Monetizing AI
Pricing as a Product Constraint
Over the weekend, Notion co-founder Akshay Kothari shared a lesson many AI startups are still learning: the pricing model itself is a product constraint.
His point about flat pricing creating a “cost ceiling” for AI capability is the part worth sitting with. When you charge $20 per seat, every model upgrade has to fit inside that constraint. You either squeeze intelligence to protect margins or you stop upgrading altogether.
This creates a misalignment that didn’t exist in traditional SaaS.
In classic SaaS, power users were quietly subsidized by casual ones. That worked because marginal cost was close to zero. Postgres doesn’t care whether someone runs ten queries or ten thousand.
AI is different.
Inference costs scale with capability and usage. The customers extracting 100× more value are often costing you 100× more to serve. Under flat pricing, your best customers are also the ones blowing up your unit economics.
What companies like Anthropic and Cursor recognized early is that hybrid pricing isn’t a billing trick. It’s segmentation.
There’s an access price. Then there’s a capability price. Everyone gets a predictable way in. Customers who want frontier performance pay for it. Willingness to pay naturally tracks how much value someone is pulling from the system.
That’s the key shift. You’re not asking customers to choose between predictability and power. Their behavior makes the choice for them.
The transition cost:
Most companies will resist this for a while—not because the math is hard, but because the org is wired around seats. Sales sells seats. CS measures adoption. Finance forecasts seat-based ARR. Usage breaks the muscle memory.
Hybrid pricing forces you to run two motions at once: seat-driven land-and-expand alongside usage-driven expansion. Nothing stays untouched—comp plans, forecasting, QBRs, revenue recognition.
The real constraint is forecasting tolerance. How much variability can finance absorb before confidence buckles?
Pure usage-based pricing is clean in theory and nerve-wracking in practice. SaaS spent two decades training CFOs to value stability. Introducing real usage volatility means reopening that entire conversation.
Hybrid pricing is the bridge. It preserves a stable base while giving product teams room to push the ceiling.
The companies that get this right won’t just price differently. They’ll end up with margin structures that let them keep investing in intelligence while others are stuck tuning legacy models.
If intelligence is the product, the business model has to reflect how that intelligence is actually consumed. The alternative is spending the next three years optimizing a pricing model that caps your product’s potential before it ships.
✌🏽SR