Contents
- Verdict
- What is Intercom Fin?
- Best for, and not for
- Strengths and limits
- Pricing
- What people actually build with it
- What Intercom Fin actually costs at small-business volume
- The gotcha: a procedure handoff is billed, an ordinary escalation is not
- What breaks in production, and how you find out
- What it connects to that you already run
- Intercom Fin alternatives: what to pick instead, and when
- Migration and lock-in: can you get your work out?
- FAQ
Fin is the most honestly priced agent in this group and, at scale, one of the most expensive. It bills $0.99 per outcome rather than per conversation, so a chat Fin cannot handle and simply passes to your team costs nothing. That alignment is real and rare. The catch is arithmetic: at 700 resolutions a month you are paying $693 before you have paid for a helpdesk, which is more than the other four in this batch and more than most flat-rate chatbots by an order of magnitude. Fin is worth it when a resolution genuinely displaces human handling time. Below roughly 200 resolutions a month, a flat-rate tool is cheaper and nearly as good.
What is Intercom Fin?
Fin is Intercom's AI customer service agent that answers support questions from your help content, follows procedures, and hands off to humans when needed. It is widely regarded as one of the strongest AI support agents on the market, with resolution rates that lead most published benchmarks. The headline is its outcome-based pricing: you pay per resolved conversation rather than per seat or per message, and Fin now works on top of other helpdesks like Zendesk as well as Intercom itself.
Best for
Support teams with solid help documentation that want to deflect a large share of tier-1 tickets and pay only for results.
Not for
Anyone with fewer than about 200 resolvable tickets a month. At that volume a flat-rate chatbot costs a tenth as much and the quality gap will not repay the difference.
Strengths
- Bills on outcomes, so a conversation Fin cannot handle and passes to a human costs nothing at all
- One charge per conversation regardless of how many questions or actions are involved, which removes the usual chat-length anxiety
- Genuinely standalone: Fin runs on Salesforce, HubSpot, Freshworks, Zendesk-style helpdesks and custom channels, so no Intercom subscription is required
- Intercom's pricing page states there are no integration, setup or platform fees on top of the outcome price
- The 50-outcome monthly minimum is about $50, which is a genuinely low floor to pilot against
- Procedures give Fin scripted multi-step behaviour rather than pure retrieval, so it can complete tasks not just answer questions
- Outcome definitions are published in plain language, which is more than any other vendor in this batch manages
Limitations
- At $0.99 an outcome it is the most expensive per-resolution price in this batch, roughly double HubSpot's $0.50
- A qualification is $9.99, ten times a resolution, and it is easy to configure Fin into charging that without intending to
- A procedure handoff is billable even though a human then does the work, so a scripted escalation is not free the way an unscripted one is
- There is no published volume discount tier, so 5,000 outcomes cost five times what 1,000 cost
- Fin Voice pricing is not published at all as of August 2026, so voice cannot be budgeted from public information
- Costs scale linearly with ticket volume forever, which is the opposite of what a growing business wants from software
- The Copilot add-on reintroduces per-seat pricing after the outcome model has trained you out of thinking that way
Intercom Fin pricing
The billing unit is an outcome, not a conversation and not a message. Intercom charges $0.99 each for a resolution, a procedure handoff or a disqualification, and $9.99 for a qualification, with a stated minimum of 50 outcomes a month. Crucially you are charged once per conversation no matter how many questions Fin answers or actions it takes inside it, and Fin's own pricing page states you are not charged when a conversation is simply passed to your team without an outcome. Seats are separate and only needed for the humans: Intercom's Essential plan is $29 per seat per month, Advanced $85 and Expert $132, with the Copilot assistant a further $35 per user per month. Intercom bills in US dollars only.
| Plan | Price | What you get |
|---|---|---|
| Fin AI Agent | $0.99/outcome | Resolution, procedure handoff or disqualification. Charged once per conversation however many actions Fin takes. 50 outcomes a month minimum |
| Qualification outcome | $9.99 each | A qualified lead is billed ten times a resolution, which is the line item that surprises marketing teams |
| Essential | $29/seat/mo | Cheapest human seat tier; annual pricing quoted with Fin outcomes billed separately |
| Advanced | $85/seat/mo | Mid tier; the seat price is independent of Fin usage |
| Expert | $132/seat/mo | Top published seat tier |
| Fin Copilot | $35/user/mo | The agent-facing assistant, priced per human rather than per outcome |
| Fin Voice | Not published | Custom pricing, described as available to selected customers working with the sales team as of August 2026 |
Pricing reflects public plans as of August 6, 2026 and can change. Check Intercom Fin for the latest.
