Quick answer: Fin, the AI agent from the company formerly called Intercom, can work inside the Zendesk you already use: it answers email, form and messaging tickets from your help content and hands the rest to your team in Zendesk, with no migration. As of November 2026 it costs $49 a month including 50 resolutions, then $0.99 per resolution, nothing for standard handovers to a person, and there is a 14-day free trial.
Your support team lives in Zendesk. The tickets, the macros, the reports and the habits are all there, and moving to another help desk just to get AI would cost you months. But the routine questions keep coming, and you'd like something to answer them without sending customers around in circles.
You don't have to switch systems. Here's how I'd add Fin to a Zendesk you already run, what to decide first, and what it really costs.
Yes. According to Fin's own documentation, Fin works inside your existing Zendesk, and Zendesk stays your system of record for tickets, conversations and reporting. Fin can cover three channels, each switched on separately:
Your team keeps working where it works today. If you're comparing this with an AI agent from another vendor, I wrote about how I'd choose between Fin and Agentforce, and about Fin with Salesforce.
The AI is the easy part. What matters is what happens when it shouldn't answer. Before connecting anything, I agree three things with the business:
Then the tools follow: connect Fin to Zendesk, import your help center, choose which tickets Fin may take, and test with real customer questions before going live. Fin's guide describes the same order: train, test, deploy, analyze.
When Fin can't resolve a question, it transfers the ticket or conversation to your team in Zendesk, and a customer who asks for "a person" is routed to an agent even if they don't use an exact phrase. I set up the rules so the ticket arrives assigned to the right group, with the history attached.
One detail worth knowing: per Fin's documentation, when Fin closes a conversation for inactivity, the Zendesk ticket status doesn't change by itself and stays, for example, Pending. I add a simple rule or review step so those tickets don't sit open forever.
As of November 2026, from Fin's billing terms and pricing page:
One exception: if you deliberately configure a procedure that ends in a handover, such as collecting and checking order details before a person approves a refund, a completed procedure is billed at $0.99. You choose whether to set it up that way.
A worked example, with my assumptions stated: your team gets 600 tickets a month, and Fin resolves 200 of them. The bill is $49 plus 150 × $0.99, about $197.50. If each of those 200 tickets would have taken an agent five minutes, that's roughly 17 hours of their month, about $415 at an assumed fully loaded $25 an hour. The alternative many teams actually weigh is overflow help from a part-time agent or an outsourced desk, and AI is usually the cheaper way to cover the routine layer. Your own numbers will differ, so I run this calculation with your ticket data first.
Zendesk's own side of the bill, your plan and agent seats, doesn't change because of Fin. Fin's documentation doesn't state which Zendesk plan is required, so I confirm that with Zendesk before we start.
The traps are almost never technical. They're these:
It answers from the help content you give it and on the topics you allow. I set it to hand over when it can't find a reliable answer, and during the first month we review a sample of conversations every week.
No. Your agents keep working in Zendesk. Fin's conversations and handovers arrive there as tickets.
Start with the 14-day trial and one channel, and judge it on the monthly numbers. Your tickets, customers and help center stay in Zendesk, so if it doesn't earn its keep, you switch Fin off and Zendesk carries on as before.
No. Fin's documentation describes importing your existing knowledge base and connecting to your existing Zendesk channels, without requiring a platform migration.
I design this kind of setup end to end, from deciding what the AI should handle to making sure your team sees everything in one place. See how I brought calls and messages into one CRM for a professional-services firm, or tell me what you want to improve.