A practical guide to Intercom's Fin audience segments

Let's be honest, generic AI support can be pretty frustrating. Customers expect conversations that feel personal, not like they're talking to a broken record. To tackle this, Intercom's Fin AI Agent has a feature called audience segmentation, which is designed to help you deliver more targeted support.

An illustration of the Intercom Fin AI Agent interface, representing a starting point for personalized AI support.

In this guide, we’ll walk through what Fin audience segments actually are, how to get them running, and, most importantly, the big limitations you need to know about. We'll also look at a more flexible approach to AI automation that gives you the control you’re really after.

What are Intercom Fin audience segments?

In a nutshell, Fin audience segments let you control which users or leads get to chat with your Fin AI Agent. It all works based on rules and attributes you set up inside Intercom. Think of it as a bouncer for your AI. Instead of letting Fin talk to every single person who contacts you, you can create specific groups for it to engage with.

The main idea here is to help teams roll out their AI agent carefully. You can test it on a small, low-risk group first, create different experiences for customers (like free users vs. VIPs), and slowly let it handle more conversations as you get more comfortable with its performance.

The rules you create can be based on all sorts of data points, including:

• User details like their email, what plan they're on, or where they're located.

• Company information if you're in the B2B space.

• What a user is doing on your site, like the specific page they're looking at.

You can set these segments up in Intercom's Workflows or right from the Fin setup pages for chat and email. It’s a good first step toward personalizing AI support, but as you’ll see, it has some serious strings attached.

How to set up and use Fin audience segments

Getting started with Fin audience segments is a mix of defining rules and planning a slow and steady rollout. While it's all built into the Intercom platform, it’s a manual process that requires a lot of thought, especially when you factor in testing and how much it’s going to cost you.

Setting up audience rules

You can find the audience settings in a couple of different spots in your Intercom dashboard. The most common places are under Fin AI Agent > Deploy > Chat (or Email) in the "Who will see Fin" section, or inside the trigger settings for a specific Workflow.

This is where you’ll add filters to create your audience. For instance, you could make a rule to only show Fin to users where the attribute ‘Plan is Free’. This is a pretty common way to dip your toes in the water with a specific group before going all in.

A gradual rollout strategy: Testing with your internal team

Intercom suggests that the first thing you should do is set Fin live for your own team. You can do this by setting up an audience segment using an email filter, like ‘Email contains @yourcompany.com’. This lets your team play around with the AI in a live setting and catch any glaring issues.

But here’s the major catch: Intercom is very clear that "Fin will be billed per resolution," even when you're just testing internally. That means you start paying the second you turn it on, turning every test conversation into a real cost. It’s a pretty expensive way to find out where your AI is falling short.

This is a huge contrast to how modern AI platforms handle testing. For example, eesel AI gives you a simulation mode that runs your AI setup on thousands of your past support tickets before it ever talks to a real person. You get a clear forecast of its performance and potential savings, all without spending a dime or risking a bad customer interaction.

Rolling out to customers

Once you feel ready to move past your internal team, you can start building more specific customer segments. Here are a few practical examples of how you might use Fin's rules:

• By tier: You could target users on a specific plan by using a custom attribute like ‘Plan is Free’, or keep the AI away from certain groups with a rule like ‘VIP customer is false’.

• By region or recency: You could combine filters like ‘Last seen’ and ‘Region’ to create a segment of new users in a particular country.

• By user type: You can also separate logged-in 'Users' from anonymous 'Leads,' and offer different AI experiences to each.

• By website behavior: Using the ‘Current Page URL’ attribute, you can turn on Fin only on pages that get a lot of traffic, like your pricing or support pages where you know people ask the same questions over and over.

While these rules give you some control, they all depend on one critical thing: having accurate and up-to-date custom data in Intercom. If your data is messy, incomplete, or just not there, your ability to create useful segments is pretty much gone.

Key limitations of Fin audience segments (and a better alternative)

While rule-based segments are a decent starting point, they have some big constraints that can stop you from building a truly smart and scalable automation system. Let’s break down the main issues and look at a more effective way to approach this.

The challenge of rigid, rule-based logic

Fin's targeting is limited to the user attributes you’ve already defined. It decides whether to jump into a conversation based on who the customer is (their plan, their location), not what their problem actually is. If a loyal customer has a really complex or sensitive issue, Fin might still try to handle it just because their profile fits a rule, leading to a clumsy handoff and a frustrated customer.

This is where a fully customizable workflow engine, like the one in eesel AI, really shines. Instead of just filtering users, you can build automation logic around the content and intent of the message itself. This allows for truly selective automation. You can set it up so the AI only handles simple, common questions and confidently passes everything else to a human agent. It gives you a level of control that just isn’t possible with basic user attributes.

Knowledge is siloed within Intercom

Fin works best with the knowledge you have stored in your Intercom help center. But if your company is like most, you have important information scattered across all sorts of places. If that info isn't in Intercom, your AI can't use it. This creates knowledge gaps and means your AI can only answer a small fraction of customer questions.

eesel AI was built to fix this exact problem by pulling all of your knowledge sources together in one click. It connects to your helpdesk, but also to tools like Confluence, Google Docs, and Notion. Even better, it can train on your past ticket history to automatically learn your business context, common solutions, and even your brand voice from day one.

Risky rollouts and unpredictable costs

The biggest headache with Intercom's "gradual live rollout" is the financial risk. You're paying for every single resolution (and every mistake) your AI makes right from the start. This makes budgeting a total guessing game and forces you to fix problems after they’ve already affected your customers and your bottom line.

This is where eesel AI offers a fundamentally better approach, one built on confidence and predictability.

• Risk-Free Simulation: Before your AI ever speaks to a customer, eesel AI lets you test it on thousands of your past tickets in a safe sandbox. You get a precise, data-backed forecast of its resolution rate and how much you could save. You can see exactly how it will respond to real questions, tweak its behavior, and launch only when you’re 100% confident.

• Predictable Pricing: Unlike Intercom's pricey per-resolution model, eesel AI has transparent plans based on interaction volume. You know exactly what you're paying each month, so you’ll never get a surprise bill after a busy week.

A comparison table: Fin audience segments vs. eesel AI

Feature	Intercom Fin Audience Segments	eesel AI
Automation Logic	Filters users by who they are	Understands what the problem is
Knowledge Sources	Limited to Intercom and connected apps	Unifies all sources (Helpdesk, Docs, Confluence, past tickets)
Testing & Rollout	Live testing with a "pay-as-you-go" risk	Free, risk-free simulation on historical tickets
Setup	Manual configuration within Intercom Workflows	Radically self-serve, go live in minutes
Pricing Model	Per-resolution (unpredictable costs)	Flat, predictable pricing based on interactions

Move beyond basic Fin audience segments to intelligent automation

So, what's the verdict? Intercom Fin audience segments give you a basic lever to control who interacts with your AI. For teams just dipping their toes into AI support, it can be a useful tool for simple targeting and cautious rollouts.

However, its dependence on rigid rules, siloed knowledge, and a risky "pay-to-test" model holds teams back from achieving truly smart and scalable automation. To build a top-tier support operation, you need more than a simple gatekeeper. You need an intelligent system that understands context, learns from all your scattered data, and lets you work with complete confidence.

For teams that want total control, risk-free testing, and an AI that learns from all of their company knowledge, a more advanced solution is the only way forward. eesel AI is a self-serve platform that plugs into your existing tools, gives you a powerful simulation engine to build confidence, and offers predictable pricing that grows with your success, not your problems.

Ready to see how you can automate support with real confidence? Sign up for eesel AI for free or book a demo to learn more.