What people actually build with it
Fin is a customer-facing support agent and almost everything worth building with it is a variation on deflect, act, or hand over cleanly. The useful distinction is between the deployments that pay for themselves at $0.99 an outcome and the ones that do not, and that turns entirely on how long the equivalent human reply would have taken.
The four builds that reliably earn their keep
Order status and delivery chasing for an ecommerce store. A customer asks where their order is on the site widget, Fin authenticates them against the email on file, calls the order management system through an action, and returns the tracking state and revised date. One outcome, $0.99, against a human reply that would have taken three to five minutes of console-hopping. This is the single highest-volume Fin deployment we see described and the arithmetic is comfortable.
Policy and entitlement questions for a subscription business. Refund windows, plan differences, cancellation terms, what is included at each tier. These are pure retrieval-augmented generation against your help centre, they arrive constantly, and the answers change rarely enough that the knowledge base stays accurate. The failure mode is content rot rather than reasoning.
Password, access and account recovery flows. Fin runs a Procedure: verify identity, trigger the reset, confirm receipt. This is where Procedures earn their name, because the agent is completing a task rather than answering a question. Watch the billing here though, because a Procedure that ends in a handoff is a billable outcome.
Pre-sales qualification on a marketing site, with the large caveat that Intercom prices a qualification at $9.99 rather than $0.99. Used deliberately on a high-value page this is fine. Left switched on across a busy site it is the fastest way to generate a bill you did not plan for.
Where Fin is genuinely the wrong tool
Internal employee helpdesks. Fin is built for customers, and an internal IT or HR agent grounded in your own systems is better served by Copilot Studio, where employee-facing usage by a Microsoft 365 Copilot licensed user is zero-rated entirely rather than billed at a dollar a go.
Anything that is really a workflow rather than a conversation. If the trigger is a webhook and the outcome is a record updated in three systems with no human on the other end, you want n8n or Make.com, not a per-outcome support agent. Fin bills for outcomes in conversations; it is not a general automation platform.
Low-volume support. Below roughly 200 resolvable tickets a month the 50-outcome minimum is not the problem, the unit price is: Chatbase and Botpress will handle a similar share of repeat questions on a flat monthly fee, and the quality difference will not repay the gap.
Bottom line: Fin pays for itself wherever a resolution displaces more than about ninety seconds of human handling time, which covers order status, entitlement questions and account recovery, and does not cover short, cheap-to-answer tickets.
What Intercom Fin actually costs at small-business volume
Take the reference workload used across this batch: 1,000 support conversations a month, of which 700 are resolved without a human. Fin bills 700 outcomes at $0.99, which is $693. The 300 that go to a person cost nothing, provided they were passed over without an outcome being recorded. That is the whole agent bill. There is no platform fee, no setup fee and no seat requirement to run Fin itself.
You still need somewhere for the 300 escalations to land. If that is Intercom, three Essential seats at $29 each add $87 and the all-in total is about $780 a month. If you already run a helpdesk Fin supports, the total stays at $693, because Fin is genuinely standalone. That distinction is worth more than it sounds: it is the only agent in this batch other than Zapier Agents that you can buy without first buying something else.
