Per-resolution pricing is cheapest when volume is low or unpredictable, because you pay only for successes. Seat-based and committed-volume deals usually win at high volume. The crossover point depends on your resolution rate, which is why you should model both before you sign.
The pricing models, in plain terms
| Model | You pay for | Example | Risk to the buyer |
|---|---|---|---|
| Per resolution | Conversations the AI closes | Intercom Fin at US$0.99 | The definition of "resolved" decides the bill. |
| Per conversation | Every conversation the AI handles | Salesforce Agentforce at US$2 | You pay for failures as well as successes. |
| Per session | A time window of interaction | Freshworks Freddy at US$0.49 | A returning customer can start a new billable session. |
| Seats plus usage | Human agent seats, plus AI usage | Zendesk from US$55 per seat | The AI price can be the part that is not published. |
| Quote only | A negotiated annual contract | Decagon, Sierra | You cannot compare before you talk to sales. |
How the bill moves with volume
Take 10,000 conversations a month with half resolved by the AI. At US$0.99 per resolution that is 5,000 billable units. At US$2 per conversation all 10,000 are billable, so the second bill is roughly four times the first even though the unit prices look similar. Now raise the resolution rate to 80%: the per-resolution bill grows by 60%, while the per-conversation bill does not change. Better performance makes one model more expensive and the other better value.
Our cost calculator runs these numbers for every product that publishes prices, at any volume and resolution rate you choose.
Costs that sit outside the unit price
- The platform underneath. Several agents only run on the vendor's own helpdesk, so the seats are part of the real price.
- Minimums and prepaid packs. Unused prepaid resolutions or sessions usually expire.
- Overage rates. Usage above your commitment is often billed at a higher rate, sometimes automatically.
- Implementation. Enterprise platforms are often sold with services. Procurement data from Vendr puts the median Decagon contract at US$432,750 a year and the median Ada contract at US$72,000.
A rule of thumb
Below a few thousand conversations a month, choose a model where you pay for outcomes and can leave without a penalty. Above tens of thousands, ask every vendor for a committed-volume quote and compare the effective cost per resolved conversation, using your own measured resolution rate rather than the vendor's average.
Frequently asked questions
Is per-resolution pricing always the fairest?
It aligns the vendor's revenue with your success, which is good, but it also gives the vendor a reason to define "resolved" generously. Read what counts as a resolution before you compare prices.
Why do so many vendors hide their prices?
Enterprise deals bundle software, implementation and volume discounts, so vendors prefer to quote. It also makes comparison harder for the buyer. Our rubric scores pricing transparency for exactly that reason.
Keep reading
- What counts as a resolution?guide
- Resolution rate vs deflection rateguide
- How to run an AI agent pilotguide
- Agentic customer service, explainedguide
- The best AI customer service agents in 2026best list
This page was researched and drafted with AI assistance from the sources listed on it. We have not run hands-on tests of these products. Method: How we review