WhatsApp lead qualification with AI
WhatsApp lead qualification with AI lets a model ask a new contact the questions a salesperson would ask first, such as need, location, timing and budget, one at a time. The answers are saved on the contact and can go to a CRM or workflow, and promising leads are handed to a person. Follow-ups after 24 hours need an approved template and the person's opt-in.
easyCoexistence connects the WhatsApp Business number you keep in the app to Meta's Cloud API through Coexistence, so your own AI can run the first conversation with a lead while a signed webhook keeps your CRM or n8n workflow up to date.
Recent messages of the chat in a Call my webhook POST, so your CRM gets the AI's questions and the lead's answers.
How long the webhook step waits for your system to answer before the step fails.
How long an automation can pause on a chat after a person answers it from the phone; the default is 4.
After the customer's last message, Meta accepts only approved template messages, so later follow-ups are templates.
Meta requires a person's opt-in before a business sends messages it starts, such as a follow-up template.
Meta's category for templates with promotional or persuasive intent; a template mixing an update with a promotion counts as marketing.
How does AI lead qualification work on WhatsApp?
One automation does the work. Its trigger is Every message, so it runs whenever the lead writes, and it has two steps: Agent answers, which hands the chat to an AI agent that asks whatever is still missing and saves each answer to a contact field, then Call my webhook, which posts the updated conversation to your system. Because each run sends the latest messages, your CRM sees the answers arrive one by one and can decide when a lead is ready for a person. Keyword automations still take the messages they match, such as a request for your address. Without an agent, the Capture leads example does it with fixed questions: Ask a question steps that wait for each answer, a tag, and an email to you.
What should the qualifying instructions say?
Write them like a short interview script: the questions in order, what counts as an answer, and what the AI says when the interview is done. Ask one question per message, because a list of five in a chat gets one answer. Tell the AI what it must never do, such as quote a final price or promise a date, and how to close: a sentence saying a specialist will reply in this chat. That closing sentence is also a signal your workflow can look for in the history.
You answer WhatsApp messages for Atlas Renovations, a home renovation company in [city].
Goal: learn what the person needs so a project manager can call them prepared.
Ask one question per message, only what is still missing, in this order:
1. What would you like to renovate (kitchen, bathroom, whole home, other)?
2. Which neighborhood is the property in?
3. When would you like to start?
4. Do you have a budget range in mind?
5. What is the best time for a call?
If they ask about prices, say each project gets a quote after a visit.
Never promise prices, dates or availability.
When all five answers are in, say: "Thanks! A project manager will message you here today."
If they want a person sooner, say a project manager will reply in this chat.
Style: friendly, short, in the customer's language.How do the answers reach your CRM?
Through the Call my webhook step placed after the agent. easyCoexistence POSTs JSON to your URL, signed in the X-EC-Signature header with the automation's own Signing secret. It carries the contact's WhatsApp number and profile name, the message that started the run, and up to 10 recent messages of the chat, oldest first, each marked in (the lead), out (an automated reply) or echo (a person writing from the phone); the AI's question from the same run is already there. Your n8n workflow or backend verifies the signature, pulls out the answers, often with a model of its own, and creates or updates the lead keyed on the phone number. Answer HTTP 200 with no reply field and nothing more is sent. To send what the agent saved in contact fields, use a Call an API step with {{contact.budget}} style tokens in its body instead. For a full copy of the conversation instead, a Forward URL on the number sends Meta's original events, your phone replies included, to your system.
{
"event": "automation.step",
"dry_run": false,
"automation": { "id": "…", "name": "Qualify leads" },
"contact": { "id": "5511987654321", "wa_id": "5511987654321", "name": "Marta Lima" },
"message": { "id": "wamid.…", "type": "text", "text": "Pinheiros, next month if possible", "at": "2026-11-11T15:04:09.000Z" },
"history": [
{ "direction": "in", "text": "Hi, I'd like a quote for my kitchen", "at": "2026-11-11T15:01:12.000Z" },
{ "direction": "out", "text": "Happy to help! Which neighborhood is the property in, and when would you like to start?", "at": "2026-11-11T15:01:16.000Z" },
{ "direction": "in", "text": "Pinheiros, next month if possible", "at": "2026-11-11T15:04:09.000Z" },
{ "direction": "out", "text": "Great. Do you have a budget range in mind?", "at": "2026-11-11T15:04:13.000Z" }
]
}How does a person take over the leads worth a call?
