Traditional SaaS pricing assumed that the cost of serving one more active user was relatively small. AI products can behave differently. Every generated response, classification, image or agent action may create a measurable variable cost. The most engaged customer can become the least profitable account if pricing and product usage are disconnected.
Pricing should follow delivered value and cost structure, not copy the previous generation of SaaS.
The main models solve different problems
A subscription gives customers predictability and founders recurring revenue. It works well when usage patterns are reasonably stable and the product delivers ongoing value that is not tied neatly to a unit.
Usage-based pricing connects revenue to consumption, such as requests, tokens, minutes or completed tasks. It protects the company when serving heavy customers is expensive, but it can make purchasing harder when customers cannot predict their bill.
Outcome-based pricing charges against a result, such as a resolved case or qualified opportunity. It can align price closely with value, but only when the outcome is measurable, attributable and resistant to gaming.
A hybrid model combines a subscription or platform fee with an included allowance and additional charges for consumption or outcomes. This can preserve budget predictability while recognising variable cost and expanding value.
Current benchmark data favours experimentation
The best pricing model is discovered through real behaviour, not copied from a category. Test willingness to pay, measure the cost of serving different customers and adjust the model before complexity becomes structural.
That is sourced benchmark data, not a universal rule. The Ryujin Labs analysis is that hybrid pricing often fits the current transition because it balances three realities: buyers want predictability, founders need protection from variable cost, and value frequently expands with usage.
Instrument usage before changing the price
An early-stage company does not need a complicated pricing system on day one. It does need data.
Track model and infrastructure cost by customer, workflow and feature. Connect those costs to product usage, retention and the outcome the customer receives. Identify the usage point at which an account becomes unprofitable and the features that create the strongest reason to renew or expand.
Without this instrumentation, outcome-based pricing is a story and usage-based pricing is a meter. With it, the company can distinguish expensive novelty from a workflow that creates durable customer value.
Decide what the customer is buying
If the customer is buying access to a dependable workspace, a subscription may remain the right base. If the customer is buying a high volume of discrete operations, usage pricing may be clearer. If the product takes responsibility for a measurable business result, an outcome component may be justified.
The practical first step is often a simple subscription with an included usage allowance, transparent limits and internal cost tracking. As evidence accumulates, the company can introduce overages, credits or outcome-linked tiers without rebuilding its entire commercial model.
Do not optimise pricing solely to recover compute cost. Customers pay for value, not tokens. The job is to find a unit that both sides understand and that scales with the result the product creates.
Ryujin Labs helps founders work through choices like these while building a company.
