How to Qualify Leads With a Custom Bot Before Fin AI or Your Sales Team Gets Involved

Most businesses running Intercom point Fin AI at support tickets and call it a day. Meanwhile, the same chat widget is sitting on their pricing page, catching pre-sales visitors who get exactly the same generic greeting as someone reporting a bug. That's a missed opportunity, and it's also a bad experience for a prospect who just wanted to know if your product fits their situation before they talk to a human.

A Custom Bot is the right tool for this job, not Fin AI. Fin AI is built to research your help content and generate an answer to a support question. Lead qualification is a different kind of task: you already know the questions you want asked, you know what a good answer looks like, and you know exactly what should happen next depending on the response. That's a job for a fixed, branching flow, not a language model improvising its way through a conversation.

Why This Belongs in a Bot, Not an AI Agent

Qualification only works if it's consistent. If you ask ten prospects about company size, industry, and timeline, you want ten clean, comparable answers sitting in your CRM afterward, not ten answers phrased ten different ways because an AI agent decided to handle the conversation loosely. A Custom Bot flow asks the same question the same way every time, captures the response in a fixed format, and only branches where you've told it to branch. That predictability is the whole point.

This also keeps Fin AI focused on what it's actually good at. If your Fin AI setup is tuned on support content, a sales conversation is off its home turf. Sending a bot to handle qualification first, then routing to Fin AI for product questions or to a rep for anything that needs a human, keeps both tools doing the job they're built for.

Pick a Short List of Questions That Actually Matter

Before building anything, decide what "qualified" means for your business. Common qualifying attributes include company size, role or seniority, industry, current tooling, and rough timeline to purchase. The mistake most people make here is asking too much. A long intake form inside a chat widget feels like a job application, and visitors bail out halfway through.

Pick three or four fields that would actually change how you respond to someone. If company size determines which plan you'd recommend, ask it. If you only sell to a specific industry, ask that early so you can route poor fits to self-serve content instead of your calendar. Everything else can wait until a real conversation.

Building the Flow

Inside Intercom, a Custom Bot flow is built around triggers, questions, and branches. Set the bot to fire on specific conditions, for example a visitor on your pricing page who hasn't been previously identified as a customer, rather than showing it to everyone site-wide. Ask your first qualifying question, then use the visitor's answer to branch the conversation. Someone from a large company might get a different next question than a solo founder browsing a starter plan.

Keep the tone conversational rather than form-like. Instead of "Please select your company size from the following options," try something closer to how you'd actually ask it face to face. Answer any question the visitor has first if they've asked one, then move into your questions. Being upfront about why you're asking, and honest if it turns out your product isn't a great fit, builds more trust than pushing everyone toward a sales call regardless of fit.

Save the Answers Somewhere Your Team Can Use Them

The step people skip is making sure the bot's answers actually go somewhere useful. Intercom lets you capture responses from a Custom Bot flow and save them directly as attributes on the lead's profile, rather than letting that information disappear once the chat ends. Set up custom attributes ahead of time for whatever you're capturing, whether that's industry, use case, or budget range, using text, number, list, or true/false fields depending on the data. Once the bot writes those answers to the profile, they're available to anyone who picks up the conversation later, and they can feed segments or nurture sequences for prospects who weren't ready to buy yet.

Route and Tag Based on the Answers

Qualification data is only valuable if it triggers action. Configure your flow so that answers indicating a strong fit route the conversation to a specific salesperson or team, while weaker fits get tagged and pointed toward help content, a pricing page, or a lower-touch signup path. Some setups take this further by pushing a new record into a connected CRM the moment a lead clears your qualification bar, so a rep sees it show up without anyone manually copying details over.

This is also where the earlier decision about which questions matter pays off. If you asked good, decision-relevant questions, your routing rules practically write themselves. If you asked ten vague questions, you'll end up building routing logic that nobody trusts.

Where a Human Still Belongs

None of this replaces manual qualification entirely. Reps chatting live can still update these same fields on the spot when a prospect mentions something the bot didn't ask about. The bot's job is to handle the repetitive first pass so your team spends time on judgment calls instead of typing the same three questions into every new conversation.

A Few Things That Trip People Up

Businesses that try this without planning tend to run into the same problems. They build one enormous bot flow trying to qualify and support at the same time, which confuses visitors who just wanted a quick answer. They ask questions that don't map to any actual routing decision, so the data sits unused. Or they never test what happens when someone gives an unexpected answer, and the flow dead-ends instead of gracefully handing off to a human.

Start narrow. Pick one page or one audience segment, build a short flow around the two or three questions that matter most, and watch how real visitors respond before expanding it.

Setting up a qualification flow that actually routes correctly and plays well alongside Fin AI takes some trial and error to get the triggers, branching, and handoff rules right. If you'd rather have someone who does this for a living set it up and tune it against your real traffic, feel free to reach out and I can help you build it out properly.

Cover photo by Headway on Unsplash.