AI & Automation
Make.com AI Agent Workflow for Local Business
Part of the guide: AI & Marketing Automation
To build an AI lead workflow in Make.com, connect a form to a custom webhook, send the lead details to an AI model that classifies the enquiry, split the result with a router, draft a reply, have a person approve it for high-value leads, send it with an email service and log everything in a spreadsheet.
On the free plan, Make includes 1,000 credits a month, and each lead in this design uses about 7.
This is an automation with an AI step, not a fully autonomous agent. Make also has a separate AI Agents app where agents use instructions, knowledge and tools. Details below were checked on 24 October 2026 against vendor pages that opened. Anything that did not open is marked.
1,000→
Make free plan credits per month
Make pricing page
about 7→
Credits per lead in this design
Assumption, see example
5 MB→
Maximum webhook payload
Make webhook docs
3,000→
Resend free emails per month
Resend pricing page
Estimates for illustration; actual numbers vary.
What Does This Workflow Do?
A visitor fills in your contact form. The form sends the answers to Make, an AI model sorts the lead into good fit, maybe or not a fit, and a router sends each type down its own path. Good fits get a personalised reply for you to approve, maybes get a softer template, and poor fits get a polite decline. Every lead lands in a sheet.
The point is speed and consistency, not removing yourself. For the first weeks, review every message before it goes out. Our overview of AI agents for local business explains where this fits.
What Do You Need Before You Start?
- A Make account. The free plan has 1,000 credits a month, 1 MB of data storage and 1 data store.
- A form tool that can send a webhook, such as Tally, Typeform, Fillout or Gravity Forms. Webhook support varies by plan, so check yours.
- An account with an AI provider and its API key.
- An email sending service with a verified domain.
- A Google Sheet for the log.
Make's pricing page now lists credits as the billing unit. Its help page says one operation uses one credit for non-AI apps and for third-party AI apps such as OpenAI, Claude and Gemini. If you connect your own AI provider, you pay Make credits for operations and pay the provider for tokens.
How Do You Set Up Each Step?
| Step | Make module type | What to do |
|---|---|---|
| 1 | Custom webhook | Create the webhook, copy its URL into your form, send a test and let Make detect the data structure. |
| 2 | HTTP request or your AI provider's app | Send the lead fields with a classification prompt. Ask for one of three labels and a short reason. |
| 3 | Parse JSON | Turn the model's reply into separate fields. |
| 4 | Router | One route per label, plus a fallback route for anything unexpected. |
| 5 | HTTP request or AI app | On the good-fit route, draft a reply from the lead's own words. |
| 6 | Approval step | Save the draft to the sheet and notify yourself before sending. |
| 7 | Email service | Send the approved message. |
| 8 | Google Sheets | Add a row with time, lead details, label, reason and whether a reply was sent. |
Make's docs say each scenario needs its own webhook, and callers can send query string, form data or JSON. You can require an API key sent in the x-make-apikey header and define a data structure so bad requests are rejected with HTTP 400. An accepted request returns HTTP 200 with the body "Accepted."
How Does the Router Behave?
Make's help says a router splits a scenario into routes, each with a filter, and that routes run one after another, not in parallel. A fallback route handles anything that matches no other filter. Use it: if the AI returns an unexpected label, the lead still reaches you instead of vanishing.
Write the classification rules in plain words in your prompt, for example the minimum monthly budget you accept, the business types you serve and what counts as a clear need. Keep the prompt short and test it on ten real past enquiries before going live.
Which AI Model Should You Use?
Pick a model that is cheap, fast and available on your plan. One caution from this check: Groq's deprecations page says llama-3.3-70b-versatile was scheduled for shutdown on 16 August 2026 for free and developer use, with openai/gpt-oss-120b and qwen/qwen3.6-27b as recommended replacements. Many older tutorials still use that model name, so do not copy it blindly.
OpenRouter exposes many models behind one OpenAI-compatible endpoint, with the base URL https://openrouter.ai/api/v1. Its live catalogue on 24 October 2026 listed these per-million-token prices.
| Model on OpenRouter | Input per 1M tokens | Output per 1M tokens |
|---|---|---|
| openai/gpt-5-nano | $0.05 | $0.40 |
| openai/gpt-4o-mini | $0.15 | $0.60 |
| qwen/qwen3.5-9b | $0.17 | $0.25 |
| openai/gpt-5-mini | $0.25 | $2.00 |
Prices change often, so check the catalogue before you commit. See our OpenRouter guide for how to pick one.
