surface gold

A few questions that reliably surface opportunities across any business:

  • "What document or email gets written over and over with small changes?"
  • "Where does information live that's hard for your team to find?"
  • "Who's the single point of failure — if X got hit by a bus, what breaks?"
  • "What process is still on paper, or on a spreadsheet being emailed around?"
  • "Where are you losing deals? Why do you think that is?"
  • "What's the complaint you hear most from customers?"
  • "What's the complaint you hear most from employees?"
  • "What do you personally spend time on that you wish you didn't?"

The last question is especially important. The owner's own time is usually valuable and usually mis-spent.

Listening for the five patterns

As you listen, mentally tag everything you hear against the five patterns from Module 5:

  • Repetitive text generation → LLM / Custom GPT
  • Scattered knowledge hard to retrieve → RAG
  • Clicking between systems manually → Automation (often + AI)
  • Prediction / classification at scale → Classical ML
  • Complex multi-step workflows with decisions → Agent (cautiously)

Every pain point the owner describes will map to one of these. If you can't map it, either they're not describing an AI opportunity, or you need to ask a clarifying question.

Phase 3: Reflect and Recommend

The last 15 minutes turn insight into momentum.

The playback

> "Let me play back what I heard. The biggest friction I'm hearing is in [function] — specifically around [specific pain]. The #2 theme is [pain]. And there's a smaller but interesting opportunity around [pain]. Does that match your read?"

When you play it back, the owner either agrees (validates your listening) or corrects you (even better — now you really understand the priority).

The prioritization

Rather than pitching, invite them to rank:

> "If you could fix only one of these in the next 90 days, which one?"

Their answer tells you two things:

  • Where the emotional priority is (may differ from the analytical priority)
  • Whether they'll pay for this

The next step

You're not closing a sale in this meeting. You're earning the right to propose. Set it up cleanly:

> "Here's what I'd like to do. I'll put together a one-page opportunity brief on your top 2–3 areas — what I'd recommend, rough sizing, what the path looks like. I'll send it over within a week and then we can do a 30-minute review call. Does that work?"

This is the sales motion. Not "want to do a proposal?" — specific, scoped, low-commitment next step.

Things That Kill Discovery Calls

1. Pitching too early

The moment you start recommending before understanding, you've shifted from trusted advisor to vendor. Stay in learning mode for at least 40 minutes.

2. Talking too much

Owners should be talking 70–80% of the time. If you're hitting 50%, you're over-talking.

3. Bringing a deck

No slides in discovery. Maybe a simple agenda on a single page. A deck shifts the dynamic to presentation.

4. Not taking notes

Take notes visibly. Ask "can I take notes?" at the start. It signals you're taking them seriously.

5. Getting sucked into technology debates

Owner says "I don't think AI is ready." Don't argue. Say "Totally fair — let me understand what you're running into regardless, and we'll figure out whether there's anything worth doing."

6. Not ending with a next step

If the meeting ends with "I'll follow up," you'll never land it. End with a specific deliverable, a specific date.

Your Note Template

Bring a structured notes page to every discovery. Here's a template you can copy:

``` Discovery Notes — [Company Name] — [Date]

Business Shape

  • What they do:
  • Revenue range:
  • Headcount / departments:
  • Customers:
  • Growth posture:
  • Systems: (ERP, CRM, comms, LOB)

Pain Inventory (by function) Sales: Marketing: Operations: Finance: HR: CS: Leadership/Owner's own time:

Owner's Top-3 Priorities (their ranking) 1. 2. 3.

Constraints / concerns

  • Budget indicators:
  • Timeline / urgency:
  • Risk concerns:
  • Change capacity:

Candidate Opportunities (my mapping) Opp 1: [pain] → [AI play, level, pattern] Opp 2: Opp 3:

Readiness observations

  • Data:
  • Tech:
  • Workflow:
  • People:
  • Leadership:

Next Step

  • Deliverable:
  • Date:
  • Attendees:

```

Using the same template every time means your output gets better with reps and your follow-up assessments get easier to write.

The Follow-Up Brief

Between discovery and formal proposal, there's a crucial middle deliverable: the opportunity brief. One page. Three sections:

  • What I heard — plain-English playback of the business and the top pain points.
  • What I'd recommend — 2–3 candidate opportunities with rough sizing, effort, and approach.
  • What I'd do next — the proposed engagement shape.

This document alone will set you apart from 80% of consultants who go from discovery straight to proposal.

---

Key Takeaways
  • Discovery is listening, not pitching.
  • Three phases: Ground Truth, Pain Inventory, Reflect & Recommend.
  • Walk the functional org, asking the same 4 questions per function.
  • Tag every pain point to one of the five AI patterns as you listen.
  • Playback + ranking question + specific next step = momentum.
  • The opportunity brief is your differentiator between meeting and proposal.

---

Next up: Module 11 — Pitching AI to an SMB Owner. How to turn discovery into a signed engagement.

---

Time: ~20 min Goal: Practice the questions and the listening skill. Get reps before you're in a real meeting.

---

Exercise 1: Build Your Discovery Script (5 min)

In your own words (not copied), write the opening you'd use in your first minute of a discovery meeting. Goal: set the tone — you're here to listen, not pitch.

Your opening: ```

```

Exercise 2: Role-Play with AI (10 min)

Use Claude or ChatGPT to run a mock discovery call. Paste this prompt:

> You're going to roleplay as the owner of a 30-person regional residential cleaning company called CleanCraft. You're 54, you started the business 20 years ago, you're tired, and you're skeptical of AI but your accountant keeps mentioning it. Revenue is about $2.8M. You have 4 office staff and 26 cleaners. You use QuickBooks Online, a scheduling tool called Jobber, and basic M365. > > I'm going to run a discovery call with you. Answer questions the way a real, slightly guarded, busy owner would — short answers first, deeper answers only if I ask follow-up questions that earn it. Don't volunteer everything. Throw in a few "I don't know" and "I'd have to ask my office manager" responses to be realistic. > > Start the meeting by saying hello.

Then actually run a 15-minute mock discovery using the three-phase structure. Go as far as you can.

After the call, answer:

  • What did you learn about CleanCraft that you couldn't have guessed? ______________
  • What were the top 2–3 pain points you surfaced? ______________
  • What AI patterns did they map to? ______________
  • What's the one opportunity you'd lead with in your brief? ______________
  • Where did you talk too much or pitch too early? (Be honest — the AI will often let you get away with it, so you have to self-correct.) ______________
Exercise 3: Write the Opportunity Brief (5 min)

Based on your CleanCraft mock discovery, draft the opportunity brief in the template:

What I heard: ```

```

What I'd recommend: ```

```

What I'd do next: ```

```

---

Discussion

Exercise 2: The AI roleplay is an underrated practice tool. You can run a fresh scenario every week, try different industries, try difficult owner personas (arrogant, distracted, skeptical), and get reps without burning real client time. Make this part of your prep routine before any important pitch.

Exercise 3: Your brief should fit on one page. If it doesn't, cut. Owners decide whether to keep meeting with you in the first 30 seconds of reading. Make every sentence earn its place.

Common mistakes to check for:

  • "What I heard" is actually "what I saw on your website" → go deeper
  • "What I'd recommend" names tools before naming opportunities → flip it
  • "What I'd do next" is vague → make it specific: date, deliverable, length

---

Check yourself: Could you run a real 60-minute discovery tomorrow and come out with a credible one-page brief 24 hours later? That's the bar heading into Module 11 and the capstone.

Previous: Sizing the Opportunity · Next: Pitching AI to an SMB Owner