"Per-Outcome" AI Pricing Is Worse for a Small Team Than Per-Seat
The vendor killed the per-seat license and replaced it with a charge per resolved ticket or per closed conversation. It sounds fairer: you pay for results, not chairs. Run the arithmetic on a small support queue and the new model is a meter you can't read until the invoice arrives.

The Per-Seat License Is Dead. Read the Replacement Carefully.
The conventional wisdom in software has been that per-seat pricing is the enemy: you pay for ten licenses and use six, so the vendor profits from your slack. Outcome-based AI pricing was sold as the correction. You stop paying for seats and start paying for results, a resolved support ticket, a closed conversation, a qualified lead. Pay for what works. It is a genuinely appealing pitch, and for a small operator it is often the more expensive deal, because it swaps a cost you control for a cost you cannot predict.
Here is what is actually happening. The vendor did not lower your bill. The vendor moved your bill onto a meter, and handed you a product designed to spin that meter.
The Numbers, Out Loud
Take the two names everyone cites. Intercom's Fin AI Agent charges $0.99 per resolution, with a lead-qualification outcome running $9.99, and a floor of 50 outcomes a month. Salesforce Agentforce lists $2 per conversation, and separately sells consumption "Flex Credits" at $500 per 100,000 credits, where a standard agent action burns around 20 credits, roughly a dime an action.
Do the arithmetic on a modest support queue. A small business handling 2,000 customer conversations a month, of which the AI resolves 1,200, is looking at roughly $1,188 a month on Fin at 99 cents a resolution. The old per-seat world might have put three human-equivalent support licenses at a few hundred dollars total. The outcome model is not automatically cheaper; it is cheaper only if your volume stays low, and the entire promise of the product is that your volume of AI-handled interactions goes up. The pricing rewards the vendor precisely when the tool succeeds.
The Meter You Can't Read Until the Invoice
Per-seat pricing has one great virtue that gets no respect: it is predictable. Ten seats cost the same in a quiet March and a frantic December. You can put the number in a budget and forget it. Outcome pricing throws that out. Your cost now floats with your demand, and demand spikes are exactly when cash is tightest and support volume is highest. A viral complaint, a product recall, a holiday rush, each one is a support surge, and under outcome pricing each surge is also a billing surge, arriving in the same month the underlying problem did.
Worse, the unit is defined by the vendor, not you. Intercom bills once per conversation no matter how many messages it takes, which sounds fair until you notice the vendor decides what counts as a resolution, a handoff, or a disqualification, and each carries its own price. Salesforce runs three parallel pricing models at once, flex credits, per-conversation, and per-user, which is not generosity. It is a company that has not settled on what to charge, changing the meter while you are standing on it.
When Outcome Pricing Actually Wins
There is a real case for it, and honesty requires naming it. If your interaction volume is genuinely low and spiky, a handful of support tickets some months and none in others, paying 99 cents when something happens beats paying for a full seat that sits idle. Outcome pricing is a good deal for the occasional user and a bad deal for the heavy one, which is the exact inverse of per-seat. The question is not which model is fairer in the abstract. It is which side of that line your actual usage sits on.
So model it before you switch. Take last year's real interaction counts, month by month, spikes included, and price them under both schemes. Not the vendor's illustrative example, your numbers. If your volume is high or growing, per-seat or a capped plan will usually cost less and, more importantly, cost a knowable amount.
Run the Number Before the Renewal
The reason outcome pricing spread so fast is not that it is fairer. It is that it uncaps the vendor's revenue: every improvement in the AI's capability becomes more billable events rather than more value delivered at a flat price. That is a good business model for the seller and a variable-cost trap for a small buyer who liked knowing the bill in advance.
Before you sign, demand a cap or a committed-volume tier, so a bad month can't produce a runaway invoice. And keep the per-seat quote in writing next to the outcome quote, because the outcome model only looks cheaper until the tool does exactly what you're paying it to do, at which point it does the expensive thing on purpose. The sales deck calls that alignment. Your accountant will call it something shorter.