The comparison below is the one no vendor publishes, and it is why the prerequisite licence matters more than the headline unit price. All figures are US dollars a month at the same 1,000 conversations and 700 resolutions, using each vendor's cheapest realistic qualifying licence and three human seats where seats are needed.
| Platform | Agent price | Licence you must already hold | All-in monthly total |
|---|---|---|---|
| Intercom Fin | $0.99 per outcome | None - Fin sells standalone | $693 standalone, ~$780 with 3 Intercom Essential seats |
| HubSpot Breeze | $0.50 per resolution | HubSpot Professional or Enterprise | ~$590 with 3 Service Hub Pro seats, plus $1,500 one-time onboarding |
| Agentforce (Conversations) | $2.00 per conversation | Salesforce Service Cloud licences | ~$2,525 with 3 Service Cloud Enterprise seats |
| Agentforce (Flex Credits) | $0.10 per action, 20 credits each | Salesforce Service Cloud licences | ~$825 at 3 actions a conversation, same 3 seats |
| Copilot Studio | ~$0.064 per conversation at 8 credits | Microsoft 365 tenant plus a credit pack | $200, the price of one 25,000-credit pack |
| Zapier Agents | 1,500 activities for $33.33/mo | None, but the cap does not reach this volume | Not viable at 1,000 conversations; Enterprise pricing unpublished |
The number that changes the answer
Fin looks expensive in that table and it is, on unit price. What it does not carry is the prerequisite. Read the HubSpot row again: the credit cost is $320, lower than Fin, but you cannot buy it without a Professional subscription and a $1,500 onboarding fee, so the first-year monthly average is nearer $715. Read the Agentforce Conversations row and note that it bills all 1,000 conversations, not the 700 that resolved, which is what makes it three times Fin.
Copilot Studio is the cheapest by a wide margin at this volume and that is not a mistake in the arithmetic. It is a consequence of Microsoft charging for compute rather than outcomes. Whether that is a bargain depends entirely on whether your business already lives inside a Microsoft tenant, which is the subject of that platform's own page.
Bottom line: At 700 resolutions a month Fin costs $693 with no prerequisite licence, which beats Agentforce outright and loses to HubSpot Breeze only if you already pay for HubSpot Professional.
The gotcha: a procedure handoff is billed, an ordinary escalation is not
Fin's headline promise is that you only pay when it works, and that promise is mostly kept. Intercom's pricing page states you are not charged when a conversation is simply passed to your team without an outcome. But the list of billable outcomes contains an item that looks like a failure and bills like a success: a procedure handoff.
The distinction is subtle enough that most buyers miss it. If Fin cannot help and gives up, that is free. If Fin runs a Procedure you authored, and that Procedure's final step is to route the conversation to a human, that is an outcome and costs $0.99, even though a person now does all the actual work. The more carefully you script your escalations, the more of them you pay for. Nothing about the model is dishonest, but the incentive it creates for you is backwards: sloppy handoffs are free and well-designed ones are not.
The second gotcha lives in the same list. A qualification is $9.99, ten times a resolution. If you enable Fin on marketing surfaces and let it qualify, a hundred qualifications is a thousand dollars, which will dwarf your entire support bill. Neither figure is hidden; both sit on the pricing page. The problem is that they are presented as a flat list of outcome types with a single price attached to three of the four, and the eye reads $0.99 and stops.
The practical control is an audit before launch rather than a cap after it. Go through every Procedure and ask which ones end in a handoff. Decide whether that handoff is worth a dollar to you, because sometimes it genuinely is - a Procedure that gathers the right details before escalating saves the human real time. Then scope qualification to specific pages and specific intents rather than leaving it on everywhere.
Bottom line: An unscripted escalation costs nothing but a Procedure that ends in a handoff bills at $0.99, so the better-designed your escalation path, the more of it you pay for.
What breaks in production, and how you find out
Deflection rate is the metric everyone watches and the one that hides the failures. An outcome is recorded when the customer stops asking, and a customer stops asking for two very different reasons: they got the answer, or they gave up. Both bill $0.99 and both look identical on the dashboard.
Confidently wrong answers on stale content
Fin grounds its answers in your help centre, so a help centre that has drifted from reality produces a fluent, well-formatted, wrong answer. This is not a hallucination in the usual sense; the retrieval worked, the source was simply out of date. It bills as a resolution, the customer acts on it, and you hear about it a week later through a complaint rather than through the agent.
