Claim
Premature gating of AI features and over-engineered launch pricing destroys the only thing pricing iteration needs: usage signal. The right shape is to build the habit first by ungating AI features, then iterate pricing as a product surface — many small changes, none precious — until the market reveals what the model should be.
Mechanism
Pricing without usage data is theory. AI features in particular have no analog reference price for the buyer; the willingness-to-pay only shows up after the habit forms. Gating before habit produces two failures at once: low adoption (so no signal) and an anchor price the market hasn't validated (so any later move looks like discounting). The inverse — ungate, observe, change pricing repeatedly — gets the market to teach you, treats price as part of the product loop rather than a strategy artifact, and normalises change so customers don't read each move as instability.
Conditions
Holds when:
- The AI feature is core enough to drive habit formation when free.
- The team can ship pricing changes quickly (no quarterly committee gating each move).
- The buyer base tolerates ongoing iteration (typical for PLG/self-serve, sometimes for SMB sales-led).
Fails when:
- Anchor pricing is contractually frozen with major accounts (enterprise / multi-year).
- The AI feature is a clear premium tier where free degrades unit economics.
- The org treats pricing as a comms event rather than a product surface.
Evidence
"Assume you do not know the perfect model in advance. Assume the market will teach you."
— Elena Verna, Stripe Sessions talk recap, 2026-05-05.
Lovable changed pricing more than ten times in year one without customer backlash, because they treated monetization as product iteration.
— Verna's recap of the Lovable case, same source.
Signals
- AI feature ungated by default; usage and habit metrics tracked before any pricing decision.
- Pricing change cadence measured in weeks, not quarters.
- Top-up / consumption tier coexists with subscription rather than replacing it.
- "What did the market teach us this month" is a recurring agenda item, not a one-off project.
Counter-evidence
Some categories require strong anchor pricing for credibility (enterprise security, regulated finance, any context where price signals seriousness). In those, frequent iteration reads as instability. Patrick Campbell's pricing research generally argues for fewer, sharper pricing decisions backed by willingness-to-pay studies — Verna's stance is the opposite end of that axis, optimised for the AI-native PLG case where the habit hasn't formed yet.
Cross-references
- PLG sales-led companies need PLG, and PLG companies need sales — both, not one — pricing-as-iterable mirrors PMM-as-infrastructure: shipping moves through, not around, the function.
- Agents are first-class product users; design for output reliability, not navigation — agent-as-user changes the unit-of-pricing question (per-agent vs per-seat), reinforcing the need to iterate.