A guide to using Intercom workflows to auto add labels based on keywords and intent

If your support inbox feels like a never-ending game of whack-a-mole, you're not alone. It’s overflowing with questions, and your team is spending way too much time just trying to sort through the chaos. Manually tagging, routing, and categorizing conversations is a massive time sink that stops your agents from tackling the complex problems where they can really shine.

Automation seems like the obvious fix. Automatically labeling tickets can seriously improve routing, reporting, and your team's overall vibe. If you're using Intercom, you’ve probably checked out their built-in tool, Workflows, to handle this. It’s a decent starting point, but as many teams find out, it comes with its own set of headaches.

This post will walk you through how Intercom's labeling system works and show you the limits of its keyword-based automation. More importantly, we'll look at a more powerful, AI-driven approach that actually understands what your customers mean, not just the words they type.

Understanding Intercom's conversation labeling system

Before we get into building workflows, it helps to know what tools you have in your Intercom toolkit. The platform gives you three main ways to categorize conversations, each with a different job. Getting these straight is the first step to a saner inbox.

• Conversation Topics: This is Intercom’s most hands-off option. You define a topic by creating a list of keywords and phrases. When a customer’s message hits one of those keywords, the conversation gets assigned that topic automatically. It’s good for a bird's-eye view of what people are talking about.

• Conversation Attributes: Think of these as custom, structured data fields you create, like dropdown menus for "Issue Type," "Priority," or "Product Area." Attributes add a layer of order because they have a set list of values, which makes them great for routing conversations to the right team or kicking off specific automations.

• Conversation Tags: Tags are the most flexible of the bunch. They're basically free-form labels you can slap onto individual replies in a conversation. This makes them perfect for one-off tracking, like flagging feedback for the product team or marking a conversation that’s part of a known bug.

Here’s a quick cheat sheet to help you decide what to use:

Feature	Best For	How It's Applied	Flexibility
Topics	High-level trend reporting.	Automatically via keyword matching.	Low (rigid keywords)
Attributes	Structured workflow management (e.g., routing, SLAs).	Manually or via Workflows.	Medium (predefined values)
Tags	Ad-hoc, specific categorization (e.g., bug reports).	Manually or via Workflows.	High (free-form)

How to set up Intercom workflows to auto add labels based on keywords and intent

Once you've got the hang of topics, attributes, and tags, you can start building automations with Intercom Workflows. This is where you can start to "auto add labels based on keywords and intent", or at least, the keyword part.

The core logic: Triggers and actions

Intercom Workflows run on a simple "if-then" logic that’s pretty intuitive if you've ever played around with automation tools.

1. Trigger: This is what starts the workflow. A classic trigger is "When a new conversation is started by a customer."

2. Condition: This is a rule the conversation has to meet for the workflow to continue. This is where your keywords come into play. For instance, you could set a condition like "If the message body contains the word 'refund'."

3. Action: If the condition is met, the workflow does something. This could be "Add the tag 'Refund Request'" or "Set the 'Issue Type' attribute to 'Billing'."

A practical example

Let's walk through a common scenario: automatically tagging a billing question. Your goal is to catch any message about payments and get it over to your finance team.

You could build a workflow that starts when a new conversation comes in. You’d then add a condition that checks the message for keywords like "invoice," "payment," "billing," or "charge." If it finds one, the workflow takes action, like setting a custom attribute "Issue Type" to "Billing." From there, another rule could automatically route any conversation with that attribute to the finance team’s inbox.

This is a solid first step for teams just dipping their toes into automation. It can definitely help tidy up the inbox and get some conversations to the right place more quickly. But as your ticket volume picks up, you'll start bumping into the walls of relying on simple keywords.

The limitations of keyword-based workflows

Relying on keywords is a bit like trying to understand a book by only reading every tenth word. You get the gist, but you miss all the important context. Here’s where keyword-based workflows in Intercom begin to fall apart.

Inaccuracy: Keywords are not the same as intent

The biggest problem with keyword systems is that they can't figure out what a customer actually wants. A customer asking for a refund might say, "I want my money back," "this charge is wrong," or "how do I return this?" A workflow that's only looking for the word "refund" is going to miss every single one of those.

This creates a domino effect of issues: tickets get missed, routing breaks, and your reports are all off because conversations aren't categorized correctly. Your agents end up spending their time manually re-tagging everything, which kind of defeats the whole point of automation.

The high maintenance of keyword lists

Your business changes. Your products get updated, policies shift, and customers find new ways to talk about you. That means your keyword lists are pretty much always going out of date.

