Automate Lead Follow-Up with AI

To automate lead follow-up with AI means replying to new enquiries quickly and consistently with a workflow: the lead is saved, consent is checked, AI drafts a helpful reply from your own information, and a person approves anything sensitive before it is sent. Simple, pre-approved messages go out immediately, while anything requiring judgement waits for a staff member.

• Speed: An instant acknowledgement at any hour, followed by a genuine reply within working hours.
• Consent first: WhatsApp, SMS and email follow-ups go only to people who opted in.
• AI drafts: The model writes replies using your own FAQ document, never from guesses or general knowledge.
• Approval: Replies about treatment, pricing or complaints need a staff member's approval before sending.
• Stop rules: The sequence ends when the person books, replies or asks to stop.

This tutorial follows one example throughout, a dental clinic in Hyderabad that runs Click-to-WhatsApp ads and a website enquiry form for clear aligners and routine check-ups. Enquiries often arrive late at night and on Sundays, and the front desk replies several hours later, by which time many people have already booked appointments elsewhere. The clinic wants faster replies without ever letting software give dental advice.

Prerequisites to Automate Lead Follow-Up

• Lead sources: A website form and, if you run them, Click-to-WhatsApp ads.
• Consent fields: Separate, unticked boxes for "Contact me on WhatsApp" and "Send me emails about offers". Record the date and source of each consent, as the DPDP Act lesson explains.
• A WhatsApp Business Platform provider: Needed to send messages from a workflow. The WhatsApp Business API lesson covers providers and approved templates.
• A CRM or sheet: One record per lead, with source, consent status, stage and owner. See CRM in marketing.
• A workflow tool: This tutorial uses the pattern from marketing automation with n8n; Make or Zapier work the same way.
• A clinic FAQ document: Timings, address, parking, what to bring, how booking works. The AI may use only this.

Setup: Write the Rules Before the Workflow

1. List message types: Write down every kind of message the clinic receives, such as an acknowledgement, general information, appointment booking, a treatment question, a price question or a complaint.
2. Mark each one: Label every type either "send automatically with approved wording" or "AI drafts, staff approves". For the dental clinic, only the acknowledgement and general information are sent automatically.
3. Write templates: Create WhatsApp templates for the acknowledgement and reminders and submit them for approval through your provider.
4. Set working hours: The front desk approves drafts between 9 am and 8 pm, and outside those hours only the automatic acknowledgement is sent.

Step-by-Step: Build the Follow-Up Flow

Step 1: Capture and Store the Lead

Connect the website form and the WhatsApp provider's incoming-message webhook to your workflow tool. Save each lead to the CRM with: name, phone, source (form or ad), message, WhatsApp consent, email consent, and stage "new". Tag ad leads with UTM parameters or the ad's ID so you know which campaign brought them.

Step 2: Check Consent

Add a branch that checks consent before any message is sent:

• Opted in to WhatsApp, or they messaged you first: Continue to Step 3. A person who starts a WhatsApp chat opens a service window in which you can reply to them.
• Form lead without WhatsApp consent: Reply to their request only by the channel they chose, such as a phone call, and do not add them to any marketing sequence.

Step 3: Send an Instant Acknowledgement

Send the approved acknowledgement: "Thank you for contacting our clinic. Our front desk will reply between 9 am and 8 pm. Reply STOP to opt out." This is the only message that the workflow sends at night.

Step 4: Let AI Sort and Draft

Pass the message to an AI step with narrow instructions:

You help the front desk of a dental clinic in Hyderabad reply to enquiries.
Use only the clinic FAQ below. Never give dental advice, diagnoses or prices.
1. Classify the message as: general_info, booking, treatment_question, price_question, complaint, other.
2. Draft a short, polite reply in the language the patient used (English, Hindi or Telugu).
3. If the message is treatment_question, price_question or complaint, set needs_approval to true.
Reply only with JSON: {"type": "...", "needs_approval": true or false, "draft": "..."}
Clinic FAQ:
[paste timings, address, parking, booking steps]

Step 5: Route Sensitive Drafts to Staff

If needs_approval is true, send the draft to the front desk's approval queue, such as a shared inbox, a Slack channel or the provider's team inbox. The staff member reads the patient's original message alongside the draft, edits it where necessary, and approves it, and only then does the workflow send the reply. This pattern is called human in the loop.

General information drafts, such as directions or timings, can be sent without approval once the clinic has checked a few weeks of them.

Step 6: Set a Short Follow-Up Sequence

Messages sent after the service window closes must use approved templates.

Step 7: Handle Stop Requests at Once

If a person replies STOP, "no" or asks not to be contacted, the workflow immediately marks them as opted out in the CRM and ends every sequence for them. Check this opt-out list before every single send, not only when the sequence begins.

Step 8: Measure What Matters

Track the time to first reply, the number of leads who book, and the number who opt out. A rising opt-out count usually means the messages are too frequent or not useful enough. The lead scoring lesson shows how to rank leads once you have enough data.

Common Mistakes

• Messaging without consent: A form submission is not permission to send WhatsApp or SMS promotions, and bought or scraped contact lists must never be used.
• AI giving medical or price answers: An incorrect answer about treatment can harm both the patient and the clinic's reputation.
• Endless sequences: Too many reminders annoy people and lead to blocks and reports, which can hurt the clinic's WhatsApp account quality.
• Checking opt-outs only once: Someone who opts out on day 2 must never receive the day 4 reminder.
• No human owner: Every lead needs a named staff member who is responsible for following it up.

Next Steps

• Decide what stays human: Read marketing tasks you should not automate before adding more automatic replies.
• Chatbots: For a full conversational assistant, see AI chatbots for WhatsApp.
• Agents: Learn how multi-step agents extend this flow in AI agents for marketing.

How AI Changes Lead Follow-Up

What AI Automates Now

AI can read a message in English, Hindi or Telugu, work out what the person wants, draft a reply from the business's own information, and suggest which leads to call first. WhatsApp providers and CRMs increasingly include these features.

What Still Needs a Human

Answers about health, pricing, eligibility and complaints, the final decision on what a patient is told, and the phone conversation that turns an enquiry into a booking all stay with staff. A person also reviews the templates and the follow-up sequence whenever the clinic's services change.

Risk to Watch

An AI draft can sound confident and still be completely wrong, for example by promising a same-day appointment that does not exist. Keep approval switched on for sensitive messages, and never send health details to AI tools that are not approved for handling that information.

Do It with AI

Use this prompt to design your follow-up rules. It works in ChatGPT, Claude or Gemini.

1. Collect real enquiries and remove names and phone numbers.
2. Run the prompt and review the message types with the front desk.
3. Build the flow with approval turned on for every message at first.
4. After a few weeks, allow automatic sending only for types that never needed edits.

Check Before You Use It

• Facts: Every timing, address and booking step must come from the clinic's own records.
• Brand fit: Replies should sound like the front desk, warm and short, in the patient's language.
• Compliance: Confirm opt-in for every channel, honour stop requests at once, and keep health details out of general AI tools.