How to Intercom automate follow ups when Fin cannot answer

Let's be real, Intercom's Fin AI can be a huge help, knocking out a ton of common questions. But what about when it just... doesn't get it? We've all been there: a chatbot gives a nonsensical answer, and the customer is left frustrated. That's not just a bad experience; it's a potential goodbye. Trying to catch all these little failures by hand is a recipe for disaster as you grow.

This guide is all about building a safety net. I'm going to show you, step by step, how to use Intercom's own tools to automatically catch these moments. You'll set up a system that triggers a follow-up and gets a real person involved whenever Fin stumbles, so no customer question is left behind.

What you'll need

Alright, before we get our hands dirty, let's make sure you have a few things lined up. This setup uses some of Intercom's more powerful automation features, so you'll need the right access.

• An Intercom plan that includes Workflows (this is usually the Advanced or Expert plan).

• Admin rights (or at least enough permission) in your Intercom account to create and mess with Workflows.

• About half an hour of your time to get the first version of this set up and tested.

How to set up the workflow

We're going to build this right inside Intercom Workflows. The trick is to ask for feedback as part of the workflow, not with your general CSAT settings. That way, you can react the instant someone gives a thumbs-down.

Step 1: Create a new workflow for Fin

First up, head over to the Workflows section in Intercom. You'll want to create a new workflow, and a good trigger for it is "When customer sends their first message."

After you've got the workflow started, add your first step and choose Let Fin answer. This is what puts Fin on the front line to take the first shot at the customer's question.

A view of the Intercom visual workflow builder, which is used to Intercom automate follow ups when Fin cannot answer.

Step 2: Add a conversation rating step

Okay, this next part is the key to the whole thing. Forget about Fin's built-in satisfaction toggle or your company-wide CSAT settings. We need a specific feedback step here.

Right after the "Let Fin answer" step, add a new action: Ask for conversation rating. Now, here's the critical bit: make sure you turn on the setting that says "Wait for customers to give a rating before continuing the Workflow." This is what freezes the automation and waits for the customer's feedback, letting you respond to it immediately.

Step 3: Branch the workflow based on the CSAT score

So now you're asking for a rating. The next move is to split the path based on what the customer says. Add a Branch step right after the rating request.

You'll set up one branch with a simple rule: "Conversation rating" "is" "Negative" (or whatever sad emoji you have set up). This creates a special route just for conversations where Fin didn't help. You can also make a "Positive" branch to automatically close the chat or send a quick "Happy to help!" note.

Step 4: Define the automated follow-up actions

Inside that "Negative" branch, it's time to decide what happens next. This is where your automated follow-up really comes to life. Here are a few of the most useful actions to add:

• Assign to a team. Get it in front of a human, fast. Send the conversation straight to your main support team so someone can pick it up.

• Add a tag. Label it for later. Tagging the chat with something clear like "fin_fail" or "negative_csat" is a huge help for tracking where Fin is struggling.

• Send a reply. Manage expectations. Send an automatic reply so the customer isn't left hanging. Something simple like, "Sorry my first answer wasn't quite right. I'm flagging this for my human colleagues, and they'll take a look shortly" works wonders.

Step 5: Test and set your workflow live

Don't just flick the switch and hope for the best. Use Intercom's Preview mode to run through the whole flow. Act like a customer, ask a question, and leave a bad rating. Make sure the handoff works just like you planned. Once you're happy with it, you're ready to go live.

Using the testing interface in Intercom to verify the automated follow-up workflow for when Fin cannot answer.

Extra tips

Setting up the workflow is a great start. But with a few small tweaks, you can make these AI-to-human handoffs even better.

Give Fin a second chance to understand

Sometimes Fin fails simply because the customer's first question was a bit fuzzy. You can give it a better shot by diving into the "Let Fin answer" settings and turning on "Ask for more information before handover." This way, if someone immediately asks to speak to a person, Fin will first try to get a bit more detail. It might be enough for Fin to solve the problem, and if not, your human agent gets way more context to work with.

Use tags for improvement

Remember those "fin_escalation" tags we set up? Don't let them just sit there. Make a habit of checking all the conversations with that tag every week or so. They're a goldmine. You'll see exactly which topics Fin fumbles, pointing you to the exact help articles you need to write or improve.

Intercom's reporting dashboard showing CSAT scores, useful for analyzing data from your efforts to Intercom automate follow ups when Fin cannot answer.

Avoid sending mixed signals

This one's a quick but important check. Make sure you've disabled any other global CSAT rules in your main Intercom settings. If you have two different satisfaction surveys firing off at once, you'll confuse your customers and potentially mess up the logic of the workflow you just built.

What if you need more control?

The Intercom workflow we just built is a great safety net. But as you handle more and more conversations, you might find it feels a bit... clunky. It's not always easy to test properly, and let's not forget you're still paying Intercom's $0.99 per-resolution fee even when Fin fails to resolve an issue. Sometimes, you just want more direct control instead of having to build clever workarounds.

That's where a tool like eesel AI comes into the picture. It's designed specifically for managing a support AI with more finesse. Instead of getting tangled up in branching workflow rules, you get a much more straightforward way to control how your AI behaves. You can hook up your Intercom account and get a much tighter grip on the whole process.

Here’s how eesel AI changes the game:

• Test on your actual conversations. Instead of one-off previews, eesel AI lets you run your AI against thousands of your past Intercom tickets. You can see exactly how it would have performed on real customer questions (the good, the bad, and the weird) before you ever let it talk to a live customer. No more guessing.

• You're in the driver's seat. You get to decide exactly what kinds of questions the AI is allowed to handle. Want it to only tackle password resets and pricing questions? Easy. Everything else can automatically go to a human, no complicated branches needed. You can also tweak its personality and tone with a simple editor so it sounds like it's actually part of your team.

• Pricing that makes sense. With eesel AI, you don't pay per resolution. That means you're not getting dinged every time the AI successfully closes a ticket or when it rightly passes a tough one to your team. The pricing is based on usage, so it's predictable and scales with you, without any weird penalties.

Wrapping up

An AI chatbot that drops the ball without a backup plan can do more harm than good. By setting up this automated follow-up in Intercom, you're building that crucial safety net. It protects your customers from frustrating dead ends and gives your team the exact feedback they need to get better.

And when you feel like you're outgrowing the workarounds and want more direct control over your support automation, tools like eesel AI are there to help you take the next step.

Want to see what a more controllable and transparent AI feels like? Give eesel AI a try for free.