The detection method is boring and works: sample twenty billed outcomes a week and read them. For the first month, read them all if the volume allows. Automated quality scoring will not catch a plausible answer sourced from a stale article, because nothing about the interaction looks wrong.
Actions failing quietly against your own systems
Fin's most valuable deployments call your systems - order management, billing, account services - through actions. When an action fails, the agent's instinct is to be helpful anyway, which means answering around the missing data. A customer asking where their order is may be told the general shipping policy rather than told that the lookup failed. That is a resolution by Fin's definition.
Instrument on your side, not Fin's. Log every action call and its outcome in your own systems, alert on error rates rather than on absolute failures, and make sure the agent is instructed to escalate rather than improvise when a lookup returns nothing. This is basic observability discipline and it is the thing most Fin deployments skip.
Billing surprises from configuration drift
The third failure is commercial. Someone adds a Procedure, someone else enables qualification on a new page, and the outcome mix shifts without anyone changing the deflection rate. Because Fin bills per outcome and outcome types have wildly different prices, the bill can double while the dashboard looks flat. Review the outcome-type breakdown monthly, not the total, and treat any change in mix as a change that needs explaining.
Bottom line: Deflection rate cannot distinguish a customer who was helped from a customer who gave up, so the only reliable quality control on a Fin deployment is reading a weekly sample of billed outcomes.
What it connects to that you already run
Fin's portability is its most underrated feature. Intercom's pricing page lists Salesforce, HubSpot, Freshworks, Dixa, Front, Zoho, Sprinklr and Gorgias as supported helpdesks, alongside custom channels, and states there are no integration fees. You can put Fin in front of the support stack you already run without migrating anything, which is not true of any other agent in this batch.
For data, Fin works from your help centre content plus whatever your actions expose. Actions are the important half and the half people underinvest in: an agent that can only read articles answers questions, while an agent that can call your order system resolves problems. If your systems already speak REST API, wiring an action is straightforward. If they do not, you are building an integration layer first, and that work is usually larger than configuring Fin.
Where Fin is deliberately narrow is outside support. It is not an automation platform and does not pretend to be, so scheduled jobs, multi-system record syncing and back-office workflows belong somewhere else. Teams commonly pair Fin for the conversational surface with n8n, Make.com or Zapier for everything that runs without a customer on the other end, and that split is a sensible architecture rather than a compromise.
| Layer | What Fin gives you | What you build |
|---|---|---|
| Helpdesk | Native support for Salesforce, HubSpot, Freshworks, Front, Zoho, Gorgias and custom channels | Routing rules for the conversations Fin passes over |
| Knowledge | Retrieval over your existing help centre articles | Keeping the articles true, which is the whole quality story |
| Systems of record | Actions that call your APIs during a conversation | The API surface itself, plus error handling when a lookup fails |
| Back office | Nothing - out of scope by design | A separate automation platform for anything without a customer in the loop |
Bottom line: Fin runs in front of the helpdesk you already own rather than replacing it, which makes its switching cost the lowest in this batch on the way in as well as on the way out.
Intercom Fin alternatives: what to pick instead, and when
Fin's competition splits into three groups: the CRM-native agents that are cheaper per unit but need a licence you may not hold, the flat-rate chatbots that are dramatically cheaper and meaningfully less capable, and the general automation platforms that do a different job entirely. The right answer depends far more on what you already pay for than on any feature comparison.
