Issue 02 · The Series

The Consumption Problem

Why seat based pricing breaks the moment AI agents become the buyer, the user, and the decision maker simultaneously.

Part of the Agentic Commercial Model essay series · Fessal Rahman

The industry's first response to the seat pricing problem was to switch to consumption: charge per conversation, per action, per credit. It is an improvement in one sense (usage at least moves in the same direction as engagement) but it trades one flawed proxy for another. Conversations, actions, and credits measure attempts, not outcomes. A model that takes five tries to solve a problem generates five times the billable activity of one that solves it in a single pass, which rewards inefficiency exactly where efficiency should be the point.

The pattern shows up repeatedly among vendors who moved first. One large CRM vendor cycled through three different pricing units in under two years (moving from a flat per conversation charge, to a credit system tied to discrete actions, to a hybrid bundled back into seats) because none of the earlier units gave enterprise buyers a number they could forecast against. A major support software vendor layered variable usage credits on top of an existing seat fee, which solved nothing for finance teams who now had to defend two unpredictable line items instead of one.

Consumption pricing has a second failure mode that only shows up under stress: it collapses in a downturn. When budgets tighten, customers don't necessarily use the product less: they get better at using it, extracting the same outcome for less usage. Two well known consumption priced infrastructure vendors have both seen this play out, where revenue growth decoupled from the value customers were actually getting, in the wrong direction.

A pricing model that survives agentic AI needs three properties that seat pricing and simple consumption pricing both lack: it has to measure something close to the actual outcome delivered, not the number of attempts it took to get there; it has to give enterprise buyers a number they can forecast with confidence a year out; and it has to make vendor revenue and customer value move in the same direction, not opposite ones during optimization. Very few companies have all three today.

In This Issue

Read the Full Essay on Medium → Back to All Writing →