Agentforce vs Fin vs Decagon vs Sierra pricing: who actually publishes a rate

Of the four AI support agents most often shortlisted together, only two publish a rate you can read without a sales call. Salesforce prints six Agentforce buying models, including Flex Credits at $500 per 100,000 and Conversations at $2 each. Fin prints $0.99 per outcome and $9.99 per qualification. Decagon and Sierra publish nothing: both decagon.ai/pricing and sierra.ai/pricing returned a 404 when we checked them. The twist is that the two vendors who do publish are becoming one company, because Salesforce agreed to buy Fin in June 2026.

Last updated August 2026. Every figure below was read first-hand on 25 August 2026.

Which AI support agents publish a price?

We checked all four pricing pages the same afternoon, from a US connection, with currency set to US Dollar where a selector existed. Two loaded with real numbers on them. Two did not load at all.

A 404 is not evidence that a vendor is expensive. It is evidence that the vendor wants the number set in a negotiation rather than read off a page, which usually means enterprise deal sizes and annual commitments. Both Decagon and Sierra sell to large support organizations, so this is ordinary for their tier. It just makes them impossible to compare on paper, and that is worth knowing before you put four names on a shortlist and promise your CFO a comparison by Friday.

How much does Fin cost per resolution?

Fin charges $0.99 per outcome. An outcome is one of four things: a resolution, a procedure handoff, a disqualification, or a qualification. The first three are $0.99 each. Qualifications are $9.99 each, which is a ten times jump most people miss on the first read. There is a 50-outcome monthly minimum, and a conversation that Fin simply passes to a human without achieving an outcome is not billed.

Seats sit on top. Running Fin on Intercom's own help desk costs $29 per helpdesk seat per month. Running it on a third-party help desk is billed at $0.99 per outcome with unlimited teammates and no seat cost, though minimum commitments apply. Copilot, the assistant for your human agents, is $35 per user per month, and a Pro add-on starts at $99 per month. Fin offers a 14-day free trial with unlimited outcomes and bills in US dollars only.

How much does Agentforce cost per conversation?

Agentforce is the opposite problem: not too little information, too much. Salesforce publishes six ways to buy the same agent. Flex Credits are $500 per 100,000, which makes one credit half a cent, a standard agent action 20 credits ($0.10) and a voice action 30 credits ($0.15). Conversations are a flat $2 each. An Agentforce User License is $5 per user per month and still requires credits. The unmetered add-on is $125 per user per month, or $150 for Industries. Agentforce 1 Editions start at $550 per user per month. Salesforce Foundations lets you build and test at $0.

The break-even between the two consumption models is exactly twenty actions. Twenty standard actions cost 20 x 20 credits = 400 credits = $2.00, which is precisely the per-conversation price. Below twenty actions per conversation, Flex Credits win. Above it, the flat $2 conversation wins.

One constraint deserves more attention than it gets. Salesforce's own FAQ states that Flex Credits and Conversations will not be supported in the same org. You choose one model for the entire organization and switching later means swapping everything across. Salesforce also lists granular agent analytics as available only with Flex Credits, so picking the simpler flat rate quietly trades away reporting depth. We walk through the rest of the buying models on our Agentforce alternatives page, and the estimating method in more detail in how to estimate an Agentforce bill.

What does Sierra actually charge?

Nobody outside Sierra's customers knows, and you should be skeptical of any article that tells you otherwise. Sierra is publicly and deliberately outcome-priced: its CEO Bret Taylor has argued on the record that a human-handled service call costs a company roughly $10 to $20 in labor, and that Sierra should charge a portion of the cost it removes when the agent resolves the case instead. That is the philosophy, and it is a coherent one.

What Sierra has never done is confirm a rate. Third-party write-ups circulate figures in the $1 to $2.50 per resolution range, along with six-figure annual contract estimates. We are not repeating those as fact, because Sierra has not published or confirmed them and neither has anyone with access to a signed contract. Treat every Sierra number you find online as an unverified secondhand estimate until Sierra tells you one directly. The same applies to Decagon.

Agentforce vs Decagon: which one can you actually price?

Only one of them. Salesforce publishes six separate buying models for Agentforce, including Flex Credits at $500 per 100,000 credits and Conversations at $2.00 each, so you can build a budget from vendor-published numbers before you ever speak to sales. Decagon publishes nothing: decagon.ai/pricing returns a 404, re-confirmed on 1 September 2026.

