Using AI to build a simple client intake form and auto-generated summary so you stop spending the first meeting asking questions you already asked

Small business owners lose 15–30 minutes per onboarding call re-asking questions clients already answered in writing, according to data from HoneyBook{:target="_blank"} and Dubsado{:target="_blank"}. This post walks you through a three-tool workflow — Tally, Zapier, and OpenAI — that converts a completed intake form into a structured meeting brief before you open your calendar. At under $0.01 per summary in AI processing costs, the math on setup time pays back after the first two clients.

What You Need to Automate Your Client Intake Form Summary with AI

Tally{:target="_blank"} — a free form builder that sends clean, labeled field data via webhook, which is what makes the AI prompt reliable downstream. The free tier covers everything in this workflow; you don't need a paid plan unless you want custom domains or removed Tally branding. Tally pricing{:target="_blank"}.

Zapier{:target="_blank"} — the automation layer connecting your form to OpenAI. Here's the catch: the free plan only supports single-step Zaps. This workflow requires at least three steps (receive form data, send to AI, deliver summary), so you need a paid plan. Zapier starts at $19.99/month{:target="_blank"} as of early 2026 — check their site, these change.

OpenAI{:target="_blank"} — specifically GPT-4o mini via Zapier's native OpenAI integration. You'll need an OpenAI account with API access. The per-summary cost runs under $0.01 for a typical 500–800 word intake form (GPT-4o mini is priced at roughly $0.15 per 1M input tokens as of early 2026).

Time required: 45–90 minutes total. Budget 15 minutes for the Tally form, 20–40 minutes for the Zapier workflow, and 10–15 minutes to test and tune your AI prompt.

Skill level: No coding required. You need a Zapier account and basic comfort clicking through a multi-step workflow builder. If you've never used Zapier before, add 20 minutes.

Step 1: Build Your Intake Form in Tally (Ask Questions That Actually Produce Useful Data)

The quality of your AI summary is determined almost entirely by your form structure. Vague questions produce vague summaries. Specific, labeled fields give the AI unambiguous input to work with.

1. Open Tally{:target="_blank"} and create a new form.
2. Add the following fields — these labels matter because Zapier passes them to OpenAI exactly as written:
   • Company name (short text)
   • Contact name and role (short text)
   • Industry / type of business (short text)
   • Primary challenge you want to solve (long text)
   • Desired outcome from working together (long text)
   • Timeline (multiple choice: Under 1 month / 1–3 months / 3–6 months / Flexible)
   • Budget range (multiple choice: Under $1K / $1K–$5K / $5K–$15K / $15K+)
   • How did you hear about us? (short text)
3. Publish the form and copy the public URL — you'll share this in booking confirmation emails or on your website.
4. Go to Integrations → Webhooks in your Tally form settings and enable the webhook. Copy the webhook URL — you'll need it in Zapier.

The trade-off between open-text fields and structured dropdowns is real: open-text lets clients elaborate, but dropdowns (for budget and timeline especially) produce clean, consistent values that the AI doesn't have to interpret. Use dropdowns wherever you want reliable extraction.

Step 2: Connect Tally to OpenAI via Zapier to Generate an Automatic Client Brief

1. Log into Zapier{:target="_blank"} and click Create Zap.
2. Set the trigger app to Webhooks by Zapier, trigger event Catch Hook. Copy the Zapier webhook URL it generates.
3. Paste that Zapier webhook URL into your Tally webhook settings (replacing the placeholder from Step 1, Step 4). Submit a test form entry in Tally so Zapier has sample data to map.
4. In Zapier, click Test Trigger — you should see all your labeled form fields appear with the test data you just submitted. If fields are missing, recheck the Tally webhook configuration.
5. Add a second step in Zapier: search for OpenAI, action event Send Prompt (or "Chat with GPT" depending on the current Zapier interface). Connect your OpenAI account — Zapier's native integration as of early 2026 supports GPT-4o and GPT-4o mini without requiring you to manually paste API keys in most cases.
6. Set the model to GPT-4o mini. This handles summarization tasks cleanly and keeps your per-run cost under $0.01. GPT-4o is not meaningfully better for this task and costs roughly 20x more per token.
7. In the Prompt field, paste your prompt (see Step 3 below), mapping Tally's labeled fields into the appropriate placeholders.

