AI: Translating Expertise into Customer Language

AI as a novice proxy

The trouble with the curse of knowledge is that you can't see it in yourself. You need an outside, naive perspective that doesn't share your automatisms. That's exactly what AI can simulate — provided you frame it correctly.

graph LR
    A[Your expert text] --> B[AI framed as 'novice proxy']
    B --> C[Jargon detection]
    B --> D[Logical-leap detection]
    B --> E[Per-level rewriting]
    C --> F[De-cursed version]
    D --> F
    E --> F
    F --> G[Test on prospects]

Prompt 1: Anti-curse audit of a sales page

You are a non-expert prospect in the field of [company sector].
You are smart but you have no technical knowledge of the field.

Here is a sales page / pitch / commercial email:
"""
[Paste the text]
"""

Run this anti-curse-of-knowledge audit:

1. JARGON: list every word, acronym or expression you don't
   spontaneously understand. For each, provide an accessible rewording.

2. LOGICAL LEAPS: spot places where the author assumes prior knowledge
   you don't have. List them as:
   "Sentence X — assumed knowledge: ..."

3. FUZZY BENEFIT: for each mentioned feature, indicate whether the
   concrete benefit for me (end customer) is explicit.

4. COMPREHENSION TEST: in 2 sentences max, state what you understood.
   Be honest: if you didn't understand, say so.

5. RECOMMENDATIONS: propose 5 concrete improvements, ranked by
   expected impact on conversion.

Why this prompt works

  • It assigns a clear role (non-expert prospect)
  • It bans complacency ("be honest")
  • It separates jargon, logical leaps, benefits: the 3 curse vectors
  • It ends on a real measurement: what's left in the head after reading?

Prompt 2: Translating a technical feature into a customer benefit

You are a copywriter expert in sales psychology.

Here is a feature of my product:
[Technical description of the feature]

My target persona:
[Persona description: job, pain, technical level]

Generate 5 rewordings of this feature, going from "feature" to
"customer benefit". For each, use a different angle:
1. Time benefit: "you save X hours per week"
2. Money benefit: "you save / earn X euros"
3. Status benefit: "you show your team / client / boss that..."
4. Emotional benefit: "you stop stressing about..."
5. Metaphorical benefit: "it's like having an X in your pocket"

For each, add:
- Residual jargon level (0 = none, 5 = very technical)
- Main action verb
- Pain avoided

Prompt 3: 3-level pitch (child, teen, peer)

Here is the description of my product:
[Full description]

Generate 3 versions of the pitch:

1. CHILD LEVEL (8 years old)
- Max 3 sentences
- Not a single technical word
- A comparison with something from everyday life

2. TEEN LEVEL (15-18)
- Max 5 sentences
- Everyday terms
- One analogy with something a teen knows
  (video game, social network, school)

3. PEER LEVEL (another founder / an investor)
- Max 7 sentences
- Business terms ok, but no tech jargon
- Must surface the market, the solution, the differentiator

Keep meaning and coherence across all 3 levels: same idea,
3 depths.

What's the child pitch for?

Not for pitching to a child — but to force you to articulate the pure essence of your offer. If you can't explain it to a child, you haven't fully understood yourself what you're selling.

"If you can't explain it simply, you don't understand it well enough." — Attributed to Einstein

Prompt 4: Detecting logical leaps in a sales argument

You are a logical-leap detector.

Here is a sales argument:
"""
[Text]
"""

Split it into numbered sentences. For each, indicate:
- Which prior knowledge it assumes
- Which previous sentence(s) it relies on (if any)
- Whether there's a logical leap (the author skips a mental step
  necessary for a reader with no context)

At the end, list the most severe logical leaps (that break overall
comprehension) and propose a transition sentence for each.

Prompt 5: Reverse "rubber duck" to prep a demo

You will play the role of a prospect in a product demo. I'm the
salesperson. Context:
- My product: [description]
- Your profile: [detailed prospect profile, technical level]
- What you want to know before signing: [criteria]

Rules:
- You ask questions ONLY as a real prospect would, not as an expert
- When I say something fuzzy or jargony, you say "I don't understand,
  can you rephrase"
- You don't help complete my sentences: if I'm unclear, you flag my
  lack of clarity
- At the end, give a clarity score /10 and 3 top improvements

Let's start. I begin with: [your opening pitch]

The trick that makes this prompt deadly

AI, by default, completes your sentences and guesses your intent. That's the opposite of what you need to detect the curse. Prompt 5 explicitly disables this tendency — AI then becomes a true mirror of prospect confusion.

Prompt 6: Generate a metaphor bank

Metaphors are the #1 antidote to the curse: they carry a complex idea via an already-known concept.

My product does: [description]
My persona: [description]

Generate 15 metaphors to explain my product. Each metaphor must:
1. Rely on something my persona already knows well
2. Preserve the essence of the product (not a side detail)
3. Be understandable in less than 2 seconds
4. Be usable in one sentence ("It's like X but...")

For each:
- The metaphor
- The semantic field (sport, cooking, transport, school...)
- Any risk (bad connotation, ambiguity)
- A ready-to-paste landing-page wording

Anti-curse AI stack (recommendation)

Use Tool / Method Frequency
Landing page audit Prompt 1 on ChatGPT / Claude Before every release
Rewriting benefits Prompt 2 At every new feature
3-level pitch Prompt 3 At every major evolution
Prospect simulation Prompt 5 (conversational mode) Before every big demo
Metaphor bank Prompt 6 1× per quarter

Limits & guardrails

  • AI is still influenced by your writing: if your prompt is itself steeped in jargon, AI may reproduce it. Forcing the "prospect voice" is crucial.
  • Test with real humans: AI is an excellent first filter, not the final truth. 5 real users are worth 100 AI iterations.
  • Beware over-simplification: dumbing down may insult a technical prospect. Calibrate to your persona.

In the final chapter, we'll see entrepreneurial strategies to structure your whole company around anti-cursing.