Every service business has the same invisible tax: the 15-minute back-and-forth to book a new client.
Phone tag. "What are your rates?" DMs. Intake forms sent by email that bounce back half-filled. If you run a law firm, dental clinic, consulting practice, or agency, you know this dance. It happens 5-20 times a week. It's never urgent enough to fix — until you multiply it out and realize you're losing 10+ hours a month just onboarding strangers.
This post walks you through eliminating that entirely using AI. No code required. Should take you under 30 minutes to set up.
What "AI customer intake" actually means
Your intake process has two jobs:
- Qualify — Is this person a good fit? Can they afford you? Do they have the problem you solve?
- Capture — Get their name, contact info, situation, and book a slot.
Right now a human (probably you) does both. AI can handle both — 24/7, instantly, without burning your attention.
The stack (all free tiers available)
- Tally.so — smart intake form with conditional logic
- Make.com (formerly Integromat) — automation backbone
- Claude or GPT-4o — AI that reads submissions and drafts personalized replies
- Calendly — scheduling
- Gmail or your email provider — delivery
You can swap most of these. The pattern is what matters.
Step 1: Build a smart intake form in Tally (10 min)
Go to tally.so and create a new form. The key is conditional logic — don't show irrelevant questions.
Example structure for a service business:
Q1: What brings you here? (dropdown)
- I need [your core service]
- I have a specific project in mind
- Just exploring
Q2 (if "specific project"): Describe it briefly
Q3: What's your timeline?
Q4: What's your budget range? (ranges, not exact)
Q5: Best email + phone
This does two things: it pre-qualifies (someone with "no timeline, no budget" is a tire-kicker), and it gives the AI something to work with.
Embed this form on your website's Contact or Services page instead of a generic contact form.
Step 2: Route submissions to Make.com (5 min)
In Make.com, create a new scenario:
- Trigger: Tally > Watch Responses
- Connect your Tally form
- Add a filter — only proceed if budget field is above your minimum
This is your first automated disqualification. You never even see tire-kickers.
Step 3: Let AI draft the reply (10 min)
Add an HTTP module in Make.com pointing to OpenAI or Anthropic's API:
POST https://api.anthropic.com/v1/messages
Headers: x-api-key: YOUR_KEY
Body:
{
"model": "claude-opus-4-7",
"max_tokens": 400,
"messages": [{
"role": "user",
"content": "You are [Your Name], a [your profession]. A potential client just submitted this intake form: [TALLY_SUBMISSION]. Write a warm, specific reply (3-4 sentences) that: (1) acknowledges their specific situation, (2) confirms you can help, (3) invites them to book a 20-min call via this link: [CALENDLY_LINK]. Sign as [Your Name]."
}]
}
Map the Tally fields into the prompt. The AI reads their actual words and writes a reply that sounds like YOU wrote it — because you told it to.
Step 4: Send the email (2 min)
Add a Gmail (or SMTP) module:
-
To:
{{tally.email}} -
Subject:
Re: Your inquiry — [Your Business Name] -
Body:
{{ai_response}}
That's it. New submission → AI reads it → personalized reply sent → Calendly link included. All within 60 seconds of them hitting Submit.
What this looks like in practice
A potential client fills out your form at 11 PM on a Sunday. By 11:01 PM they have a reply in their inbox that references their specific project, confirms you work with clients like them, and includes a booking link.
Your competitor — also a solo operator — responds Monday morning. You already have the call booked.
Speed-to-lead is one of the most underrated conversion levers in service businesses. A study by HBR found that responding within an hour makes you 7x more likely to qualify a lead vs. responding an hour later. AI makes this trivial.
Common objections
"What if the AI says something wrong?"
You're not letting AI make promises — just write a warm acknowledgment + book a call. The real conversation happens live with you.
"It feels impersonal."
Test it. Send a test submission yourself. The output is more specific to the person than the generic "Thanks for reaching out!" 99% of businesses send.
"I don't know how to call APIs in Make."
Make has a visual interface. The HTTP module is drag-and-drop. You paste the JSON above and map fields by clicking.
Cost
| Tool | Free tier covers |
|---|---|
| Tally | Unlimited forms, up to 100 responses/mo |
| Make | 1,000 operations/mo |
| Claude/GPT | ~$0.01-0.03 per intake processed |
| Calendly | 1 event type free |
For most service businesses with <50 leads/month, this runs essentially $0/month.
One more thing: the follow-up
Add a second path in Make: if someone fills the form but doesn't book within 48 hours, trigger a follow-up email. Same AI prompt, different angle:
"You are [Your Name]. This person submitted an intake 48 hours ago but hasn't booked yet: [SUBMISSION]. Write a 2-sentence follow-up checking if they have questions. Keep it casual."
Most service businesses never follow up after the initial reply. This alone can recover 20-30% of stalled leads.
The result
You go from: phone tag → eventually book call → maybe convert
To: submit form → instant personalized reply → self-book call → show up qualified
Your intake is now an asset that runs while you sleep. You spend zero time on the pre-qualification dance. Every lead gets an immediate, specific response.
If you want a done-for-you version of this (we build the automation, write the prompts, integrate your calendar), we set this up for service businesses at MidasTools. Takes us about a day.
Otherwise — the tools are all free to start. Go build it.
Have you automated your intake? Drop what worked (or didn't) in the comments.
Top comments (0)