Step 3: Write the AI Prompt That Turns Form Answers into a Meeting Brief

This is the highest-leverage step in the entire workflow. A prompt that just says "summarize this intake form" produces a paragraph you could have written yourself. A prompt that specifies output sections produces a structured brief you can open 90 seconds before a call.

System prompt: You are a professional client intake analyst. Your job is to convert raw intake form responses into a concise, structured meeting brief for the business owner. Be specific. Do not invent information not present in the form. Flag anything that is vague or missing.

User prompt: A new client has submitted an intake form. Here are their responses:

• Company name: [Tally field: Company name]
• Contact name and role: [Tally field: Contact name and role]
• Industry: [Tally field: Industry / type of business]
• Primary challenge: [Tally field: Primary challenge you want to solve]
• Desired outcome: [Tally field: Desired outcome from working together]
• Timeline: [Tally field: Timeline]
• Budget range: [Tally field: Budget range]
• Referral source: [Tally field: How did you hear about us?]

Write a meeting brief with the following sections: Client Background (2–3 sentences on who they are and their context) Problem Statement (what they say they need to solve) Goals and Desired Outcomes (what success looks like to them) Constraints (timeline and budget, flagged clearly) Suggested Questions (3–4 follow-up questions to ask in the meeting, based on gaps in the intake)

Keep the total length to 150–250 words. Use plain language.

After pasting this into Zapier's prompt field and running a test, you should receive a five-section brief in the Zapier output data. If the output is a raw paragraph without section headers, check that you included the formatting instructions in the user prompt. If GPT-4o mini is flagging fields as missing, your Tally-to-Zapier field mapping likely has a mismatch — go back to Step 2, Step 4 and re-test the trigger with a fresh form submission.

For businesses with a consistent client type — a bookkeeper who only works with product-based businesses, for example — add one sentence to the system prompt: "This business owner works exclusively with [industry type]. Use that context to make the Suggested Questions more specific." The numbers say this produces noticeably more targeted output without adding any complexity to setup.

Step 4: Deliver the Summary Where You'll Actually Use It

Add a third step in Zapier to route the AI output somewhere you'll actually look before a meeting.

1. Email to yourself: Add a Zapier step using Gmail or Email by Zapier, with the subject line set to "New Client Brief — [Company name]" and the body mapped to the OpenAI output field. Fast to set up, zero ongoing maintenance.
2. Google Docs: Add a step using Google Docs, action Create Document from Template or Append to Document. Name the doc with the client name and date. This creates a searchable archive — better than email for any business managing more than 5–6 active clients.
3. Notion: If your team already uses Notion{:target="_blank"}, Zapier's Notion integration can create a new page in a client database with the brief pre-populated. Requires Notion's free tier and a 1–2 minute database setup in Notion.

The honest answer is that email is fine for solo operators. Google Docs or Notion become worth the extra five minutes of setup once you have a team member who also needs the brief before the meeting.

When Something Goes Wrong

Symptom: The AI summary is generic — phrases like "the client wants to improve their business" with no specifics. Root cause: One or more of your intake questions is too open-ended (e.g., "Tell us about your goals"), and the client's answer was correspondingly vague. Fix: Replace open-text questions for goals and challenges with structured prompts: "Describe the single most important outcome you need from this engagement" and add a second field: "What have you already tried?" Specific questions produce specific answers, which produce specific summaries.

