Intercom Fin vs a Custom Support Chatbot: Which to Pay For in 2026

Short answer: if your customers' questions are already answered in a help centre and your support team lives in a helpdesk, buy Fin (or the equivalent on your helpdesk) and stop reading. Intercom prices Fin at $0.99 per resolved conversation on top of plan seats at $29, $85 or $132 per agent per month billed annually (intercom.com/pricing, read 29 September 2026). Build a custom support chatbot only when the bot has to look things up in your own systems or do things for the customer: check an order, change a booking, issue a refund, read an account. That is a $15,000 to $50,000 build for a first production version, $50,000 to $100,000 when it acts across several systems, plus $5,000 to $10,000 per month to keep improving it. Those are KUMO's published bands. The rest of this page is the four questions that put you on one side or the other.

The comparison in one table

Question 1: where do the answers live?

If 80% of the questions your team gets are answered by an article you could write, a platform chatbot is the right product. Fin reads your help centre and answers from it, and its pricing only charges you when it resolves the conversation. Writing the articles is the work.

If the answers live in your systems ("where is my order", "why was I charged twice", "can I move my appointment"), no help article answers them. The bot needs to read your data for this customer, safely, and that is integration work whichever logo is on the widget. This is the line between the two columns, and most buyers are on the platform side of it.

Question 2: does the bot need to do anything?

Answering is cheap. Acting is not. Changing a booking, issuing a refund, updating an address or cancelling a plan means validation, duplicate prevention, approval rules for high-value cases, an audit record and a way to undo. Platform tools do offer actions through their integrations; the question is whether your systems and rules fit the actions they offer. When they do, use them. When your process has exceptions the platform cannot express, that is where custom earns its price, and why it is priced as workflow software rather than a chat feature. AI automation approval workflows covers how approvals should work.

Question 3: what does volume do to the bill?

At $0.99 per resolution, 2,000 resolved conversations a month is about $2,000 plus seats; 10,000 is about $10,000 plus seats. Volume pricing is fair when volume is modest and the alternative is hiring. It becomes the reason to build when resolutions are high and predictable: a custom bot's build cost is fixed and its running cost is model usage, usually tens to low hundreds of dollars a month at Starter scale, plus the engineering retainer. Run both numbers over 24 months before deciding; the crossover is usually somewhere between a few thousand and ten thousand resolutions a month, depending on seats.

Question 4: who will own it in month four?

A platform chatbot is owned by the vendor: you configure, they operate. A custom chatbot is yours, which is the point and the cost. Someone reviews unanswered questions, refreshes knowledge, watches handoff rates and keeps integrations working. If nobody in your company will do that and you do not want a retainer, buy the platform. If you want the bot to keep getting better around your own process, the production AI support cost breakdown shows what ownership involves.

Where each one wins

• Platform wins: standard support, help-centre answers, a team already on the helpdesk, modest volume, no appetite to operate software.
• Custom wins: account-specific answers, actions inside your systems, a workflow that is not a helpdesk (bookings, claims, onboarding, internal operations), high predictable volume, or a requirement that the data never leaves your cloud. Private AI agent deployment covers that last case.
• Both: many companies run a platform bot for the help-centre questions and a custom integration behind it for the account ones. The AI support copilot vs chatbot guide covers the third option, assisting agents rather than replacing them.

What the custom side costs, precisely

A KUMO Starter Build ($15,000 to $50,000) for a support chatbot is a launched first production version: retrieval over your knowledge with citations, one or two systems read or written under stated rules, handoff into your helpdesk with the conversation attached, an evaluation set with pass criteria, deployment in your own cloud, and a handover another engineer can continue from. Two to three months. Code yours from day one. The full breakdown, with what moves the number, is on chatbot cost to build, and the AI development cost calculator gives a range for your scope in about a minute.

KUMO's founders built Volopay's first production version as its founding engineers; how KUMO builds AI products is on AI product and platform engineering.

How to decide this week

1. Pull last month's support conversations and tag each one: answerable from an article, needs account data, needs an action.
2. If the first group is over 80%, trial the platform bot. Intercom says it sets up in under an hour.
3. If the second and third groups are large, write one page listing the systems involved and the actions, run it through the calculator, and get two quotes.
4. Compare both routes over 24 months at your real volume, including seats and the people who will own it.

If you want a second opinion on the tagging, book a free 30-minute call with an engineer and bring the export. We will tell you which column you are in, including when it is the platform.

FAQ

How much does Intercom Fin cost? $0.99 per resolved conversation ("Fin outcome"), on top of Intercom plan seats at $29, $85 or $132 per agent per month billed annually, read from intercom.com/pricing on 29 September 2026. Fin is also sold standalone on other helpdesks at $0.99 per outcome.

How much does a custom support chatbot cost? $15,000 to $50,000 for a first production version that answers from your knowledge, reads one or two systems and hands off to people; $50,000 to $100,000 when it takes actions across several systems; $5,000 to $10,000 per month to keep improving it without in-house engineers.

Is a custom chatbot better than Intercom Fin? Only when the bot must read your systems or act for the customer, or when your workflow is not a standard helpdesk one. For help-centre questions on an existing helpdesk, the platform is better and cheaper.

Can I use both? Yes. A common pattern is the platform bot for help-centre questions with a custom integration behind it for account lookups and actions.

How long does each take? Intercom says Fin sets up in under an hour on an existing helpdesk. A custom Starter Build takes two to three months.

Last verified 29 September 2026. Intercom prices and claims read from intercom.com/pricing on that date; they change, so check the page. KUMO bands are KUMO's published prices.