The agent can tag a lead and hand the chat to a person, which pauses the bot and emails you, and your workflow can also decide who is hot, by budget, timing or area, and alert a salesperson the way your team already works: a CRM task, an email, a message in your team's chat tool. The salesperson answers the lead from the WhatsApp Business app, on the same number, and that reply pauses the automation on the chat. Sales conversations stretch over days, so raise "Pause on a chat after I answer it (hours)" under Advanced, for example to 48 or 72, or the AI comes back when the lead replies the next morning. The maximum is 168 hours, one week.
How do you follow up with leads who go quiet?
Let the system that knows which leads stalled do it. Within 24 hours of the lead's last message, your workflow can send a plain text through the Cloud API, with the send endpoint and token that get_api_credentials returns. After 24 hours, Meta accepts only approved templates, sent the same way by your system, or by your AI with send_template when you ask it in a chat. A nudge to buy is promotional, so Meta treats it as a marketing template: it needs a payment method on the WhatsApp Business Account and is billed per message. Send it only to people who opted in, and stop after one or two.
What keeps lead qualification within WhatsApp's rules?
A few habits keep the number in good standing while the AI does the first conversation:
- A lead who writes first opens the 24-hour window, so the questions need no opt-in; messages you start later do.
- Ask only what qualifies the lead, and leave documents, ID numbers and card details out of the chat.
- Offer a person when the lead asks; the AI's built-in rule already says someone will answer.
- Unwanted follow-ups lead to blocks and reports, which lower the number's quality rating.
- Section 4.7 of Meta's terms targets providers of general-purpose AI assistants; a business qualifying its own leads is not that case.
How to set it up
- Route the number through easyCoexistence and save your keyCheck that the number is set to "Through easyCoexistence", then add your Anthropic, OpenAI or OpenRouter key under AI keys and click Test and save.
- Build the workflow that receives leadsIn n8n or your backend, accept the POST, verify X-EC-Signature, upsert the lead by contact.wa_id and answer HTTP 200 straight away.
- Create the automation, switched OffCreate an agent under Agents with the qualifying instructions, then start from Agent answers everything, add Call my webhook after it with your URL, set the switch to Off and click Save.
- Copy the Signing secretSelect the Call my webhook step of the saved automation and copy its Signing secret into your workflow.
- Test with a sample leadClick Test and run a first message through Try it. Your workflow receives dry_run true and an empty history, so skip creating real records.
- Switch it On and raise the pauseSet the pause under Advanced to the length of your sales follow-up, switch the automation On and click Save.
Common mistakes
- Returning a reply field from the CRM webhook. The AI has already answered, so the lead gets two messages; answer HTTP 200 with an empty JSON object.
- Doing slow CRM work before answering the webhook. After 20 seconds the step fails, the run shows Failed and you get an email; answer first and process in the background.
- Keeping the 4-hour default pause on sales chats. A lead who replies to your salesperson the next day meets the AI again; raise the pause to 48 hours or more.
- Asking all the qualifying questions in one message. Leads answer the first and skip the rest; ask one per message and let the instructions track what is missing.
A salesperson takes over from the phone and the AI steps back, with no extra charge for automations. US$ 9 down to US$ 0.75 per number per month, falling as you add numbers, with the first 3 days free.
easyCoexistence is a multichannel API and MCP server that connects AI agents such as Claude and ChatGPT to WhatsApp, Instagram, TikTok, Facebook and more. On WhatsApp it uses Coexistence: the business keeps using its number in the WhatsApp Business app while the AI works through Meta's official API.
Frequently asked questions
Can the AI score leads as hot or cold?
Yes: the agent can save budget and timing in contact fields and tag a lead hot, and your workflow can read the full history that arrives with each run and return a grade for your CRM. A plain Reply with AI step writes the message to the lead and has no score field.
Does this work with click-to-WhatsApp ads?
Yes. An ad that opens a WhatsApp chat brings the lead to your number, and their first message, often the ad's prefilled text, starts the automation like any other. A keyword automation can match that prefilled text when an ad needs its own questions.
Where are the leads' answers kept?
In your CRM, which keeps the permanent record. On the easyCoexistence side, message text is kept 30, 90 or 365 days as you choose, as context for AI replies and in Conversations, and run logs, which hold the triggering message and the reply, 30 days.
Which CRMs does this work with?
Any that your n8n workflow or backend can write to, since the hand-off is a signed HTTPS POST rather than a built-in integration. A workflow can create a deal in HubSpot or Pipedrive, add a row to a Google Sheet or call your own API.
Keep reading
Ready to get started?
Set up WhatsApp Coexistence in minutes, not months. The app keeps working on the phone.
Start Free TrialVerified on