Symptom: Zapier's OpenAI step fails with a "no data" or "undefined field" error. Root cause: Your Tally webhook didn't pass the field data cleanly, usually because the test submission happened before the webhook was properly saved in Tally. Fix: Go to Tally → Integrations → Webhooks, confirm the webhook URL matches your Zapier trigger URL exactly, then submit a fresh test form entry and re-run the Zapier trigger test. Do not use Zapier's "skip test" option — the field mapping won't work without real sample data.

Symptom: The summary is accurate but shows up 20–30 minutes after form submission rather than immediately. Root cause: Zapier's polling interval on lower-tier plans can introduce delays; webhook-triggered Zaps should fire in under 2 minutes but occasionally lag during high-traffic periods. Fix: Confirm your Zap is using a Webhook trigger (instant), not a polling trigger. If you set up the Tally connection through the standard Zapier app directory rather than the webhook method, delete and rebuild using Webhooks by Zapier as the trigger.

What to Do Next

Once the workflow is running and producing clean briefs, the natural next optimization is adding a lightweight CRM as the summary destination. This workflow does not replace a CRM — it doesn't track follow-ups, pipeline stages, or client history. For that, layer these briefs into HubSpot's free CRM tier{:target="_blank"} using Zapier's HubSpot integration, creating a new contact and attaching the brief as a note. That combination covers both the pre-meeting prep and the longer-term relationship record.

One more thing: if your intake form collects sensitive data — anything medical, legal, or financial — review OpenAI's API data usage policy{:target="_blank"} before going live. Client responses routed through Zapier to the OpenAI API are processed by OpenAI. Adding a one-line consent notice to the form footer ("Your responses may be processed by AI tools to prepare for our meeting") is a reasonable precaution.

For businesses looking to go further with AI-assisted client work, see how to use AI to write client-facing project summaries and status updates, and automating follow-up emails after client meetings with ChatGPT.

FAQ

How much does this workflow actually cost per month? The AI processing cost is negligible — under $0.01 per summary at GPT-4o mini's pricing of ~$0.15 per 1M input tokens (pricing checked January 2026). The real cost is Zapier's paid plan, which starts at $19.99/month. If you're running 20 new client intakes per month, your cost per automated brief is roughly $1.00 in platform fees plus fractions of a cent in AI processing. That's the math against 15–30 minutes of your time per intake.

Can I use Google Forms instead of Tally? Technically yes, but the trade-off is reliability. Google Forms sends responses as a flat string rather than clean labeled fields, which makes your AI prompt harder to write and more likely to produce inconsistent output. Tally's Zapier integration passes each field with its label intact, which is what makes the prompt predictable. If you're already deep in Google Workspace and prefer to stay there, use JotForm{:target="_blank"} as a middle ground — it offers labeled field output and a Zapier integration with a more structured data format than Google Forms.

What if I don't want to pay for Zapier? The no-code alternative is Make (formerly Integromat){:target="_blank"}, which offers more generous multi-step automation on its free tier and supports both Tally and OpenAI. The trade-off is a steeper learning curve — Make's visual canvas interface is more powerful but less intuitive than Zapier for first-time users. Budget an extra 30–45 minutes for setup if you go that route.

Does this workflow work for service businesses that don't do formal discovery calls? Yes, with a modified form. If your process is closer to "client books a service → you deliver it" rather than a consultative sales cycle, the brief becomes a service brief rather than a meeting brief. Replace "Desired outcome from working together" with "Specific deliverable requested" and remove the Suggested Questions section from the prompt. The workflow is identical; only the form fields and prompt sections change.

What if clients submit the form but the answers are too short to be useful? This is a form design problem, not an AI problem. Add minimum character counts to long-text fields in Tally (available in field settings under "Validation"). Setting a 50-character minimum on the "Primary challenge" field forces clients to write more than "need help with marketing." You can also add instructional placeholder text inside the field: "Example: We're losing clients after the first month and don't know why." That framing produces dramatically more specific responses than a blank field.