How to integrate WhatsApp AI agent with n8n and CRM
By Catalina Jaramillo · AI Commerce & Revenue Operations Consultant · 2026-10-19
You can connect a WhatsApp AI agent to n8n in a few hours and push every conversation into your CRM automatically. The workflow uses the WhatsApp Business Cloud API, n8n’s visual editor, and a simple webhook to keep lead records up‑to‑date.
What does a WhatsApp AI agent do and why use n8n?
A WhatsApp AI agent answers common questions, qualifies leads, and creates or updates records in your CRM without human intervention. It works 24/7, reduces manual data entry, and ensures every prospect follows the same qualification path.
- Instant replies powered by OpenAI or other LLMs
- Lead capture (name, phone, intent) directly from chat
- Automatic routing to a human agent when needed
- Scalable orchestration using n8n’s low‑code workflow engine
Which tools and credentials do you need?
All components are cloud‑based, so you only need accounts and API keys. The setup works with any CRM that offers a REST API (HubSpot, Salesforce, Zoho, Pipedrive, etc.).
- WhatsApp Business Cloud API account (Meta for Developers)
- n8n Cloud or self‑hosted instance (free tier is enough for testing)
- OpenAI API key (or another LLM provider) for the AI layer
- CRM API token or OAuth client credentials
- A publicly reachable URL for n8n webhooks (ngrok works for local testing)
Carolina, Catalina's AI agent, reviews your business, shows which system solves it and what it would cost. If there is a fit, she books a call with Catalina.
Tell Carolina about your case →How to build the WhatsApp AI flow in n8n?
Create a new workflow that starts with a webhook triggered by incoming WhatsApp messages. The webhook parses the payload, sends the user’s text to an LLM, and returns the AI response back to WhatsApp.
- Add a **Webhook** node – set the HTTP method to POST and copy the generated URL.
- Configure the **Meta WhatsApp** node (or HTTP Request) to send the AI reply using the /messages endpoint.
- Insert a **ChatGPT** (or OpenAI) node – map the incoming message text to the prompt field.
- Use **IF** nodes to detect intent (e.g., “price request”, “schedule demo”) based on keyword matching or LLM confidence.
- For qualified leads, trigger the CRM sync nodes (see next section).
How to connect n8n to your CRM and sync lead data?
After the AI determines that a conversation contains a new lead, the workflow pushes the data to the CRM. Most CRMs expose a simple create‑or‑update endpoint.
- Add the CRM’s **HTTP Request** node – set method to POST/PUT, include the API token in the header.
- Map fields: phone number, name (if collected), last message, lead status, source = “WhatsApp”.
- Use an **Set** node to format dates and add a unique identifier (e.g., WhatsApp message ID).
- Optionally, add a **Delay** node to respect rate limits, then a **Send Email** node to notify the sales team of a hot lead.
What are the common pitfalls and how to avoid them?
Even a short workflow can break if you overlook API limits or data mapping errors. Below are the most frequent issues and quick fixes.
- **Message throttling** – Meta caps outbound messages per phone number. Use a queue (n8n’s “Wait” node) to spread sends over time.
- **Incorrect phone formatting** – Store numbers in E.164 format (+1XXXXXXXXXX) before sending to the CRM.
- **LLM hallucinations** – Keep prompts short and add a “Only answer with Yes/No or a short phrase” guard clause.
- **Missing webhook verification** – Meta expects a 200 OK response within 10 seconds; return it immediately before processing the AI step.
- **CRM field mismatches** – Test the payload with a sandbox record first; n8n’s “Execute Workflow” debug view helps spot missing keys.
How much does it cost to implement this solution with Catalina?
Catalina offers a modular pricing model that matches the scope of the automation. Below is a quick reference for the items you’ll need.
- Strategic diagnosis (30‑minute audit) – **USD 490**
- Automation express (basic WhatsApp‑n8n‑CRM flow) – **USD 500 – 1,500**
- Revenue Agent Sprint (full AI prompt engineering + monthly monitoring) – **USD 2,500** configuration + **USD 750/month**
- Ongoing optimization and continuous growth – **USD 1,000 – 2,000/month**
How would Catalina approach the project?
Catalina starts with a data‑driven diagnosis, then designs a pilot that can be measured before scaling. The process is transparent and hands‑off after the handover.
- 1️⃣ **Diagnostic session** – Review your current lead capture, WhatsApp usage, and CRM schema (USD 490).
- 2️⃣ **Solution design** – Map each user intent to a workflow node, define success metrics (conversion rate, response time).
- 3️⃣ **Pilot build** – Implement a minimal viable flow covering greeting, qualification, and CRM sync; run for 2 weeks.
- 4️⃣ **Metrics review** – Compare inbound leads, duplicate records, and response latency before vs. after.
- 5️⃣ **Scale** – Add advanced features (payment links, multi‑agent handoff) and set up a monthly monitoring package.
FAQ
Do I need a dedicated phone number for WhatsApp Business?
Yes. Meta requires a verified Business Phone Number for the Cloud API. The number can be a virtual line, but it must be approved through the WhatsApp Business Manager.
Can I use any CRM, or is it limited to specific platforms?
Any CRM that provides a REST API can be connected. n8n has built‑in nodes for HubSpot, Salesforce, Zoho, and Pipedrive, and a generic HTTP node for custom systems.
How long does the initial setup usually take?
With the required credentials ready, a basic WhatsApp‑n8n‑CRM flow can be built and tested in 3‑5 hours. The diagnostic and pilot phases add 1‑2 days of planning and validation.
What if the AI gives an incorrect answer to a prospect?
Add an “Escalation” branch that routes the conversation to a human agent when confidence is below a threshold or when the user types “agent”. This keeps the experience safe and professional.
Is there a measurable ROI for this automation?
Catalina measures ROI by tracking lead conversion rate, average response time, and manual effort saved. In the pilot stage you’ll see concrete numbers to decide on further investment.
Carolina, Catalina's AI agent, reviews your business, shows which system solves it and what it would cost. If there is a fit, she books a call with Catalina.
Tell Carolina about your case →