How I Auto Generate Proposals with ChatGPT and Make
The Late-Night Pitching Trap
I spent four hours on a Thursday night in my Seoul office manually typing out a sponsorship proposal for a tech brand wanting to partner across two of my content channels. I had to pull stats, outline deliverables, estimate timelines, and format a clean PDF. By the time I hit send, my eyes were burning, and the client took three days to respond with a simple request to adjust the scope.
That was the moment I decided to build a workflow to auto generate proposals with ChatGPT and Make. Pitching clients, sponsors, or agency partners is essential when running solo operations, but manual proposal writing is a massive drain on deep work. Customizing every single proposal from scratch doesn’t scale when you are a team of one.
After testing several approaches, I built an automated pipeline: an inbound intake form sends details straight into a Make scenario, which passes the context to ChatGPT to draft customized scope, deliverables, and terms. The draft is then injected into a clean document template ready for my final approval. Here is exact setup I use to save hours on client onboarding without sacrificing personalization.
The Workflow Architecture
Before jumping into the step-by-step setup, it helps to understand how the data flows through this automation pipeline. You do not want a fully uncontrolled system that emails clients automatically without human eyes checking it first. Instead, this system builds a full draft and drops it into your working folder.
| Component | Tool Used | Primary Function |
|---|---|---|
| Intake Form | Tally / Typeform | Collects client budget, goals, timeline, and scope notes |
| Automation Hub | Make (formerly Integromat) | Routes data between intake form, OpenAI API, and Google Docs |
| AI Generator | ChatGPT (OpenAI API) | Drafts proposal sections tailored to client inputs |
| Doc Template | Google Docs | Formats generated text into a polished document for review |
This four-tier stack ensures you maintain control over the output while offloading 90 percent of the drafting heavy lifting to AI.
Step 1: Build the Intake Form
Your proposal is only as good as the input data you collect. A vague intake form forces ChatGPT to guess, leading to generic proposals that read like boilerplates. I use a simple form tool like Tally or Typeform, but Google Forms works equally well.
Make sure your intake form captures these specific data points:
- Client Information: Company name, contact person, website, and industry.
- Project Overview: The core problem they want solved or campaign goal.
- Requested Deliverables: Checkboxes for specific services (e.g., dedicated video sponsorship, long-form blog review, SEO audit, monthly maintenance).
- Budget Range: Dropdown ranges so you know how to anchor your package options.
- Target Timeline: When they need the work completed.
- Custom Notes: A free-text field for any special requests discussed during discovery calls.
Once your form is published, submit a test response with realistic client data. This test submission provides the payload you need to map variables inside Make.
Step 2: Create Your Google Docs Template
Rather than having ChatGPT generate raw unformatted text or complex code, let Google Docs handle the styling. Create a template document in Google Drive that reflects your brand identity.
In your template, place variable placeholders using double curly braces where dynamic text should go. For instance:
{{ClientName}}{{ProjectDate}}{{ExecutiveSummary}}{{ScopeOfWork}}{{DeliverablesTable}}{{PricingSection}}{{NextSteps}}
Style your document template with your preferred typography, brand colors, header image, and static legal terms upfront. When Make runs the scenario, it creates a copy of this template and swaps those curly brackets with the response from ChatGPT.
Step 3: Set Up the OpenAI Prompt Schema in Make
Log in to Make and create a new scenario. Add your form tool as the trigger module (for example, Tally’s “Watch Responses” module). Once connected to your form, add the OpenAI module set to Create a Chat Completion.
To prevent ChatGPT from hallucinating inaccurate promises or wildly wrong prices, split your prompt instructions carefully between the System Role and User Role.
The System Prompt
In the System field, define the persona and establish clear operational boundaries:
You are a professional business manager drafting a client proposal. Your tone is direct, modern, and confident. Use clean markdown structure with bullet points. Never make up specific pricing unless given explicit rules. Follow the structural requirements strictly.
