AI Chatbot Lead Qualification: Qualify and Route Leads in Chat

AI chatbot lead qualification means your website chatbot asks a visitor the few questions a salesperson would ask first (what they need, how big the team is, how soon), saves the answers with their contact details, and sends good-fit leads to the right next step: a booked call, your CRM, or a person in your team inbox. It runs while the visitor is still on the page, so nobody waits for a reply to a form.

This guide shows how to set it up, what to ask, how routing works, and exactly what Dante AI does today, including how leads reach HubSpot.

What is AI chatbot lead qualification?

Lead qualification is deciding whether someone is worth a sales conversation now, later, or never. A chatbot does it inside the conversation instead of on a static form:

• It answers first. The visitor gets a real answer to their question, trained on your website and documents, before being asked for anything.
• It asks at the right moment. When the visitor asks about a demo, pricing or a call back, the chatbot asks for contact details and one or two qualifying answers.
• It routes. A strong fit is offered a meeting or handed to a person; a weak fit gets a self-serve answer; every captured lead lands in a list your team (or your CRM) can act on.

The point is not to replace your sales team. It is to make sure the people they talk to are the ones who are ready, and that nobody who was ready leaves because the form felt like homework.

How does a chatbot compare with a form and a human SDR?

Three common ways to qualify inbound leads, side by side. This is a comparison of approaches, not of vendors.

Most teams end up with the chatbot doing first-touch triage and the SDR taking the strong fits. For the form side of this in more depth, see AI chatbot vs contact form.

What should a chatbot ask to qualify a lead?

Fewer questions than you think. Each one costs you some visitors, so ask only what changes the route:

1. Contact: name and email, so someone can follow up.
2. One fit question: the thing that decides whether sales should call, such as team size, number of locations or monthly volume.
3. Intent, from the conversation itself: a visitor asking about pricing for 50 seats has told you more than any dropdown. Let the chatbot pick that up instead of asking for it.

In Dante AI, the lead capture form holds up to 3 fields. The defaults are Name and Email (required) and Phone (optional), and you can rename a field, so the third one can become "Team size" or "When do you want to start?". Our own docs give the same advice: ask for name and email on first contact and save deeper qualifying questions for the follow-up.

How do you route a qualified lead to sales or your CRM?

Routing happens in two places: inside the chat, and after the lead is saved.

Inside the chat

• Book a meeting. Connect Calendly, Cal.com, Google Calendar or Outlook and turn booking on for the agent. The agent shows live openings and the visitor picks a slot in the chat (Starter and up).
• Hand over to a person. With Human Handover on, a "Talk to a human" pill sits under every reply, and a typed request works too. The chat lands in your team Inbox as Awaiting, a teammate clicks Take it, and the AI stops replying while they talk (Advanced and Pro).
• Self-serve. For a weak fit, the agent keeps answering and points to your sign-up or help pages. You steer this with the agent's Behavior instructions, for example "If the visitor has fewer than 10 staff, point them to the self-serve plan instead of offering a demo."

After the lead is saved

• Leads page: every capture appears under Monitor, then Leads, with the full chat one click away in Chat history. Export all or selected rows as CSV, JSON or PDF.
• Contacts: the same person across agents and channels is one contact, matched on exact email or phone.
• Zapier: the New lead trigger fires each time a visitor's details are captured, carrying the form fields, the agent name and the conversation id. From there you build the Zap you want with the apps in the Zapier catalog, for example adding the lead to your CRM or posting it to a sales channel, with the fit answer mapped across.

One gap to plan for: Dante AI does not send email alerts for new leads yet. If you want an instant alert, add a notification step to the Zap. For more Zapier setups, see how to connect your AI chatbot to Zapier and a CRM.

Can Dante AI send leads to HubSpot?

