Intercom Fin

What the price actually is

$0.99 per Fin outcome, running standalone against an existing helpdesk with no seats. The $297 here is NOT a published price: it is 300 outcomes at $0.99, so this row can sit in a table of monthly figures. Change the volume and the row changes with it. Fin is also bundled into Intercom seat plans at $29, $99 and $139 per seat.

The usage this is priced against

300 conversations resolved a month. Each resolution is assumed to take several model turns over retrieved help-centre content — roughly 8,000 tokens of context and 500 of reply — plus keeping that content embedded for retrieval.

Someone using it half as much sees double the multiple. The assumption is the argument — if you disagree with it, the number below is not about you. Change it in the calculator.

The cost math, in full

Rates come from a dated rate card of representative published API prices, not from the vendor. Nobody outside these companies knows what they actually pay; volume discounts and in-house serving both push real costs below these figures, which makes every multiple here a floor rather than a ceiling.

Why this verdict, not the number

This is the entry that most tests the site's own method. Divide the price by the tokens and the multiple is large. But the comparison the arithmetic implies — that you could do this with an API key — is the wrong one twice over. First, the vendor absorbs every conversation Fin fails to resolve, so the price includes all the attempts you are not billed for. Second, the thing being replaced is a person answering a ticket, which costs many times $0.99 in any market. Outcome pricing also inverts the usual incentive: a vendor paid per resolution loses money when its model is bad, which is a stronger quality guarantee than any benchmark. The multiple is real and the verdict still goes the other way.

What the price buys besides tokens

• Eval harness
• Integrations
• Orchestration

Deciding whether an answer actually resolved a conversation is the hard problem, and the entire pricing model rests on getting it right — that is an evaluation harness with money attached. Around it sit helpdesk integrations, escalation handling, and the retrieval layer over a customer's own documentation.

What you lose if you leave

Billing that only charges for outcomes, the judgement about when to escalate to a human, and connectors into the helpdesk you already run. A self-built bot bills you for every failed attempt, because your API provider charges by the token regardless of whether the customer went away happy.

The honest cheaper path

A retrieval bot over your help centre with your own key is an afternoon's work and would cost a fraction of this per conversation. It will also answer confidently when it should escalate, and you pay for those answers too — which is the difference the price is charging for.

Recompute it for yourself

Where every number came from

• pricingIntercom pricing
• rate-cardAnthropic API pricing
• rateRate card entry: frontier-mid
• rateRate card entry: embedding-standard

Price recorded 18 Sept 2026 · entry last reviewed 18 Sept 2026. Think something here is wrong? File a correction — we publish them, including the ones that embarrass us.