The User Prompt
In the User message field, pass the variables coming from your intake form alongside strict structural guidelines:
Draft a tailored project proposal using the following details:
- Client Name: {{1.ClientName}}
- Company: {{1.CompanyName}}
- Goal: {{1.ProjectGoal}}
- Selected Services: {{1.Deliverables}}
- Budget Range: {{1.Budget}}
- Notes: {{1.Notes}}
Return the response formatted into three clean sections:
1. Executive Summary: Explain their problem and how our services solve it.
2. Scope of Work: Break down detailed steps and milestones based on their selected deliverables.
3. Recommended Package: Outline a primary tier matching their budget, plus a higher-value add-on option.
Using GPT-4o or GPT-4o-mini via the API gives you fast, high-quality output while keeping token costs to fractions of a cent per proposal run.
Step 4: Map ChatGPT Output into Google Docs
Add the Google Docs: Create a Document from a Template module after your OpenAI module in the Make scenario line.
- Select your document template from Google Drive.
- Set the Destination Folder where new proposal drafts should be saved.
- Give the new file a dynamic title, like
Proposal - {{1.CompanyName}} - {{now}}. - Map the ChatGPT output fields to the template variables you defined in Step 2.
If you want to keep your pipeline completely clean, add a final notification module—like a Slack message or an email via Gmail—to alert you whenever a new proposal document is drafted, complete with a direct link to the editable Google Doc.
Handling Edge Cases and Limitations
When I first set up this automation across my media properties, I made a key mistake: I trusted the AI to format tables and price math natively. AI models are natural language predictors, not accounting software. If you leave pricing structures fully open to ChatGPT, it will eventually output confusing package numbers or miscalculate deliverables.
Here are three crucial operational guardrails I learned the hard way:
- Keep Static Pricing Fixed: Do not rely on ChatGPT to calculate total costs dynamically. Use fixed package pricing inside your Google Doc template, or pass pre-calculated fee numbers from your form logic directly into the document. Use ChatGPT only to customize the text descriptions, scope context, and value justification.
- Watch the Formatting: Complex markdown syntax output by ChatGPT sometimes pastes raw formatting characters into Google Docs template variables. To keep things looking clean, instruct the model to avoid heavy nested formatting inside string outputs.
- Always Keep a Human in the Loop: Never configure a Make scenario to auto-email a generated proposal directly to a client. Always review the output draft first to refine the tone, check edge-case scope details, and add personal touches.
My Take on Proposal Automation
Automating proposal drafts using ChatGPT and Make strikes a solid balance between operational efficiency and custom communication. It takes a tedious 90-minute writing process and reduces it to a 5-minute editing task.
However, automation should never replace your strategic judgment. The goal of using AI here isn’t to disengage from your prospective clients—it is to eliminate repetitive formatting and blank-page friction. The time you save on manual drafting should be reinvested into talking to clients, refining your service offerings, and delivering great work.
Regarding costs, this setup remains very accessible for solo operators. Make offers a functional free plan for basic operations, with paid tiers starting around $9 to $15 per month depending on usage. The OpenAI API works on a pay-as-you-go model where individual proposal generations typically cost under ten cents. Always check the official pricing pages for Make and OpenAI, as tiers and operation limits update frequently.
Frequently Asked Questions
Can I auto generate proposals with ChatGPT using Make without coding knowledge?
Yes. Make uses a visual, drag-and-drop workflow editor. You do not need to write code to connect your form, OpenAI, and Google Docs. Mapping variables between steps is done entirely by selecting input tags inside the user interface.
How do I keep ChatGPT from making up inaccurate prices or services?
The best way to prevent hallucinated details is to handle pricing through fixed variables in your form or doc template rather than letting the AI calculate numbers. Use system prompts to explicitly tell ChatGPT to stick strictly to the services specified in the user context without adding extra deliverables.
Should I auto-send the generated proposal document directly to clients?
No, auto-sending generated proposals directly to prospective clients is risky. AI models can misunderstand nuanced client notes or generate awkward phrasing. Set up your workflow so Make creates a draft document and notifies you, giving you a chance to review and polish the proposal before sending it out.