Yes, through Zapier. Dante AI does not have a native, direct HubSpot integration today. The shipped path is:

1. In Dante AI, go to Integrations and generate a Zapier key (Starter, Advanced and Pro).
2. In Zapier, connect the current Dante AI app with that key and pick the New lead trigger.
3. Add HubSpot as the action app in the same Zap and map the Name, Email and your renamed third field to HubSpot contact properties.

Dante signs each event it sends to Zapier and retries delivery, up to three attempts, if Zapier is briefly unreachable. There is no raw webhook to configure by hand: Zapier registers it for you. If you prefer no Zap at all, export the Leads list as CSV and import it into HubSpot. The same Zapier route works for Salesforce, Pipedrive, Google Sheets and other apps in the Zapier catalog.

Note the plans: Zapier works from Starter, but the lead capture form is on Advanced and Pro. On any plan, a visitor who asks for a person while Human Handover is off can leave a name and email (the "Offer to take contact details" switch, on by default), and that is saved as a lead too.

What does it look like end to end? A worked example

An invented business: a company selling scheduling software to clinics, on the Advanced plan. Sales wants demos with clinics of 10 or more staff; smaller clinics should sign up on their own.

Setup:

1. Lead capture fields: Name and Email (required), and the third field renamed to "Number of staff".
2. When to show: "Show on demo, pricing, or booking asks."
3. Behavior instruction: "Answer the question first. If the visitor mentions 10 or more staff, offer to book a demo. If fewer, recommend the self-serve plan and link the sign-up page."
4. Booking: Calendly connected, a 30-minute demo set as the default event type.
5. Human Handover: on, so a visitor can ask for a person at any point.
6. Zapier: New lead trigger, then HubSpot, with "Number of staff" mapped to a contact property so sales can see the fit at a glance.

The conversation, at 21:40 on a Tuesday:

• Visitor: "Does this work for a clinic with three sites? We have about 25 staff."
• Agent answers from the website: multi-site support, how staff calendars work.
• Visitor: "What would it cost us?" That matches the When to show instruction, so the agent shows the lead form first.
• Visitor fills in name, email and 25 for "Number of staff". The agent answers the pricing question, then offers demo times from the live calendar. The visitor picks Thursday at 10:00.

What happens next, without anyone touching it: the lead appears on the Leads page, the meeting is on the calendar, and the Zap creates the contact in HubSpot with 25 in the staff field. On Wednesday morning the sales rep opens the lead, reads the full chat, and goes into Thursday's demo already knowing the clinic has three sites.

A different visitor with 4 staff gets the self-serve recommendation, and if they leave details they still land on the Leads page and in HubSpot, marked with 4 staff, so nobody wastes a demo slot on them. Test both paths in the preview before you publish: the trigger instructions use the agent's judgement, so check that your wording does what you expect.

Before you build anything, it helps to know what slow or missing replies cost you today. The calculator further down this page turns your own numbers into a monthly figure.

What usually goes wrong, and how do you avoid it?

• Asking before answering. A form before any answer feels like a gate. With the default When to show instruction, Dante answers first; a custom instruction shows the form before the answer when a message matches it, so keep custom triggers to real buying signals such as pricing or demo asks. Switch on "Require details before answering" only if you really need details up front.
• Too many questions. Three fields is the limit in Dante for a reason. Keep the rest for the follow-up.
• Promising a person nobody provides. If no one reads the Inbox or the leads list, turn off the offer to take contact details rather than leave visitors waiting.
• No alert. Leads sitting unread overnight lose their edge. Add a Slack or email step in the Zap.
• Routing rules nobody tested. Run a strong-fit and a weak-fit chat in the preview, then check where each lead landed.

For when a chat should go to a person, see AI chatbots with human handover. For booking in the chat, see AI chatbots for appointment scheduling. For the wider sales picture, see AI sales agents.

Sources

• Dante AI: Lead capture docs, Leads docs, Contacts docs, Integrations docs, Inbox docs, Agent behaviour docs, Plans docs and the pricing page (each read 25 October 2026).