| Platform | Billing unit | Cost at the reference workload | Who it suits | Does better than Fin |
|---|---|---|---|---|
| Intercom Fin | Outcome ($0.99) | $693, no prerequisite | Businesses with expensive support labour and any helpdesk | Charges nothing for a conversation it fails to resolve |
| HubSpot Breeze | Resolution ($0.50) | ~$590 all-in, HubSpot Pro required | Companies already on HubSpot Professional | Half the unit price, and grounded in CRM records not just articles |
| Agentforce | Conversation ($2) or action (20 Flex Credits) | ~$825 to ~$2,525 depending on model | Salesforce shops with real CRM depth | Acts on the CRM data model with your existing permissions |
| Chatbase | Flat monthly plan | Tens of dollars, not hundreds | Low-volume support, simple FAQ deflection | Costs a fraction as much for the easy 60 percent of questions |
| Botpress | Flat plan plus usage | Materially below Fin at this volume | Teams who want to own the conversation design | Far more control over flow logic and channels |
When to pick the alternative
Pick HubSpot Breeze if you already pay for HubSpot Professional. You are past the prerequisite, the unit price is half, and the agent can ground answers in the customer record rather than only in help articles. If you are not on HubSpot, do not buy HubSpot to save $0.49 a resolution; the licence and the $1,500 onboarding fee erase the saving for years.
Pick Agentforce only if Salesforce is genuinely your system of record and you will use Flex Credits rather than Conversations. Conversation pricing at $2 bills every chat including the failures, which is the exact opposite of Fin's model and roughly triple the cost at this workload.
Pick Chatbase or Botpress if your support volume is modest or your questions are simple. Below a couple of hundred resolvable tickets a month, a flat-rate tool is the rational choice and the honest recommendation. Fin's advantage is real but it is an advantage on hard tickets, and a business without many hard tickets should not pay for it.
Pick Copilot Studio if the agent is employee-facing rather than customer-facing. Employee usage by a Microsoft 365 Copilot licensed user is zero-rated, which makes an internal helpdesk agent close to free in a Microsoft estate and makes paying Fin a dollar per internal question hard to justify.
Bottom line: Fin wins when you have expensive support labour and no incumbent suite; HubSpot Breeze wins if you already hold HubSpot Professional, and a flat-rate tool wins below roughly 200 resolutions a month.
Migration and lock-in: can you get your work out?
Fin has the lowest vendor lock-in of anything in this batch, and it is structural rather than generous. Because Fin sits in front of a helpdesk you already own, the two expensive assets in a support operation - your ticket history and your help centre content - never move into Intercom's exclusive custody in the first place. Turn Fin off and your helpdesk carries on.
What you would lose is the configuration: Procedures, action definitions, tone and behaviour instructions, and whatever tuning you did against real conversations. None of that has an export format another vendor can read, and none of it is portable in any meaningful sense. In practice that is a few weeks of work to reconstruct rather than a business-critical dependency, which is a very different position from Agentforce or Copilot Studio, where the agent is fused to the platform's own data model.
The dependency that is genuinely worth planning around is commercial rather than technical. Outcome pricing scales linearly with your ticket volume, which means Fin gets more expensive precisely as your business grows, and there is no published volume discount to bend that curve. A company doing 5,000 resolutions a month pays about $4,950, which is well past the point where a purpose-built agent you own outright becomes cheaper. That is not lock-in, but it is a reason to keep the exit warm.
The practical hedge costs almost nothing. Keep your help centre content in a system you control rather than authoring it inside a vendor's editor. Document every Procedure as plain-language business logic alongside the implementation, so a rebuild starts from intent rather than from reverse engineering. And keep a set of fifty real conversations with the right answers written down, so that if you ever evaluate a replacement you can measure it rather than trust a demo.
Bottom line: Nothing valuable is trapped inside Fin because your tickets and help content stay in systems you already own, so the exit cost is a few weeks of reconfiguration rather than a migration project.
Intercom Fin FAQ
How much does Intercom Fin cost per resolution?
$0.99 per outcome as of August 2026, where an outcome is a resolution, a procedure handoff or a disqualification. A qualification is billed separately at $9.99. You are charged once per conversation however many questions Fin answers inside it, and there is a 50-outcome monthly minimum, which puts the effective floor at about $50 a month. Seats for your human team are priced separately, from $29 each on the Essential plan.
If Fin fails and a human takes over, do we still pay?
Not if no outcome was reached. Intercom's Fin pricing page states plainly that you are not charged when a conversation is simply passed to your team without an outcome, which is the single best thing about the model. The exception matters though: a procedure handoff is itself a billable outcome. So an unscripted escalation is free and a scripted one costs $0.99 even though a person still does the work. Audit your procedures before launch to see which handoffs you have accidentally made chargeable.
