Uncategorized

AI Marketing Agent Pricing Models: Per Action, Per Outcome or Per Seat

Two quotes for the same AI agent can look similar on a price list and diverge once you apply your real volume. Vendors now sell agents by the action, by the outcome, by the conversation or by the seat, and each model puts the cost risk in a different place. Comparing them requires a common unit, not a price list.

Key Takeaways
  • Salesforce lists Agentforce at $500 per 100,000 Flex Credits, with a standard action costing 20 credits (about $0.10), or $2 per conversation.
  • Intercom charges $0.99 per Fin outcome, and counts an outcome when a customer stops asking for help, not only when they confirm resolution.
  • Gartner predicted in June 2025 that over 40% of agentic AI projects will be canceled by end of 2027, citing escalating costs, unclear value or weak risk controls.
  • Convert every quote to cost per business result at your own volume before comparing.

Scope note: the two public price lists we could verify, from Salesforce and Intercom, are for CRM and customer-service agents. Marketing-agent vendors are adopting similar structures, but we did not find comparable published rate cards for SEO or ad-management agents, so use these as models of the pricing logic, not as marketing benchmarks.

Per-action pricing

Per-action pricing charges for each step an agent takes. Salesforce’s Agentforce pricing page lists Flex Credits at $500 per 100,000 credits, and states that a standard Agentforce action uses 20 Flex Credits, while a voice action uses 30. At those rates a standard action costs about $0.10, by simple division.

Flex Credits: a shared pool of prepaid or metered usage units that any agent action draws from, rather than a fixed fee per agent.

Salesforce also sells the same capability per conversation at $2, and per user, from $5 per user per month plus credits up to $125 per user per month for add-ons. Buyers can pre-purchase usage, pre-commit to a baseline, or pay as they go. The page notes that pricing is subject to change.

The implication is that a per-action model charges for effort, not results. An agent that takes eight steps to answer one request costs more than one that takes two, so cost scales with how the agent is configured.

Per-outcome pricing

Per-outcome pricing charges only when the agent reaches a defined result. Intercom’s pricing page lists $0.99 per Fin outcome, with no seat costs and no setup or platform fees when Fin runs on an existing helpdesk. A minimum monthly commitment applies, and the page gives 50 outcomes as an example.

The definition of the outcome is the contract. Intercom counts one when a customer confirms the issue is resolved, when the customer stops asking for help after Fin responds, or when Fin completes a workflow, including handoffs. Each conversation is charged once.

That second condition deserves attention. A customer who gives up and a customer who is satisfied can both generate a charge, so a billed outcome is not proof of a solved problem. Pair any outcome-priced agent with your own quality sampling, so you pay for results you would also count as results.

The models compared

The table summarizes where each model places risk. It reflects the structure of the vendors named above and general tradeoffs, not a ranking.

ModelYou pay forCost riskAsk the vendor
Per action (credits)Each step the agent takesVerbose agents and retries raise the billHow many credits does a typical task use?
Per conversationEach interaction, whatever the resultFailed conversations still cost moneyAre escalated chats billed?
Per outcomeA defined resultLoose definitions inflate the countExactly what counts as an outcome?
Per seatNamed users, plus usage in some plansPays the same when the agent is idleIs usage billed on top of the seat?

Why cost control matters

Gartner’s June 25, 2025 press release predicted that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value or inadequate risk controls. It also described “agent washing,” where vendors relabel existing products as agentic, and estimated that only about 130 of the thousands of agentic AI vendors are real.

This is a prediction from 2025, not new 2026 data, and the release is about enterprise agentic projects in general. Still, it names cost and value as two of its three failure causes, which are the two a pricing model controls.

A usage-based price makes the first risk visible and the second testable, because you can see spend and results move together. A flat seat fee hides both until renewal. For that reason, we suggest tying any pilot to a spend cap, as in our 90-day pilot scope.

How to compare quotes

Converting each quote to a single unit exposes the real difference. As an illustration only, $500 buys about 5,000 standard Agentforce actions, 250 Agentforce conversations, or roughly 505 Fin outcomes at list price, so which is cheaper depends entirely on how many actions a task needs and how you define a result.

  1. Define the business result in your own terms, such as a qualified lead or a resolved enquiry.
  2. Estimate monthly volume and the average number of agent steps per result from a small trial.
  3. Convert each quote to cost per result at that volume, including minimum commitments and add-ons.
  4. Ask for the billing definition in writing, including how handoffs, retries and abandoned chats are counted.
  5. Set a monthly spend cap and a review date before signing.

Our guide to the business case for an AI marketing agent covers the questions finance will ask about these numbers.

Frequently Asked Questions

What is the difference between per-action and per-outcome pricing?

Per-action pricing bills each step an agent takes, so cost follows effort. Per-outcome pricing bills only when a defined result occurs, so the contract definition of that result decides the bill.

How much does Salesforce Agentforce cost per action?

Salesforce lists Flex Credits at $500 per 100,000 and states a standard action uses 20 credits, which works out to about $0.10 per action. It also offers $2 per conversation, and says pricing is subject to change.

Is outcome-based pricing always cheaper?

No. It depends on how the vendor defines an outcome. Intercom counts a customer who stops asking for help, so the billed count can exceed confirmed resolutions. Check the definition before comparing price.

Pricing structure is only half of an agent decision; the other half is whether the work it does moves a number you care about. At Grow with Hafsa we help US and UK teams scope AI marketing agent pilots with measurable cost and result targets. Our checklist for evaluating an agentic marketing vendor is a good next read before requesting quotes.

Related reading: If you are buying GEO work specifically, see what to put in the engagement scope, from deliverables to reporting.