How Much Does It Cost? A Worked Example
All inputs are assumptions for a small business that gets 100 enquiries a month. Credit use per lead depends on how many modules run.
| Input | Value (assumption) |
|---|---|
| Modules that run per lead | 7 (webhook, classify, parse, draft, save draft, send, log row) |
| Credits per lead | 7 |
| Leads per month | 100 |
| Credits used | 700 of the free 1,000 |
| AI tokens per lead | 1,500 in and 400 out |
| AI cost per lead on gpt-5-nano | about $0.0002 |
| AI cost per lead on gpt-4o-mini | about $0.0005 |
| AI cost for 100 leads | about $0.02 to $0.05 |
At that volume the free Make plan and the free Resend plan, which allows 3,000 emails a month but only 100 a day, would cover the workflow. If you reach about 140 leads a month you hit the 1,000-credit limit. The official page did not show paid plan prices when opened, so check Make's pricing before you upgrade.
If you spend 10 minutes on each manual reply (assumption), 100 leads take about 17 hours a month. Even with approval, a draft in front of you cuts that sharply, but measure your own time.
How Do You Keep a Human in the Loop?
Do not let an AI send first-contact emails unseen. A simple approval pattern works: Make writes the draft to a sheet column, notifies you, and a second small scenario sends only rows you mark as approved. Make's AI Agents docs also list best practices on tools, prompts and data security, which are worth reading before you give an agent more freedom.
Use an email service with duplicate protection. Resend's API requires from, to and subject, accepts up to 50 recipients in the to field, and supports an optional idempotency key that expires after 24 hours, which helps prevent duplicate sends if a scenario retries.
Do This Today
Write the three rules that define a good-fit lead for your business, then test them by hand on your last ten enquiries before you build anything in Make.
What Mistakes Break This Workflow?
- No fallback route, so unexpected AI output silently stops the scenario.
- No data structure on the webhook, so a changed form field breaks later modules.
- Pasting API keys into prompts or shared notes instead of using connections or stored variables.
- Sending marketing follow-ups with no unsubscribe. The FTC's CAN-SPAM guide asks for accurate sender details, a physical postal address and an opt-out honoured within 10 business days for commercial messages.
- Treating a low-credit warning as noise and losing leads at month end.
For ready-made scenarios in other areas, see 10 Make.com workflows for local business, Make.com for salons and our no-code AI agent guide. For reminders, read the appointment reminder script.
FAQ
Do I need a paid Make plan for this workflow?
Not at low volume. The free plan includes 1,000 credits a month. If each lead uses about 7 credits, that covers roughly 140 leads. Paid prices could not be confirmed on the official page, so check it.
What form tools can trigger the webhook?
Any tool that can send a POST request to the webhook URL, in query string, form data or JSON. Check whether webhooks are on your form plan.
Can I send a follow-up if the lead does not reply?
Yes, with a second scheduled scenario that reads the sheet and checks a replied column. Make's Sleep delay limit was not verified, so avoid long waits inside one scenario.
What if the AI sends a bad email?
Use the approval step described above and review each draft until you trust the prompt.
Is this the same as Make AI Agents?
No. Make AI Agents is a separate app where an agent has instructions, knowledge and tools. The workflow here is a fixed scenario with an AI step.
Sources
- Make: pricing (free plan credits; paid prices were not shown when opened)
- Make Help: credits
- Make: custom webhook module
- Make Help: router
- Make Help: Make AI Agents
- Groq: deprecations
- OpenRouter: quickstart and models API (prices read on 24 October 2026)
- Resend: pricing
- Resend: send email API
- FTC: CAN-SPAM Act compliance guide
- Not used: Make paid plan prices (seen only in third-party blogs via search, not on an official page); Groq free-plan limits (not shown on the page)
Want this built and tested for your business? See our AI agents service or book a free audit.
Free for local businesses
Want this applied to your business?
I'll review your Google presence, local SEO, and ad accounts — and send you a specific action plan within 48 hours. No pitch, no pressure.
Want hands-on help?
See how DataLatte handles AI Agents & Automation for local businesses.
Nataliia
Local marketing strategist with 10+ years at global agencies — OMD, Dentsu, GroupM, and BBDO. Now helping small businesses get the same data-driven edge. Based in Europe, working with clients in the US, UK, Australia, and beyond.
About NataliiaRelated articles
AI & Automation
n8n AI Agent Workflow for Local Business Lead Follow-Up
10 min read
AI & Automation
Voice AI Agents for Local Business: Handle Phone Calls Without Staff
14 min read
AI & Automation
n8n for Local Business: 10 Automations That Save You 10 Hours a Week
15 min read
AI & Automation
Make.com Automation for Local Businesses: 10 Workflows That Save 5hr/Week
12 min readWant this applied to your business?
Let's review your current marketing setup together — free, no obligations.
Get Your Free Marketing Audit