Blog

AI support agent pricing: per resolution, credits or seats?

Published October 26, 2026 · By Jose · 6 min read

  • AI ROI
  • Customer Support
  • CX Operations
Illustration of the ghost mascot examining a short "Per action" invoice and a long "Per conversation" invoice beside a stack of 20,000 chats, under the title "Same work, a $32,000 difference".

Quick summary

  • AI support agents are priced in three ways: per resolution or outcome, per conversation or action (often through prepaid credits), and per seat for the humans who work alongside the AI.

  • Published list prices in October 2026: Fin charges $0.99 per outcome, Zendesk $1.50 to $2.00 per automated resolution, and Salesforce Agentforce $2 per conversation or $0.10 per action.

  • The definition of a resolution moves the bill more than the price does. "No further help requested" and "resolved without a human" are different contracts.

  • At 20,000 conversations a month with half resolved by the AI, the same workload costs between about $8,000 and $40,000 a month depending on the model.

  • Zentesimal charges only for conversations the AI resolved, inferred from the conversation or confirmed by the customer. When the outcome is unclear, the conversation isn't charged and is marked as uncertain.

AI support agents (software that answers customers and can act in other systems) are priced per resolution, per conversation or action, or per seat. Per resolution ties the bill to results. Per conversation and per action tie it to volume. Per seat covers the humans who work alongside it.

The model matters more than the headline price. Two vendors at a similar rate can produce bills several times apart, because they count different things.

The price per unit is on the website. The unit itself is in the contract.

What are the pricing models for AI support agents?

There are three models, and most vendors combine at least two. The table uses list prices published on each vendor's pricing page in October 2026, before platform fees and discounts.

AI support agent pricing models, October 2026

Model

What you pay for

Published example

What it rewards

Per resolution or outcome

Chats the AI closes successfully

Fin: $0.99 per outcome. Zendesk: $1.50 per automated resolution committed, $2.00 pay as you go

Resolving, as the vendor defines it

Per conversation

Every conversation the AI handles

Agentforce: $2 per conversation

Volume handled, resolved or not

Per action, through credits

Each step the AI takes, from a prepaid credit pack

Agentforce: 20 Flex Credits per action, $0.10 at $500 per 100,000 credits

Fewer, cheaper steps per interaction

Per seat

Each human using the helpdesk or an AI copilot

Fin with Intercom's helpdesk: $29 per seat per month on top of outcomes

Headcount

Sources: Fin pricing, Zendesk pricing, Salesforce Agentforce pricing.

Enterprise vendors such as Sierra and Decagon don't publish rates. Their pricing is negotiated per contract.

Diagram of what each AI support agent pricing model bills for the same 10 chats, 5 resolved by the AI and 5 not: per resolution bills the 5 resolved chats, per conversation bills all 10, and per action bills every step of every chat.

Why does the definition of a resolution matter more than the price?

The definition decides which interactions you pay for. Each vendor writes its own, and they don't match.

  • Fin counts a resolution when "no further help is requested after Fin's last answer". It also bills a handoff to a person when it completes a configured procedure, and charges one outcome per conversation at most.
  • Zendesk bills requests "successfully resolved by the AI agent, without any escalation to a human agent".
  • Agentforce per conversation bills the conversation whether it ends resolved or not.

A user who gets an answer, says nothing and churns next week counts as resolved under a "no further help requested" rule.

Our guide on why "not escalated" is a misleading success metric covers that blind spot from the quality side.

How does the cost change as volume grows?

Cost grows in line with volume in every approach, but at different slopes. The example below takes a support team handling 5,000 and then 20,000 AI conversations a month. Half of them are resolved by the AI, with four AI actions per chat on average.

Monthly cost of the same workload under each model (illustrative)

Model, list price

5,000 conversations

20,000 conversations

Per outcome, $0.99 (Fin)

$2,475

$9,900

Per automated resolution, $1.50 committed (Zendesk)

$3,750

$15,000

Per automated resolution, $2.00 pay as you go (Zendesk)

$5,000

$20,000

Per action, 4 × $0.10 (Agentforce credits)

$2,000

$8,000

Per conversation, $2 (Agentforce)

$10,000

$40,000

The figures leave out platform fees, seats, monthly minimums and included allowances, such as Zendesk's 5 to 10 automated resolutions per agent per month. They show the slope of each scheme and are no substitute for a quote.

