Fin

Solo

suitable with caveats

The floor is 50 outcomes a month, roughly 50 dollars, provided Fin runs on the helpdesk you already have. That route carries no seat cost at all, since the vendor promises unlimited teammates at no extra charge. Only if you also buy the vendor's own helpdesk do you add from 29 dollars per seat per month. Anyone with fewer than 50 cases a month pays for unused volume and solves a problem they do not yet have.

Mid-market

suitable

This is where the model fits best. The bill grows with cases resolved rather than with team size, and in exactly this range the question of whether growth forces you to hire more support staff is a real decision.

Enterprise

suitable with caveats

Technically sound, commercially restless. At five-figure monthly ticket volumes, 99 cents per case turns into a five-figure monthly line item that cannot be capped. On top of that, EU hosting carries documented exceptions and the vendor is in the middle of being acquired.

Off-the-shelf

Seat-priced helpdesk with knowledge base and rule-based bot

The grown route: ticket system, self-service articles, a bot that answers on keywords. The advantage is predictability, since the bill hangs on team size rather than customer growth. The price is a low deflection rate. Every case the rule-based bot does not know reaches a human, so headcount keeps growing with the business.

AI-native

Outcome-billed answering agent, the class this tool belongs to

An agent reads your knowledge base, answers in its own words and escalates on its own. You pay per case. That shifts risk toward the vendor and is more honest than seats, as long as the definition of success holds up. Anyone taking this route should pin down three things: what counts as an outcome, what happens when the rate drops, and whether the agent may run on a third-party helpdesk. Without those three, you are buying a variable bill with nothing promised in return.

Build it yourself

Your own agent on a model API and your own knowledge search

Technically feasible today, and raw model cost per conversation sits below the per-outcome price. What you build and then maintain forever is everything else: connectors for every channel, human handover, answer quality across languages, a regression suite so a new model version does not quietly get worse, and the evidence your regulator will ask for. The rule of thumb is simple: building your own pays off when support is the product, or when the outcome bill permanently exceeds the cost of one engineer.

Entry price

0.99 USD per outcome, meaning per resolution, per procedure handoff and per disqualification. 9.99 USD per qualification. Minimum 50 outcomes a month, which puts the floor at roughly 50 USD. The pricing page sets out two routes that exclude one another: on the helpdesk you already run, Fin costs outcomes only, with unlimited teammates at no extra charge. Take the vendor's own helpdesk instead and you pay from 0.99 USD per outcome plus 29 USD per seat per month. A seat is therefore not a mandatory part of the entry price but a consequence of that choice. Billing is in US dollars only. Fourteen-day trial without a payment method.

Source · As of2026-07-30

What gets expensive

In two places. First the definition: an outcome is billed even when the customer simply stops writing after the last answer. You are paying for silence rather than proven success, and anyone who does not hold that against their own satisfaction measurement is buying a metric instead of a result. Second, growth: at 10,000 requests a month and the advertised rate you land at roughly 7,500 USD monthly for outcomes alone, plus 35 USD per user for the copilot, 99 USD for the analysis of 1,000 conversations, telephony on custom terms and channel fees for SMS and WhatsApp. Run the vendor's helpdesk as well and the seats come on top. That bill grows exactly in step with your customer count, which is the opposite of a fixed cost block.