The units vendors use
As of the dates on each comparison page, the main ones are:
- Per resolution or outcome: you pay when the AI finishes a conversation without a person. Fin (formerly Intercom) lists $0.99 per outcome, Help Scout $0.75 per resolution, and HubSpot announced $0.50 per resolved conversation for its Customer Agent.
- Per session: you pay whenever the AI replies, whether or not it solved anything. Freshdesk lists 49¢ per Freddy AI Agent session after a one-time starter pack.
- Per conversation: you pay for each conversation the AI handles. Front lists AI from $0.05 per conversation.
- Per interaction: Gorgias prices AI by automated interaction on top of its ticket fees; third parties report roughly $0.90 to $1.00.
- Included allowance, then a cap: a number of AI answers is part of the seat and the AI stops at the limit. That is how Flatdesk works: 100 per agent per month, pooled.
Questions to ask any vendor
Put these in writing before you rely on a number.
- What exactly counts as one? Does a hand-off to a person count? A customer who never replied?
- Is there an allowance in the seat, and is it per agent or per account?
- What happens at the limit: the AI stops, or you are billed for overage automatically?
- Can you see each charged conversation, and dispute one the AI got wrong?
- Does the rate change with volume, and can it change mid-contract?
- Are the AI features gated to higher plans?
How to estimate your bill
Take the tickets you get in a month, multiply by the share you expect the AI to finish, and multiply by the vendor's unit price. Then try it at half and at twice that share, because the share is the number you are least sure of and the bill moves one for one with it.
A worked example: 1,500 tickets a month with the AI finishing 30% is 450 resolutions. At $0.99 that is about $446 on top of seats; at $0.50 it is $225. On a plan that includes 100 per agent for 5 agents (500), it is the seat price and nothing more. The bill calculator does this for the vendors in the comparison pages.
Why the unit matters more than the rate
A lower rate counted per session can cost more than a higher rate counted per resolution, because sessions include every conversation the AI touched. Ask what the unit is before you compare rates.
Test the AI before you commit
A resolution count is only worth paying for if the AI is good. Run it on a sample of your own past tickets and read what it would have said. The guide on testing AI support covers how.