Step-by-Step: How to Build Your First n8n Automation for a Small Business Client

You've signed your first client. Congratulations — now comes the part that actually matters: delivering. The gap between knowing n8n conceptually and building a production automation for a real paying client can feel daunting. This guide closes that gap.

We're going to walk through the most common and highest-value automation for a local small business client: a complete missed call text-back and lead follow-up system with AI-personalized responses and appointment booking. This is the workflow that agencies sell for $1,500-$3,000 setup and $500-$997/month recurring, and it typically takes 3-6 hours to build once you know the steps.

Before diving in, if you're still evaluating whether to use n8n, Make.com, or Zapier, read our comparison guide first. For this walkthrough, we're using n8n self-hosted because it offers the best combination of flexibility, cost, and AI capability.

What We're Building

The complete automation flow we're building handles three scenarios:

Scenario A — Missed call: Someone calls the business, doesn't leave a voicemail or gets sent to voicemail. Twilio detects the missed call and triggers our workflow. Within 60 seconds, the caller gets an AI-personalized text message.
Scenario B — Web form submission: Someone fills out the contact form on the client's website. The form data triggers the workflow. Within 60 seconds, the prospect gets a text from the business number asking a qualifying question.
Scenario C — Multi-step follow-up: If the prospect doesn't respond to the first message in 2 hours, they get a follow-up. Then again at 24 hours, 48 hours, and 5 days. The AI adapts the message at each step to avoid sounding robotic.

Prerequisites: What You Need Before Starting

n8n instance: Self-hosted on Railway, Render, or DigitalOcean (costs $5-$20/month). You can also use n8n Cloud ($20/month) to get started quickly.
Twilio account: A Twilio phone number for the client's business ($1/month). Twilio credentials (Account SID and Auth Token) from twilio.com/console.
OpenAI API key: From platform.openai.com. You'll use GPT-4o-mini for text generation (costs about $0.01-$0.05 per message generated).
Cal.com account: For appointment booking. Free tier works. You'll need the client's API key from their Cal.com account.
Airtable or Google Sheets: To log every lead and track conversation status (which stage of follow-up each lead is in).
Client's website contact form: You'll need to add a webhook URL to their form to send submissions to n8n. This works with Gravity Forms, Elementor, Webflow, or any form that supports webhooks.

Part 1: Building the Missed Call Text-Back (Scenario A)

Step 1: Configure Twilio for the Client

Buy a local phone number from Twilio that matches the client's area code. In the phone number settings, set the "A Call Comes In" webhook URL to point to a TwiML endpoint that you control — or simply set the number to forward calls to the client's existing business line while Twilio logs the call.

The critical configuration: in Twilio, set up a "Status Callback URL" for calls. This URL will receive a webhook every time a call is completed, including calls that went to voicemail or were missed. Set this status callback URL to your n8n webhook endpoint (which you'll create in the next step).

Step 2: Create the n8n Webhook Trigger

In n8n, create a new workflow. Add a "Webhook" node as your trigger. Set it to POST method. Copy the generated webhook URL — this is what you'll paste into Twilio's status callback URL field.

In the webhook node settings, set "Response Mode" to "Immediately" so Twilio doesn't time out waiting for your workflow to complete.

Step 3: Add a Filter for Missed/No-Answer Calls

Not every call status update should trigger a text. You only want to respond to missed calls and calls that went to voicemail. Add an "IF" node after the webhook with this condition:

Condition: CallStatus equals "no-answer" OR CallStatus equals "busy"
If TRUE: continue to the AI message generation step
If FALSE: stop the workflow (do nothing)

Step 4: Generate the AI Text Message

Add an "OpenAI" node. Select the "Message a Model" operation. Use "gpt-4o-mini" as the model (fast and cheap).

System prompt for the AI node:

"You are a friendly assistant for [Client Business Name], a [niche] business. When a customer misses a call, your job is to send them a warm, helpful text message that sounds natural and human — never robotic or automated. Keep it under 160 characters. The message should acknowledge that they tried to reach us, express that we're sorry we missed them, and ask a simple open-ended question to start a conversation. Do not mention AI or automation. Sign off with the business name."

User message: "Write a missed call text message for a customer who just called [Client Business Name] at [current time]. Their phone number is [caller phone number from Twilio webhook data]."

Step 5: Send the SMS via Twilio

Add a "Twilio" node. Select "Send SMS" operation. Fill in:

From: The client's Twilio number
To: The caller's phone number (from the Twilio webhook data: CallerNumber field)
Body: The AI-generated message from the previous node

Step 6: Log the Lead

Add a Google Sheets or Airtable node to log the interaction. Create a row with: phone number, call time, text sent (yes/no), timestamp of text, follow-up status (set to "stage 1"), and any other relevant data. This log is critical for the follow-up sequence in Part 3.

