Per-seat pricing is cheapest when a few people each answer a lot of email. Per-conversation pricing is cheapest when several people share a moderate or seasonal volume, because you pay for the work, not the logins. Per-resolution pricing only fits when the AI answers customers on its own and you accept the vendor’s definition of “resolved”. Work out the break-even point with your own numbers before choosing.
What does each pricing model charge for?
| Model | You pay for | Cost goes up when |
|---|---|---|
| Per seat (per user) | Each person with access, every month | You add people, even if they answer little |
| Per conversation (per email) | Each customer email the tool handles, under the vendor’s counting rule | Customers write more |
| Per resolution | Each conversation the AI closes without a person | The AI answers more on its own |
Each model rewards something different. Per seat rewards a few heavy users. Per conversation rewards sharing. Per resolution rewards the vendor for closing conversations, which is why its definition matters so much.
Where is the break-even between per seat and per conversation?
Divide the monthly price per seat by the price per conversation. The result is the number of conversations per person per month at which both models cost the same.
An example with round numbers (use your vendors’ actual prices):
- Per seat: €20 per user per month
- Per conversation: €0.25
€20 ÷ €0.25 = 80 conversations per person per month.
| Your team | Conversations per month | Per seat at €20 | Per conversation at €0.25 | Cheaper |
|---|---|---|---|---|
| 1 person | 300 | €20 | €75 | Per seat |
| 3 people | 150 | €60 | €37.50 | Per conversation |
| 3 people | 600 | €60 | €150 | Per seat |
| 5 people, busy season | 400 | €100 | €100 | Equal |
| 5 people, quiet month | 100 | €100 | €25 | Per conversation |
If each person handles far more than the break-even number, per seat wins. If the work is spread thin across people, or volume swings, per conversation wins. The numbers above only illustrate the calculation; they are not any particular vendor’s prices except the €0.25, which is ReplyPen’s launch price.
What does per-seat pricing hide?
- Occasional users cost the same as daily ones. The colleague who covers holidays needs a seat all year.
- A base plan may be required. Some AI add-ons are only sold on top of a particular office or email plan, which adds to the per-seat price.
- It says nothing about volume. That is good when volume grows, but it also means the price does not fall in a quiet month.
What does per-conversation pricing hide?
- The counting rule. One vendor counts a thread once; another counts every customer email that gets a draft. A thread where the customer writes three times can be one unit or three. What counts as a conversation in AI support pricing? covers the variations.
- Skipped mail. Check that newsletters, receipts and auto-replies are not counted.
- Spikes. A product problem that triples your inbox for a week triples the cost for that week. Ask whether you can limit spend.
What does per-resolution pricing hide?
A resolution usually means the customer did not ask for more help after the AI’s answer. Intercom, for example, describes an “assumed resolution” as a customer who “exits the conversation without requesting further assistance”, and says such a conversation “will be charged” (Intercom Help: Fin outcomes, checked 30 September 2026). So you may pay for conversations where the customer simply gave up.
Per-resolution pricing also only applies when the AI replies on its own, usually in a chat widget or helpdesk. If you want drafts that a person checks and sends from Gmail or Outlook, there is no resolution to charge for.
Which model fits which small business?
| Situation | Usually fits |
|---|---|
| A founder answering everything, with steady volume | Per seat, or the AI in your existing email plan |
| Two to five people sharing support@ | Per conversation |
| Strong seasonal peaks | Per conversation |
| High volume of simple questions, AI answering on its own in a helpdesk | Per resolution, if you accept the definition |
| You are still testing whether AI drafts help at all | Whatever lets you try on your real inbox without a contract |
How do you compare quotes fairly?
- Take one real month of your customer email.
- Count customer emails, threads, follow-ups within threads, and the people who answered.
- Apply each vendor’s counting rule to that month, not to an estimate.
- Add required base plans, minimums and commitments.
- Add checking time, measured during a trial.
The cheapest price per unit is not always the cheapest month.
How does ReplyPen price?
Per conversation: €0.25 at launch, with every feature included and no seats, so you can connect a shared support@ address without paying per person. A conversation is one customer thread in which you use a ReplyPen Draft; follow-ups in the same thread, Drafts you never send and regenerated Drafts do not count, and neither does email skipped by your Rules. See the pricing page for the calculator.
For the full picture of what AI for customer email costs, read How much does AI for customer email cost?.
