AI support pricing: per resolution vs flat plans

How is AI customer support priced?

Most tools use one of three models. Per resolution pricing charges each time the AI resolves a conversation without a person. Per seat pricing charges for each support agent, with AI either included or billed on top. Flat pricing charges one monthly price with an allowance of AI conversations. Many tools combine the first two.

Prices are list prices from each vendor's pricing page, checked 2026-10-06; some seat prices assume yearly billing.

What counts as a resolution?

Each vendor defines it differently, and the definition decides your bill. Intercom counts an outcome when a customer confirms the answer resolved their question, doesn't ask for more help, or Fin completes a workflow. Help Scout counts a conversation resolved by AI Answers without human help. Zendesk counts automated resolutions against an allowance per agent.

Two questions to ask any vendor: does a conversation that ends without the customer replying count as resolved, and do follow up questions in the same conversation count again? The answers can move your real cost more than the headline price.

What does each model cost at real volumes?

Here is the monthly cost for a team of three support seats at three levels of AI resolutions, using list prices. Seat costs are included because per resolution fees always sit on top of them.

The math: three seats at each tool's entry price, plus the per resolution fee for every resolution (Zendesk includes 5 per agent, then charges $2.00 each pay as you go; committed plans are cheaper). Flat plans stay the same across these rows: Ferndesk is $149/mo with 1,000 AI conversations included, and usedocs is $99/mo with 2,000 included and 5 seats.

Is per resolution pricing fair?

In one sense, yes: you pay when the AI does the job. That aligns the vendor with results, and at low volume it's often the cheapest option. The catch is that success makes it more expensive. The better your docs and your AI get, the more conversations resolve, and the bigger the bill, which makes the cost hard to budget and punishes exactly the investment you want to make.

It also creates an odd incentive around docs. Writing a great article that deflects a hundred questions a month adds a hundred billable resolutions.

When is a flat plan better?

A flat plan is better when your volume is steady or growing, when you want a predictable bill, and when you'd rather spend on better docs than on metered answers. It's worse if your volume is tiny and you'd only use a fraction of the allowance, or if you need far more conversations than the plan includes.

Check what happens at the limit. Some flat plans charge for extra conversations; usedocs keeps answering with docs search instead of AI past the 2,000 allowance, so the bill doesn't change, and adding a second bot costs $29/mo.

Resolutions and conversations aren't the same unit

Be careful comparing per resolution prices with flat conversation allowances, because they count different things. A per resolution tool only bills conversations the AI resolved; unresolved ones that go to a person are free on the AI side but cost agent time. A conversation allowance counts every AI conversation, resolved or not.

usedocs counts AI conversations, with follow up questions in the same conversation free. If, say, two out of three AI conversations resolve, 1,000 resolutions correspond to about 1,500 conversations, still inside the 2,000 allowance. Your own resolution rate is the number to plug in.

What else affects the real cost?

The AI fee is rarely the whole bill. Seats, minimums, add ons, and the cost of wrong answers all matter.

• Seats: per seat tools charge for every agent; flat plans usually include a few.
• Minimums: some vendors require a minimum monthly commitment, especially when you use their AI with another helpdesk.
• Add ons: AI features are sometimes a separate add on, as with Help Scout's AI Answers.
• Wrong answers: an AI that guesses resolves conversations it shouldn't, and you pay for the resolution and for the follow up when the customer comes back. An assistant that refuses when the docs are silent avoids that.
• Docs upkeep: every AI answers from your docs. Stale docs mean wrong answers whatever the pricing model.

How should you choose?

Estimate your monthly AI conversations, your resolution rate, and your seat count, then run each candidate's pricing for this month and for twice this volume. Pick the one that stays sensible at twice the volume, because that's where you're headed if the AI works.

1. Count last month's support conversations.
2. Estimate the share an AI could answer from your docs today.
3. Price each tool at that volume and at double it.
4. Add seats, minimums, and add ons.
5. Ask each vendor what counts as billable, in writing.

Where does usedocs fit?

usedocs is one flat plan: Growth at $99/mo, or 20% less billed yearly, with 5 seats and 2,000 AI conversations a month pooled across the workspace, follow ups free. The assistant answers from your help center with a link to the source and says so when the docs don't cover a question. It also keeps those docs current from your code and your customers' questions, which is what keeps the answers right.

Related reading