Most business AI tools are sold per user per month, inherited from the software pricing that preceded them. The model fits procurement well and fits the usage pattern badly.
Why per-seat became the default
A fixed monthly figure per user is predictable, easy to forecast and simple to approve, which matters more to a buyer than to a user.
It also decouples the vendor's revenue from its costs, which is comfortable while models are cheap and uncomfortable when a heavy user costs more than they pay.
The result is a pricing model designed around the finance department's requirements rather than around how the product is used.
How it suppresses occasional use
AI tool usage inside a company is heavily skewed. A minority use it daily, and a much larger group would use it a few times a month.
Per-seat pricing forces a decision on that larger group: pay a full monthly fee for occasional value, or go without.
Most managers cut the occasional users to control cost, which removes exactly the population that would have discovered new applications for the tool.
Why usage-based pricing shifts the risk
Charging per request aligns cost with value and lets everyone have access, which solves the occasional-user problem directly.
It creates a new one. Spend becomes unpredictable, and unpredictable spend is harder to approve than a larger predictable figure.
Teams also behave differently under metered pricing, hesitating before requests in a way that undermines the habit formation the vendor wants.
What procurement optimises for
Procurement is measured on negotiated discount and contract certainty, not on realised value per user.
That favours large annual commitments with per-seat allocations decided in advance, months before anyone knows who will actually use the tool.
Unused seats are the predictable consequence, and their existence then argues against renewal, regardless of how much value the active users obtained.
Where pricing changes the tool
Vendors build what their pricing rewards. Per-seat revenue encourages features that justify a personal login rather than features that serve a whole team.
Shared assets, team-level automation and background processing all sit awkwardly against a model that charges for individual humans.
This is why several vendors have moved to hybrid arrangements with a small seat fee and metered usage above it, which prices access cheaply enough to be universal while recovering cost from the users who generate it.