That asymmetry matters more than any feature comparison at this stage. With Agentforce you can check a quote against a list price and argue about it. With Decagon every number in the conversation originates from Decagon, and the per-resolution figures circulating in third-party write-ups are not vendor-confirmed. If you are running a formal procurement process, ask Decagon for a written rate card with the unit definition attached, and treat any figure you found in a blog post as unverified.

One practical note on the Agentforce side: Flex Credits and Conversations cannot both run in the same Salesforce org, so this is an organization-wide decision rather than a per-team one, and granular agent analytics are only available on the Flex Credits model. The break-even between them is exactly 20 actions, because a standard action costs 20 credits at $0.10 against $2.00 for a conversation.

Which one is cheapest for 5,000 resolved conversations a month?

Here is the same workload priced with each vendor's published rate. Assume 5,000 conversations a month that the AI resolves on its own, and five human agents behind it. Blank cells are blank because the vendor publishes nothing, not because we could not be bothered.

Two things fall out of that table. The first is that the widest gap is not between vendors, it is inside one vendor: the same Agentforce workload costs $1,500 or $10,000 depending purely on which consumption model you signed, a spread of nearly seven times. The second is that Agentforce on Flex Credits genuinely undercuts Fin at typical service action counts, which is not the intuition most people carry into the evaluation.

Both caveats matter. The Agentforce figures exclude licensing, and Salesforce prints its own footnote saying its examples exclude Data 360 credits and other consumption services, so treat them as a floor rather than a quote. They also assume you are already paying for Salesforce, because Agentforce is built around Salesforce data and permissions. If your customer record does not live there, you are buying an integration advantage you cannot use. Fin's number, by contrast, is close to all-in for the AI itself.

Does outcome-based pricing actually save money?

It saves money when volumes are low and it stops saving money exactly when the product starts working. That is the structural problem with every per-resolution meter: the better your agent gets at deflecting tickets, the more you pay, and the savings you were promised arrive as a bigger invoice. A team that doubles automated resolutions from 5,000 to 10,000 has done everything right and just added $4,950 to a Fin bill.

Outcome pricing is still the fairest model when you are starting out, because you genuinely pay nothing for an agent that does not work. It is worth pressure-testing at the volume you expect in eighteen months, not the volume you have today. If you are still narrowing a shortlist, it helps to look past the four names everyone lists and browse the wider field of ready-made AI agents by category before you commit to a platform decision that is hard to reverse.

What to ask on the demo call

1. What exactly counts as a billable event? Get the definition in writing. Fin bills four outcome types at two very different rates. Agentforce bills actions, and one customer question can be several actions.
2. What sits outside the meter? Salesforce's own footnote excludes Data 360 credits and other consumption services from its examples. Ask what else is excluded.
3. Can we change pricing models later? With Agentforce the answer is that Flex Credits and Conversations cannot coexist in one org, so ask what a switch costs operationally.
4. What is the minimum commitment? Fin publishes a 50-outcome monthly minimum and notes that minimum commitments apply on third-party help desks. Enterprise vendors negotiate annual floors.
5. What does this bill look like at three times today's volume? Ask them to model it. A vendor who will not put that number on paper is telling you something.

Where this leaves a smaller US team

If you are running a support operation with thousands of monthly conversations and a Salesforce estate behind it, Agentforce on Flex Credits is a defensible answer and the arithmetic above is genuinely competitive. If you are on Intercom already, Fin is the least disruptive option and the only one you can start on a free trial today. Decagon and Sierra are worth a call if your deal size supports enterprise procurement, and worth skipping if it does not.

If your objection is the meter itself rather than any one vendor, the shape you want is a flat subscription that does not charge more when the agent succeeds. That is how we built MessageAgent: one AI persona across SMS, WhatsApp, Instagram and Facebook DM, web chat, and email, with a single subscription and carrier fees passed through at cost. Being straight about it, MessageAgent is pre-launch and signing up joins a waitlist, so it is not a candidate if you need something in production this quarter. When you are comparing the field on cost rather than on features, our breakdown of AI customer service agent pricing counts the meters across the whole category, and the Intercom pricing and competitors page covers the Fin side in more depth.