Does Fin require an Intercom subscription?
No, and this is unusual in this category. Intercom's own pricing page says Fin works with any helpdesk including Salesforce, HubSpot, Freshworks, Front, Zoho, Gorgias and custom channels, and can be bought with no seat costs or platform fees against a 50-outcome monthly minimum. Compare that with HubSpot Breeze, which needs a Professional subscription first, or Agentforce, which needs Salesforce licences. If you already have a helpdesk you like, Fin is the only agent in this batch you can bolt on without changing it.
Does Intercom Fin have a free plan?
No. There is a 14-day free trial of Intercom with no credit card required, but Fin itself is billed per outcome with a 50-outcome monthly minimum, so the cheapest real month is roughly $50. That is still the lowest entry point of anything in this batch other than Zapier Agents, and far below Copilot Studio's $200 prepaid pack or HubSpot's Professional tier. Treat the first month as a measurement exercise rather than a deployment.
Intercom Fin vs HubSpot Breeze: which is cheaper?
HubSpot on unit price, Intercom on total cost unless you already run HubSpot. Breeze charges $0.50 per resolution against Fin's $0.99, but Breeze is only available on HubSpot Professional or Enterprise, which starts at $90 per seat per month plus a $1,500 one-time onboarding fee. Fin needs no prerequisite licence at all. At 700 resolutions a month Fin is about $693 standalone; HubSpot is about $320 in credits plus roughly $270 of seats plus onboarding. If HubSpot is already your CRM, Breeze wins. If it is not, Fin wins easily.
Is Intercom Fin worth it?
It depends on what a resolution is worth to you, and the sum is quick. Multiply the minutes a human takes on a typical ticket by your fully loaded hourly cost. If that number is meaningfully above a dollar, Fin's outcome pricing is good value and gets better as your tickets get longer. If your tickets are short and cheap to answer, a flat-rate tool like Chatbase or Botpress will deflect most of the same questions for a fraction of the money. Fin is not overpriced; it is priced for businesses whose support labour is genuinely expensive.
What is a qualification and why is it $9.99?
A qualification is Fin establishing that a visitor is a viable prospect, which Intercom prices as a sales outcome rather than a support one at $9.99 each. The logic is defensible because a qualified lead is worth far more than an answered question. The risk is operational: if you point Fin at a high-traffic marketing page and let it qualify freely, you can generate a four-figure line item from a workflow you thought of as chat. Cap it, scope it to specific pages, and watch the first month's usage closely.
What breaks with Fin in production?
Three things, in the order they bite. First, outcome definitions drifting from your intent, usually via procedures that turn free escalations into billed handoffs. Second, content rot: Fin is only as good as the knowledge base behind it, and a stale article produces a confident wrong answer that still bills as a resolution. Third, the resolution-then-reopen pattern, where a customer accepts an answer, thinks about it, and comes back. Review a sample of billed outcomes weekly for the first month rather than trusting the deflection percentage.
Can you set this up and tune it for us?
Yes, and the tuning is where the money is with an outcome-priced agent. We audit which of your ticket types are genuinely resolvable, clean the knowledge base those answers depend on, configure procedures so that handoffs you do not want billed are not billed, and set a monitored sample review so a confidently wrong answer does not run for a month. That starts with a free automation audit, and the handover includes documentation you own plus a recurring plain-language report of what ran, what it resolved and what it cost.
Looking at alternatives? Fin has the cleanest incentive alignment of anything in this directory and we will say so plainly: if the agent does not resolve the ticket, you do not pay. That is worth a premium. What we would push back on is the assumption that a resolution is worth a dollar in every business. For a company where a support reply takes four minutes of a $20-an-hour person, a resolution is worth about $1.30, so Fin's margin for you is thin. For a company where tickets take twenty minutes, it is excellent value. Do that sum before you commit. Where Autonoly differs is that we build a fixed-scope automation you own outright rather than renting per outcome forever, which wins on cost past a few thousand resolutions and loses below a few hundred. See the Autonoly review.
