How to price AI products: value vs cost frameworks, usage-based, seat-based, or outcome-based billing models, and why low build cost should not become a discount.
Price does not come from the cost of running the model; it comes from the value generated for the user. Map the real value (hours saved, revenue generated, risk avoided), look at what the market pays for similar outcomes, choose between usage-based, seat-based, or outcome-based billing, test willingness to pay, and adjust.
No. Charging for infrastructure cost is leaving money on the table. An AI feature that saves ten hours of work per week is worth far more than the cents spent on tokens. Price should reflect the value delivered, not the cost of delivery.
It depends on the case. Usage-based billing works for variable-consumption features. Seat-based works for productivity tools. Outcome-based works when value is measurable and directly attributable to AI. The Mercury framework suggests testing all three and adjusting based on real willingness to pay.