Part 2: Building the Web Form Lead Response (Scenario B)

Step 1: Create a Second Webhook for Form Submissions

Create a new workflow in n8n (or add a branch to your existing one). Add a webhook node for POST requests. This webhook URL goes into the client's contact form webhook setting.

Test it by submitting the contact form yourself. You should see the form data appear in n8n's execution log. Common fields you'll receive: name, email, phone, message, and sometimes service type.

Step 2: Generate a Personalized Response

Add an OpenAI node. This time, you can use more context because you have their name and message:

System prompt: "You are a friendly assistant for [Client Business Name]. A potential customer just submitted an inquiry. Write a warm, personalized SMS response that acknowledges their specific inquiry, introduces the business, and asks one qualifying question. Keep it under 200 characters. Do not mention AI."

User message: "Customer name: [name from form]. Their message: [message from form]. Service they're interested in: [service field if available]. Write the response."

Step 3: Send SMS, Log Lead, Update CRM

Same as Part 1 Steps 5-6. Additionally, if the client uses a CRM like HubSpot, GoHighLevel, or Pipedrive, add a node to create a contact and deal in their CRM with all the form data pre-populated.

Part 3: Building the Multi-Step Follow-Up Sequence (Scenario C)

This is where the real revenue protection happens. Most of your clients' leads go cold not because prospects aren't interested, but because nobody followed up consistently.

The Follow-Up Architecture

Build a scheduled workflow that runs every hour and checks your lead log for any leads that are:

In "stage 1" and haven't responded in 2 hours → send follow-up text #1, update to "stage 2"
In "stage 2" and haven't responded in 24 hours → send follow-up #2, update to "stage 3"
In "stage 3" and haven't responded in 48 hours → send follow-up #3, update to "stage 4"
In "stage 4" and haven't responded in 5 days → send final follow-up, update to "stage 5 - closed"

The Follow-Up Message Sequence

Each follow-up needs a different angle to avoid feeling spammy. Here's the sequence:

Follow-up #1 (2 hours): Gentle check-in. "Hey [Name], just wanted to make sure my last message came through! Happy to answer any questions about [service]. — [Business Name]"
Follow-up #2 (24 hours): Value add. "Hi [Name], here's a quick tip for [their specific situation]: [relevant tip]. Still happy to help whenever you're ready. — [Business]"
Follow-up #3 (48 hours): Social proof. "[Name], we helped a customer in [area] with [similar problem] last week and they saved [benefit]. Whenever you want to chat, I'm here. — [Business]"
Follow-up #4 (5 days): Breaking up. "Hi [Name], I don't want to keep bothering you — this will be my last message. If things change and you need [service], we'd love to help. Take care! — [Business]"

Generate each of these dynamically with OpenAI using the context from the original lead record.

Handling Replies

When a prospect replies to any of your texts, Twilio fires a webhook. Create a fourth workflow that triggers on incoming SMS. The AI reads the reply and decides: is this person interested and ready to book, or do they have a question to answer first?

If they're ready to book: generate a reply that includes the Cal.com booking link. Something like: "Great! Here's a link to pick a time that works for you: [cal.com link]. Looking forward to talking!"

If they have a question: use OpenAI to generate a relevant answer based on a knowledge base about the client's business (a simple Airtable table or Google Sheet with FAQ responses works great here). Then ask if they'd like to book.

Testing Your Automation Before Going Live

Never launch an automation you haven't tested end-to-end. Here's your testing checklist:

Call the Twilio number from a test phone and hang up. Did the text arrive within 60 seconds? Does it read naturally?
Submit a test form with fake data. Did the text arrive? Does it reference the fake data correctly?
Test the follow-up logic by manually setting a lead to "stage 1" with a timestamp 3 hours in the past. Does the hourly workflow pick it up and send follow-up #1?
Reply to one of the test texts with "I'm interested." Does the booking link appear in the response?
Reply with a specific question. Does the AI answer it correctly?
Test error handling: what happens if Twilio fails to send? Set up an error notification so you're alerted immediately.

Client Onboarding: What You Need From Them

Here's the complete list of what you need from your client to build this system:

Their current business phone number (you'll forward calls from the Twilio number to this)
Access to their website backend to add the webhook URL to their contact form
Login to their CRM (if they use one)
A Cal.com account (you can set one up for them in 10 minutes)
Their business hours so the AI knows when to respond differently after hours
3-5 sentences describing their business, services, and typical customer
Any FAQ content: pricing, service area, turnaround times

Once you have a client result, document it and use it in your next sales conversation. See our guide on getting your first AI agency client for how to turn your results into case studies that close deals.

For agencies looking to systematize their delivery process across multiple clients, see our post on how to productize your AI automation service into packages.

Ready to build your AI agency? Join 215+ AI agency owners in our free community. Join the free AI Agency Sprint on Skool.