AI · Reforge Foundation
How to Price Your AI Product
AI introduces variable costs that wreck traditional SaaS pricing. The full pricing framework for AI products — feature gating, pricing levels, payment timing, and the four levers that compose a coherent monetization system.
SaaS pricing assumptions don't hold for AI
Traditional SaaS pricing was built on fixed-cost software with marginal cost approaching zero. Add a user, add nearly zero marginal cost. Charge per seat, capture LTV across the seat's tenure. AI products have variable costs that scale with usage — inference cost per query, model cost per token, infrastructure cost per active user. Per-seat pricing breaks when a power user's inference cost exceeds their seat fee.
The four levers of AI monetization
- What you charge for. Features, tiers, capabilities, capacity. AI products often need separate gating for "access to better models" vs "more usage of standard models."
- How much you charge. Price levels. AI features command $4–30/mo premium in analysis of 44 leading tech companies — but only when the willingness-to-pay actually exists.
- How you charge. Per-seat, per-usage, per-outcome, hybrid. Pure usage-based pricing matches AI cost structure but creates customer-side budget unpredictability that limits adoption.
- When you charge. Timing of payment matters for conversion and willingness-to-pay. Free trial → paid is different from freemium → upgrade is different from paid trial → continued paid.
Why AI features command premium pricing
AI features have shown commanding premium pricing in many categories — $4–30/mo over baseline — when (a) the use case has clear ROI for the user, (b) the AI capability is genuinely differentiated from what users could get for free elsewhere, and (c) the pricing aligns with how the user perceives value (per-output vs per-time-saved vs per-quality-improved).