You've learned the hard way.
Vocabulary Test taught you precision. Whiteboard Flood taught you restraint.
Now you want to make sure you never skip the verification step again. And you want to make sure you never over-explain in a demo again.
So here's how I use Claude to lock both of those things in before I ever sit down with the customer.
The workflow
Step 1: Transcribe the discovery call
Record it. Transcribe it. (I use Cursor's voice notes + Claude.)
Step 2: Extract their language
I paste the transcript into Claude with this prompt:
"Review this discovery call transcript. Identify every technical term the customer used to describe their problem, their constraints, and their goals. For each term, note:
- The exact phrase they used
- What they meant when they said it (based on context)
- Any differences between standard industry definition and their use
Focus on: 'integration,' 'real-time,' 'scalability,' 'automation,' 'reporting,' anything that could mean multiple things."
Claude returns a map of their vocabulary. Not my vocabulary. Theirs.
Step 3: Build a focused narrative
Now I ask Claude to help me build a demo outline:
"Based on this customer's specific language and constraints, create a demo outline that:
- Only covers features that directly address their stated problems
- Uses their terminology, not industry standard terms
- Avoids showing anything that wasn't mentioned in discovery
- Includes 2-3 'explain it simply' sections where I'll use a whiteboard (just the essential architecture)
Format: scene → problem they described → how we solve it → show it working → what this means for them"
Claude gives me a structured outline. Clean. Focused. Impossible to flood.
Step 4: Identify dangerous assumptions
Then I ask:
"Review this outline and discovery call. Identify anything where I might be making assumptions about what the customer meant. Where might there be a vocabulary mismatch?"
Claude catches the edge cases. The moments where I think I understood but might not have.
Why this works
Before this workflow, I was running on intuition.
Now I'm running on evidence.
I have a map of their vocabulary. I have an outline that forces me to only include what they actually asked for. And I have a list of assumptions I need to validate before the demo.
The result: demos that feel like conversations, not performances.
Three moments where this workflow saves you
Before the discovery call
Use Claude to research the customer. Ask: "Based on this company and this role, what are the most common definitions of [their key term]?" You'll walk into discovery knowing what confusions might happen, and you can prevent them.
After discovery, before the demo
This is where the workflow above shines. You've got the transcript. You can extract their language and build a focused demo before you ever open the product.
During a demo that's going sideways
If you sense you've misunderstood, ask to pause. Pull up the discovery notes. Ask Claude: "They said X, does this feature address that?" You get clarity in real time instead of after the meeting.
The pattern to remember
Precision in discovery doesn't guarantee clarity in demos.
But precision plus focus does.
Use AI to force yourself to listen—not to assume.
Use AI to force yourself to simplify—not to impress.
Use AI to catch the moments where you think you understood but might not have.
That's not replacing your judgment.
That's amplifying it.
Are you using AI in your sales process in ways that actually work? Reply and tell me what's working for you and what's not. (Anonymized, always.)
— Dan
