levelpartner.ai
AI agents and AI for business

Automated chatbot for website: how it works and how to set it up

An automated chatbot for website enquiries is only half automated if the conversation ends when the chat window closes. The useful part is what happens next: the contact is created, the opportunity lands in the right pipeline stage, the right person is told, and a follow-up goes out if the visitor did not book. This article covers that second half on the levelpartner.ai platform, wiring the chatbot into workflows so no conversation is left sitting in an inbox.

Where chat leads usually go missing

Most firms judge their chatbot on conversations held. The real loss happens after. A visitor gives their name and email, asks two questions and leaves without booking. The transcript sits in the inbox. Nobody owns it. Three days later someone notices, sends a generic email, and the visitor has already spoken to a competitor.

Automation fixes this by giving every chat outcome a defined next step. Booked, handed off, qualified but not booked, and not a fit each trigger something different, and none of them depend on someone remembering.

The four outcomes to plan for

  • Booked. Confirmation, reminders and an opportunity in the booked stage.
  • Handed off. An assigned owner, a notification and a deadline for the first reply.
  • Qualified, not booked. A short follow-up sequence with a booking link.
  • Not a fit. A polite close, a tag and no further sales messages.

Write these down before you build anything. If your team cannot agree what "qualified" means, the workflow will not fix that.

Before you start

  • A live chat agent with qualifying questions and Appointment Booking enabled.
  • A pipeline with stages that match how your team actually sells.
  • Tags for each chat outcome, created in advance.
  • Follow-up message copy for the qualified-but-not-booked path.
  • A named owner for hand-offs during business hours.

Where each outcome comes from

Each path needs a signal the workflow can react to, and the agent's own settings supply most of them. A booking made through Appointment Booking creates an appointment, which a workflow can use as its trigger. The Human Handover action adds a tag of your choice and assigns a user, so the hand-off path can start from that tag. In Bot Goals, Trigger Workflow starts an automation once the agent has collected the details you asked for, which suits the qualified-but-not-booked path, and Update Contact Field saves each qualifying answer, such as budget and timescale, to the record. The not-a-fit path is usually a workflow that reads those fields and closes the opportunity.

How to set it up

Build one workflow per outcome, using the same pattern each time.

  1. Open Automation and select Workflows.
  2. Choose Create Workflow to start from scratch.
  3. Click Add New Trigger and pick a trigger for one outcome.
  4. Add trigger filters to narrow when it fires.
  5. Click Add Action and select the action you want.
  6. Test with a new contact that has never entered the workflow.

What changes for the sales team

With these workflows in place, the sales team stops reading chat transcripts to decide what to do. They open the pipeline and see booked calls, assigned hand-offs and follow-ups in progress. The chatbot becomes a front end to the pipeline rather than a separate tool someone has to check.

The weekly effort is small: about fifteen minutes to check that each path fired correctly and that no hand-off sat unanswered. If one path never triggers, look first at the signal it depends on in the agent's settings.

Illustration: a 50-person agency adding outcome workflows

Imagine the chatbot handles about 120 conversations a month and books 18 calls, while around 40 visitors leave contact details without booking and receive nothing. If a qualified-not-booked sequence turned 11 of those 40 into bookings within a week, calls booked from chat would rise by more than half. Notifying the owner directly on a hand-off would also cut the reply time from most of a day to under an hour.

The common mistake

Sending the not-booked follow-up the moment the chat closes feels like pressure. Visitors often leave to check a diary or ask a colleague. Wait an hour or two before the first message, and make it helpful rather than a repeat of the booking link.

When not to automate the follow-up

If your chat volume is under a handful a week, a person can follow up each one personally and do it better. Automate when the volume is too high for someone to read every transcript the same day. Avoid automated sales follow-up entirely on support conversations with existing clients; route those to account owners instead.

Troubleshooting

Workflows never trigger. The signal is missing. Check that the handover action adds the tag the workflow listens for, and that Trigger Workflow in Bot Goals points at the right workflow.

Contacts receive two confirmations. Both the calendar and the workflow send one. Turn off one of them.

People who booked still get follow-ups. The workflow has no goal or exit condition. Add a goal for an appointment booked so the contact leaves the sequence.

Opportunities pile up in the wrong stage. The workflow creates a new opportunity each time. Set it to update an existing opportunity for the contact when one exists.

Frequently asked questions

Does the chatbot need to be AI for this to work?

No. The workflows react to tags and fields, so they work with a simpler rules-based chat too. AI simply makes the conversation and the qualification more accurate. The AI sales assistant article compares the two.

Can the follow-up go by SMS instead of email?

Yes, if the contact gave a mobile number and consent. SMS usually gets a faster reply. See AI text messaging for wording that suits the channel.

How fast should a hand-off be answered?

Within the hour during business hours is a sensible target. Set the alert to match whatever standard you choose, and review misses weekly. The AI follow up guide covers chasing rules.

Where do I start if the chatbot is not live yet?

Build the conversation first with the website chat bot setup guide, then come back and add these workflows once you have a week of real transcripts.

The journey this guide describes

Instant lead response

  • Trigger

    Form or ad lead

    Website form, ad form or portal email

  • Trigger

    Message or missed call

    Chat, social, WhatsApp or phone

  1. Enquiry captured

    Every channel lands in one contact record with its source, so nothing sits in a personal inbox.

    Instant
  2. Find the contact

    Match on email or phone

    • New contactNot in the CRM yet
      1. First reply sent

        A personal reply goes out on the channel the enquiry came in on, with a link to book.

        Under 60 seconds
      2. Wait for an answer

        Up to 2 minutes

      3. Intent read

        AI reads the message and labels what the person wants, so the right follow-up starts.

        Seconds
        • Ready to talkWants a call or a time
          1. Call-back within 5 minutes

            The owner is connected to the lead

          2. Owner alerted

            The person responsible gets the enquiry, the history and the next step on their phone.

            Instant
          3. Deal opened

            A deal is created in the pipeline with its source, so the enquiry is tracked to the outcome.

            Instant
          4. Done
        • Just lookingNo time yet
          1. Moved to nurture

            Useful follow-up on a rhythm

          2. End
    • Existing clientAlready in the CRM
      1. Sent to their account owner

        With the full history attached

      2. Owner notified

        By text and in the app

      3. End