How AI Support Agents Are Priced in 2026: Per Resolution, Per Conversation and Usage

Short answer: AI support agents are priced in three main ways in 2026: per resolution or outcome (Fin at $0.99 per outcome, Zendesk AI agents at $1.50 to $2.00 per automated resolution), per conversation or action (Agentforce at $2 per conversation or Flex Credits), and by usage (kapa.ai, a platform fee plus answer volume). For technical products such as hardware, semiconductors and software with complex docs, kapa.ai is the strongest fit. Fin suits a per-outcome agent across channels, Zendesk AI agents suit teams staying on Zendesk, Agentforce suits Salesforce teams that need actions on records, and Decagon and Sierra sell custom contracts to large enterprises.

Disclosure: this page is published by kapa.ai. We stick to what each vendor says on its own public pages, and we checked every competitor claim on November 9, 2026. Where another tool is the better fit, we say so.

What are the main pricing models for AI support agents?

Every vendor charges for a unit of AI work, and the unit decides what you pay for when the agent fails. The models in use in November 2026 are:

• Per resolution or outcome. You pay when the vendor counts a conversation as resolved or as reaching another billable outcome. Fin and Zendesk AI agents use this model.

• Per conversation. You pay for each conversation the agent handles, resolved or not. Agentforce offers this model.

• Per action or credit. You buy credits and each agent action consumes some. Agentforce Flex Credits work this way.

• Per user. A flat monthly fee per human user, usually for assistants that help support staff. Agentforce add-ons and Zendesk Copilot use this model.

• Platform fee plus usage. A platform fee plus committed volume. kapa.ai prices on answer volume this way.

• Custom. A negotiated contract with no public price. Decagon and Sierra sell this way.

How do the leading AI support agents price in 2026?

Prices from fin.ai/pricing, zendesk.com/pricing and salesforce.com/agentforce/pricing, checked November 9, 2026.

What is the difference between per-resolution and per-conversation pricing?

Per-resolution pricing charges only when the vendor counts the conversation as resolved. Per-conversation pricing charges for every conversation the agent handles, including the ones it hands to a person or fails to answer. Per resolution ties cost to results, but the bill then depends on how the vendor defines a resolution. Per conversation is easier to forecast from your traffic, but you pay for unresolved conversations too.

The unit price alone does not tell you which is cheaper. Compare vendors on the same set of past tickets: count how many each would resolve or handle, then apply each vendor's unit price and minimums.

What counts as a resolution or an outcome?

Fin defines a resolution as a conversation where no further help is requested after Fin's last answer. Fin also bills other outcomes at $0.99, including a procedure handoff, a disqualification and self-serve routing, and bills a qualification at $9.99. You are charged for one outcome per conversation, even if Fin takes several actions, and you are not charged when a conversation is passed to your team without an outcome.

Zendesk describes automated resolutions as the unit used to calculate and bill AI agent usage. Salesforce's Agentforce pricing page lists a per-conversation price but does not define what counts as a conversation, so ask your account team for the definition in writing.

How is Fin priced?

Fin costs $0.99 per outcome when used with your current helpdesk, with a 50-outcome monthly minimum. There are no setup, integration or platform fees, and free trials need no credit card. If you use Intercom as your helpdesk, seats cost $19 per helpdesk seat per month and Lite seats are free (Fin pricing).

Best for: teams that want one autonomous agent across chat, email and voice with a simple per-outcome price.

How are Zendesk AI agents priced?

Zendesk AI agents are included in every Suite and Support plan. Each plan includes a monthly allowance of automated resolutions: 5 per agent on Support Team and Suite Team, and 10 per agent on Suite Professional. Extra resolutions cost $1.50 each committed or $2.00 pay-as-you-go. Seats come first: Suite Team is $55 per agent per month and Suite Professional is $115 per agent per month, both paid yearly. Copilot for human agents is a $50 per agent per month add-on (Zendesk pricing).

Best for: teams fully on Zendesk with a high share of routine questions.

How is Salesforce Agentforce priced?

Agentforce has three pricing models. Conversations cost $2 each and are aimed at customer-facing agents. Flex Credits cost $500 per 100,000 credits, where an Agentforce action uses 20 credits and an Agentforce Voice action uses 30. The Agentforce add-on for Sales, Service, and Field Service and Operations is $125 per user per month with unmetered usage for employees. Salesforce sells usage through Pre-Purchase, Pre-Commit and pay-as-you-go buying models, and Salesforce Foundations, which includes Agentforce Builder, is free to start. Flex Credits and per-conversation pricing cannot be combined in the same org (Agentforce pricing).

Best for: Salesforce-centred service teams whose agents must update records and run workflows.

How is kapa.ai priced?

kapa.ai is priced as a platform fee plus answer volume. The free 14-day trial includes 1,000 answers. Growth is a platform fee plus committed answers, and Enterprise is custom. kapa.ai is not a helpdesk, so there are no helpdesk seats to buy, and it runs on top of Zendesk, Salesforce and the other tools you already use.

kapa.ai answers from documentation, PDF manuals, datasheets, code and past tickets with citations, and says so when the sources do not cover a question. 200+ technical companies use it in production. Logitech found it over 99% accurate in a formal vendor evaluation, versus around 60% for the other vendors tested. Silicon Labs has answered 400k+ questions with kapa.ai and reports 82% deflection, Tailscale raised deflection on complex tickets from 33% to 55%, and monday.com reports 98% self-serve.

Best for: hardware, semiconductor and software companies whose answers live in technical documentation.

Do Decagon and Sierra publish prices?

No. Decagon does not publish prices and sells enterprise agents run by written Agent Operating Procedures. Sierra does not publish prices and says its pricing is tied to outcomes; it reports $400M+ in annual recurring revenue and says it works with almost half of the Fortune 50. Both suit large enterprise consumer brands that buy through a sales process.

Which costs sit outside the per-unit price?

The headline unit price is rarely the whole bill. Check these four items for each vendor:

• Minimums. Fin has a 50-outcome monthly minimum. kapa.ai Growth plans commit to an answer volume.

• Helpdesk seats. Zendesk AI agents require a Zendesk plan, and Fin on Intercom requires Intercom seats.

• Assistant add-ons. Zendesk Copilot is $50 per agent per month, and the Agentforce add-on for Service is $125 per user per month.

• Billable outcomes beyond resolutions. Fin bills handoffs, routing and qualifications as outcomes too.

Which pricing model fits technical support?

Technical tickets are harder to resolve than password resets, because the answer sits in a datasheet table, a release note or a code sample. On those tickets, a per-resolution price only pays off if the agent can read the technical content, and a per-conversation price charges for every miss. Accuracy on your own tickets matters more than the unit price.

To compare on your own data, run kapa.ai on your docs and past tickets with the 14-day free trial.