At 20,000 conversations, the cheapest and the most expensive option are $32,000 a month apart for the same work.

Per action looks cheapest in the table, but it bills unresolved tickets too.

If the AI resolves 30% instead of 50%, the per-outcome bills fall and the per-action bill stays the same.

What does per-seat pricing still pay for?

Per-seat pricing pays for the people. The helpdesk seats of the human team and any AI copilot (an assistant that drafts replies for a human agent) are usually billed per person per month.

Seats and AI usage are separate lines on the same invoice. A team that moves half its conversations to an AI agent keeps paying for the seats until the headcount or the plan changes.

How Zentesimal charges for its AI agents

Zentesimal gives you a team of AI agents for support, conversion, expansion and retention. It charges per conversation the AI resolved. Chats the AI only took part in are free.

What counts as a resolved conversation

The AI's work counts as a resolution in one of two ways: the result is clear from the exchange itself, or the customer confirms it.

When the outcome is unclear, nothing is charged. It is marked as uncertain instead, and stays visible for review.

If nobody can tell whether the AI solved it, the customer shouldn't pay for it.

Optional consultancy for enterprise customers

Enterprise customers (100k+ users) can add optional consultancy on top. It is not included by default.

The Zentesimal team works with them on getting the AI agent to the outcomes they set, from its first skills (the written instructions it follows for each type of request) to the changes that follow. It is separate from usage.

What to ask before signing an AI agent contract

Five questions decide most of the bill, and every vendor can answer them in writing.

Questions that decide the cost of an AI support agent

  1. What exactly counts as a resolution? Silence after an answer, customer confirmation, or no handoff to a person.
  2. What happens when a resolved conversation comes back? Whether a reopened case within a week is billed twice.
  3. What is billed when the AI hands over to a person? Some contracts bill the handoff as an outcome.
  4. What are the fixed costs? Platform fees, monthly minimums (Fin sets 50 outcomes a month with an external helpdesk) and seats.
  5. What happens to uncertain outcomes? Whether they're billed by default or left out.

The answers turn a list price into a monthly cost.

Multiplying expected AI conversations by the expected resolution rate and the price per unit, then adding fixed fees, gives a working estimate. The support use case shows where the AI agent sits in that workload.

Recap

AI support agents are billed per resolution, per conversation or action, and per seat for the humans. Per resolution follows results, per conversation and per action follow volume, and seats follow headcount.

The vendor's definition of a resolution shapes the invoice more than the unit price, so the questions about reopened cases, handoffs and uncertain outcomes come before the price comparison. Zentesimal bills only resolved conversations and leaves uncertain ones out.

Book a demo

FAQ

How much does an AI customer support agent cost?

Published list prices in October 2026 range from $0.99 per outcome (Fin) and $1.50 to $2.00 per automated resolution (Zendesk) to $2 per conversation or $0.10 per action (Salesforce Agentforce). Platform fees, seats and minimums come on top, and enterprise vendors such as Sierra and Decagon negotiate prices per contract.

What is outcome-based pricing for AI agents?

Outcome-based pricing charges only when the AI agent achieves a defined result, usually a resolved conversation. The vendor's definition of that result, such as "no further help requested" or "resolved without a human", decides which conversations are billed.

Is per-resolution pricing cheaper than per-conversation pricing?

At the same rate, per resolution is cheaper because unresolved conversations are free. In practice it depends on the rates and the resolution rate: at $0.99 per outcome and 50% resolved, 20,000 conversations cost about $9,900, against $40,000 at $2 per conversation.

What should count as a resolved conversation?

A conversation where the customer's problem was solved, shown by the conversation itself or confirmed by the customer. Silence alone is weak evidence. A fair rule leaves conversations with an unclear outcome unbilled.

Do I still pay for seats with an AI support agent?

Usually yes. Helpdesk seats for the human team and AI copilots for human agents are billed per person, separately from what the AI agent resolves. Fin with Intercom's helpdesk, for example, adds $29 per seat per month to its outcome price.

See how CX teams deploy Plays they actually rely on.