n-Fin Answers From Your Help Center by Default. That's Not the Same as Knowing a Customer's Account.
Fin is Intercom's AI agent: it reads your help center content and whatever conversation data it has access to, answers customer questions directly, and hands off to a human agent when it can't resolve something confidently. Out of the box, it's limited to what Intercom itself knows. Connected to real billing, order, or account data through Custom Actions, it answers specific questions instead of generic ones, which is the gap we actually close.
What Fin actually does well by default
Fin is genuinely good at first-response triage: answering common questions from your existing help center articles, understanding intent well enough to route correctly, and recognizing when it's out of its depth and should hand off to a human rather than guessing. For a business whose support volume is dominated by "how do I reset my password" or "what's your refund policy," that alone handles a real chunk of incoming conversations without a human touching them.
Where it struggles is anything specific to the person asking. A question like "where's my order" or "why was I charged twice" needs account-specific data Fin doesn't have unless someone's connected it, and by default, that's exactly the case.
Giving Fin the data it needs to actually answer
Intercom built Custom Actions specifically for this: a webhook Fin can call mid-conversation to fetch or push real data, like checking an order's shipping status or looking up an invoice. Once that's wired up, Fin's answers stop being generic and start reflecting the actual account it's talking to. This is the same underlying work as any backend integration, a webhook endpoint that returns the right data in the right shape, and it's genuinely where most of the value in deploying Fin well actually sits, more than the base chatbot setup itself.
What that connection looks like
Order and billing status, answered directly
A Custom Action tied to your order and billing systems lets Fin answer "where's my order" or "why was I charged" with the real answer, not a deflection to a human.
A handoff that already has the context
When Fin does escalate to a human, the account data it already pulled stays attached to the conversation, so the agent isn't starting from zero either.
What's Fin punting to a human that it could actually answer?
Tell us the specific question types Fin can't resolve on its own right now, and we'll scope the Custom Action that would close that gap.
Running Fin and still seeing it hand off questions it should be able to answer?
Tell us what Fin punts to a human most often, and where that answer actually lives. Usually it's one Custom Action away from being resolved automatically. And if the real manual bottleneck in your business isn't your chatbot at all, we build on whatever it actually is.
Let's talk