Keeping these workflows running effectively turns into a full-time job for a support manager. They have to constantly dig through conversations, find new keywords, and manually update a mountain of rules. It’s a tedious cycle that just doesn't work as your company gets bigger.

Intercom's AI evolution: Fin attributes

To be fair, Intercom knows that keywords have their limits and introduced a more advanced feature called Fin Attributes. This uses AI to classify conversations based on natural language descriptions you write for each attribute. So instead of just listing keywords, you describe what a "Billing Issue" looks like in a few sentences.

This is a step up, for sure. But it still leaves the heavy lifting to you. You have to manually write out detailed descriptions for every single category and sub-category you want Fin to recognize. It's better than a simple keyword list, but it's not the smart, self-learning system that modern teams really need.

The cost barrier: Advanced features come with a price tag

Powerful automation in Intercom isn't cheap. The main Workflows builder, which you need for any real automation, is only on the Advanced plan and up. If you want to use the AI-powered Fin Attributes, you'll have to pay an extra fee for every conversation the AI resolves.

Here's a quick look at Intercom's pricing. As you can see, the features that make a real difference are locked behind the more expensive plans, and the AI itself has a usage-based cost that can be hard to predict.

Plan	Price (per seat/mo, billed annually)	Key Automation Features
Essential	$29	Basic Help Center, Shared Inbox
Advanced	$85	Workflows automation builder, Multiple team Inboxes
Expert	$132	SLAs, Multibrand Help Center
Fin AI Agent	$0.99 per resolution	AI-powered answers and classification (Fin Attributes)

While Intercom gives you some tools to get started, they can be rigid, a pain to manage, and expensive to scale.

A better approach: True AI

Instead of trying to guess every possible keyword your customers might use, what if your system could just... understand them? That’s what a true AI-powered platform does. It goes beyond keywords to get the context and intent, which leads to much more accurate and reliable automation.

Go beyond keywords with true intent detection

eesel AI was designed from the start to get the nuances of customer conversations. It doesn't just scan for keywords; it analyzes the whole message to figure out what the customer is actually trying to do.

This means it can correctly spot a refund request whether the customer says "refund," "money back," or "dispute this charge." The AI understands what they mean, so your ticket tagging, routing, and reporting are always accurate without you having to lift a finger.

How eesel AI improves on Intercom workflows

eesel AI plugs right into your help desk, like Intercom, and gives it a brain boost. Here’s how it gets around the limitations of the native workflows:

• Go live in minutes, not months: Forget about spending weeks writing rules and descriptions. With eesel AI, you just connect your Intercom account with a single click, and the AI starts learning from your past conversations right away. There's no complicated setup or need to switch tools.

• Train on your real history: eesel AI's AI Agent looks at thousands of your old support tickets to learn your company's specific issues, brand voice, and common solutions. It gets your business context from day one, so it can start tagging conversations with an accuracy that would take months to build manually.

• Test with confidence: Rolling out new automation can be nerve-wracking. eesel AI takes the guesswork out of it with a powerful simulation mode. You can test your setup on thousands of past tickets to see exactly how the AI would have tagged, routed, and responded before you turn it on for live customers. You get a clear picture of its performance and can make tweaks without any risk.

This approach is a world away from the rigid, manual systems of the past.

Feature	Intercom Workflows	eesel AI
Accuracy	Relies on exact keyword matches.	Understands customer intent and context.
Setup	Manual creation of every rule and keyword list.	Goes live in minutes. Learns automatically from past tickets.
Maintenance	Requires constant updates to keyword lists.	Self-improving based on new conversations.
Testing	Limited to testing individual workflows.	Powerful simulation on thousands of historical tickets.
Knowledge	Limited to Intercom and manually added info.	Connects to Intercom, Confluence, Google Docs, and more.

Stop managing keywords, start automating with intelligence

Manually labeling support tickets is a surefire way to burn out your team and kill efficiency. While keyword-based rules in tools like Intercom are a step up from nothing, they're a brittle solution that’s hard to scale. They require constant babysitting and often miss what customers are actually saying.

True AI-powered automation is a smarter way forward. By understanding customer intent and learning directly from your business data, it offers a more reliable, efficient, and scalable way to manage your support inbox. eesel AI plugs into the tools you already use to provide this intelligence, letting you finally get off the keyword hamster wheel.

Ready to see how real AI can transform your Intercom support? Simulate eesel AI on your historical tickets for free and get an instant report on your automation potential.