Fin AI Copilot vs. Fin AI Agent: What's the Difference and Which One Does Your Team Need

Intercom sells two different products with "Fin" in the name, and mixing them up leads to real budget mistakes. One replaces your human agents on routine conversations. The other makes your human agents faster at the conversations that are left. If you're evaluating Intercom and only look at one of them, you're likely missing the piece that actually solves your problem.

Fin AI Agent is the customer-facing one

Fin AI Agent is what most people mean when they talk about "AI support" in Intercom. It sits inside the same chat window your customers already use, and it tries to resolve a conversation on its own before a human ever sees it. It answers from your Knowledge Hub, which is the shared pool of help center articles, uploaded documents, snippets, and any external sources you've connected. When it's set up to take action rather than just answer, it can also do things like look up an order, issue a refund, or update an account through connected tools, not just describe how to do those things.

Billing for Fin AI Agent is usage-based. Intercom currently lists it as priced per resolved conversation rather than per seat, at a flat rate regardless of which plan tier you're on. That structure matters for planning: a support team with high ticket volume but a thin knowledge base can end up paying for a lot of resolutions that customers immediately have to escalate anyway, because a bad answer still gets billed if the customer doesn't ask a follow-up question.

Copilot is the agent-facing one

Copilot is a separate product built for your human support team, not your customers. It lives inside the inbox your agents already work in, as a side panel they can open while handling a live conversation. Instead of answering the customer, it answers the agent: an agent can type a question into Copilot the same way they'd ask a coworker, and it searches across help center articles, internal-only articles, PDFs, macros, and connected knowledge tools like Notion, Guru, and Confluence to generate a response, complete with linked sources so the agent can verify it before using it.

Beyond direct Q&A, Copilot can summarize a long conversation thread, suggest next steps on a complicated issue, and adjust a drafted reply's tone, length, or language before it goes out. New agents in particular lean on it heavily during onboarding, since it effectively puts your entire internal knowledge base one question away instead of requiring them to know which document to search.

One detail worth knowing if you're on another helpdesk: Copilot isn't locked to Intercom's own inbox. It's also available as an add-on app inside platforms like Zendesk and Salesforce, where an agent installs it, signs in with a Fin account, and gets the same answer-and-summarize functionality without your team having to migrate its whole ticketing system. That makes it a lower-commitment way to test whether Intercom's AI quality is good enough to build around, before deciding whether to move the rest of your support stack over.

Pricing for Copilot works differently from Fin AI Agent. It's charged per agent seat for unlimited use, on top of whatever base helpdesk plan you're already paying for, rather than scaling with conversation volume. That makes the cost predictable but means it only pays for itself if your agents are actually opening it regularly, not sitting unused as a line item.

Why the confusion causes real problems

The mistake I see most often is a business buying one when they needed the other, or assuming one includes the other's functionality. A team drowning in repetitive tier-one tickets doesn't need every agent to have a faster way to look things up. They need fewer of those tickets reaching a human at all, which is a knowledge base and Fin AI Agent configuration problem. Conversely, a team whose volume is already reasonable but whose agents are slow, inconsistent, or spend too much time hunting through old macros and docs isn't going to fix that by tuning an autonomous agent that customers barely interact with; that's a Copilot fit.

Plenty of mature setups eventually run both, and the two are designed to hand off to each other cleanly. Fin AI Agent takes the first pass on incoming conversations and resolves what it can. Whatever it escalates lands with a human agent who has Copilot open in the sidebar, drawing on the exact same underlying knowledge sources to answer the harder question quickly and consistently. Run that way, the two products aren't competing for the same job, they're covering the two different jobs support teams have always had: keep the easy stuff off a person's plate, and make the person faster on what's left.

How to decide which to start with

Look at where the pain actually is before choosing. If your resolution rate is low and agents are buried under conversations that a decent knowledge base should be able to handle without them, start with Fin AI Agent and put the real work into your Knowledge Hub content, since a weak knowledge base will undercut either product equally. If your ticket volume is manageable but response quality and speed vary a lot by agent, or onboarding new hires takes too long, Copilot addresses that more directly and doesn't require handing anything customer-facing over to an AI agent first.

I help businesses configure and tune both sides of Intercom's Fin AI setup, from getting an agent's Knowledge Hub actually resolution-ready to rolling out Copilot in a way agents will use. If you're trying to figure out which one fits your team, feel free to get in touch.

Cover photo by Petr Macháček on Unsplash.