The task
You’re pre-launch (or considering a price change) and pricing is unclear. There are dozens of plausible plans: free + paid, freemium + premium, per-seat, usage-based, value-based packaging. The wrong shape is more expensive than the wrong number. The goal isn’t to guess a price — it’s to produce 3 distinct hypotheses with clear validation experiments.
This use case is for SaaS founders before pricing pages go live, indie devs deciding between subscription and one-time, and product teams introducing a new tier.
When AI is the right tool
Use AI for the structuring step: given your product, audience, and willingness-to-pay signals, models can generate distinct pricing shapes (not just numbers) and reason about which segments each shape serves best.
It’s also useful for stress-testing: paste your current pricing hypothesis and ask “what’s the most likely failure mode?”
When not to rely on AI alone
AI doesn’t know your real conversion data, your support cost per customer, or your gross margin reality. Pricing math has to be done by you with actual numbers, not by the model with assumptions.
Skip pure AI for regulated industries where pricing must comply with specific rules (insurance, healthcare).
What to feed the AI
- Product: what it does, the core unit of value (a project, a seat, an API call, a saved hour)
- Audience: specific segment + buying motion (self-serve vs. sales-led)
- WTP signals: what 5-10 target customers said they’d pay, in their own words
- Competitor pricing: 3-5 closest competitors, with shape (not just price)
- Constraints: gross margin floor, billing system, geographic targeting
- One thing you’re certain about and one thing you’re unsure about
The two certainty/uncertainty items keep the model from defaulting to lazy answers.
Copy-ready prompt
You are a pricing strategist. Generate 3 distinct pricing hypotheses for the product below.
Product: {product_with_value_unit}
Audience: {segment_and_buying_motion}
WTP signals: {customer_quotes_with_numbers}
Competitor pricing (with shape): {comp_list}
Constraints: {margin_floor}, {billing_system}, {geo}
Certain: {certainty}
Unsure: {uncertainty}
For each of the 3 hypotheses output:
- One-line summary of the pricing shape
- Plan structure: tiers, what's in each, anchor price + currency
- Target segment + their JTBD this shape serves
- Why this shape (1 paragraph, contrarian if it fits)
- Key risk: what would make this fail
- Leading indicators to monitor in week 1
- A 1-week validation experiment (specific: who, how, what to measure)
End with a "which to test first and why" recommendation, and a "what we'd learn even if it fails" line for each.Recommended output structure
Three side-by-side hypotheses, with the validation experiment as the action item per hypothesis. The “what we’d learn even if it fails” line is the bar for a worthwhile experiment.
Always keep one of the three hypotheses contrarian — e.g., the one nobody on the team likes. That’s often where the biggest signal lives.
How to check the output
Validate each price against your gross margin floor. Check that each hypothesis maps to a segment you can actually reach this month (not “enterprise CIOs” if you have no enterprise motion). Ensure each validation experiment is something you can run in a week without engineering.
Run all three hypotheses past 10 target customers in real conversations. The conversation data beats any model output.
Common mistakes
- Treating pricing like one number — shape matters more than dollars
- No validation experiment, just “guess and watch the dashboard”
- Cargo-culted competitor pricing without considering your unit cost
- A “free tier” that’s so generous nobody upgrades, or so stingy nobody signs up
- Skipping the contrarian hypothesis (the one that surprises you)
Next steps to keep improving
After your first 30 days of pricing live, paste your actual data back into the model with what you learned. Ask for “the next 3 hypotheses to test.” Pricing is a series of tests, not a one-time decision.
Practical depth notes
For How to Generate Pricing Hypotheses With AI Before You Launch, the difference between a usable AI result and a generic one is the input packet. Give the model the audience, the current draft or raw material, the desired format, the decision you need to make, and two examples of what good and bad output look like. Ask it to preserve facts first, then improve structure or wording second.
After the first response, do a separate review pass. Look for missing constraints, invented details, weak calls to action, and language that sounds plausible but does not match the real situation. The best final output should be easy to use immediately: clear owner, clear next step, and no hidden assumption that someone else has to untangle.
FAQ
- How many hypotheses should I run at once? Two, max. More splits your traffic and confuses the signal.
- What about discounts? Don’t bake them into the base price. Use them as a separate experiment.
- When do I lock pricing in? When the same hypothesis wins across two segments over 60 days. Then revisit annually.
Related
Reuse the structure with pricing hypothesis prompts, align pricing to message with product positioning prompts, and pressure-test the underlying idea using startup idea evaluation prompts.