Send a One-Page Proposal Minutes After Every Sales Call

Start in Granola. Pick (or create) a folder you can drop meeting notes and transcripts into after a call ends. In Step 2, you'll point Zapier's When a note is shared to Zapier trigger at this same folder, so sharing a note there becomes the intentional "go" signal for the whole workflow. Nothing fires until you decide a call is worth following up on.

Next, build a one-page proposal template in [your proposal template] tool of choice. Google Docs is what we used, since Zapier can generate a document straight from it. Add a double-bracket placeholder for every value you want filled in automatically: {{prospect_name}}, {{company_name}}, {{key_needs}}, {{budget_range}}, {{next_steps}}, and {{decision_timeline}}.

Pro tip: Those double brackets aren't decoration: that's how automation platforms like Zapier and Make.com know where to insert a variable. Anytime you want the Zap to drop a value into a document, wrap the field name in double brackets and the platform will find it.

If you don't already have a proposal template, ask your AI assistant to draft one using the field names you plan to extract. If you do have one, just swap in the double-bracket placeholders where the old static text used to be.

Open a new Zap and use Zapier's Copilot to build it conversationally instead of wiring up each step by hand. Tell it what you actually want, in plain language:

When I get off a sales call, I want the Granola notes to come here. Then AI
will analyze it, pull out the important info, log it in Google Sheets, and
use that to build a custom proposal based on an existing template. It'd be
great if the template were a Google Doc, but it doesn't have to be. The
end goal is a PDF emailed to the prospect from my email.

Copilot turned that into a six-step chain: Granola (When a note is shared to Zapier) → AI by Zapier (extract the fields) → Google Sheets (log a row) → Google Docs (create a document from the template) → Google Drive (export the doc as a PDF) → Gmail (create a draft with the PDF attached).

Copilot then asked follow-up questions to wire the details together: what the spreadsheet was called, and how to name each generated file. We told it:

Google Sheets: Match the AI-extracted fields to the fields they should go
into, and put a date and run time into that first column.

Google Docs: Do the same thing with the property matching, and also
generate a document name based on what we extracted. It should follow
this format: [Client Name] - [Company Proposal] [Clean Date].

The first time you run this, Zapier will ask you to authenticate every connected app: Granola, Google Sheets, Google Docs, Google Drive, and Gmail. You only have to do this once. If Zapier finds a Google Doc or sheet you already created for this workflow, it can connect to it automatically; otherwise, you'll pick your Drive and the right file manually in each step.

Pro tip: If you're wary of connecting your real Gmail account, you don't have to grant full access. During the Gmail authentication step, you can toggle off the permission to send or read email and only allow the scopes needed to create drafts and look up contact addresses.

Open the AI by Zapier step and look at the prompt Copilot wrote for the extraction:

Extract key information from this sales call to populate a proposal.
Return the following fields:

1. prospect_name: Name of the person/decision maker
2. company_name: Company or organization
3. key_needs: Top 1-3 pain points or needs identified
4. budget_range: Budget mentioned (if any), or note if not discussed
5. next_steps: Agreed-upon next actions
6. decision_timeline: When they need a decision/proposal
7. call_summary: 2-3 sentence summary of the call

Provide clear, concise, proposal-ready language.

This is the part worth getting right. Pulling every field explicitly, instead of asking the AI to "summarize the call," is what made the rest of the Zap reliable. Feed it the note, the call transcript, and the attendee list as input, so it has everything it needs to fill each field.

Check the output against your own call before you trust it. We caught two gaps in testing:

• The first pass didn't pull the attendee's email address, which you need if the Zap is going to email the proposal to anyone. Add it as its own field.
• call_summary is useful for your log, but it doesn't belong on a client-facing proposal. Either drop that field from the document template, or make sure it only maps into the Google Sheets row and not the Google Doc.

What You Need To Get Started already covers the accounts; from here, it's about verifying the fields, not the connections.

Run the test built into each step. Zapier will step through Granola, extraction, the sheet row, the document, the PDF export, and the Gmail draft, and show you the output of each one as it goes.

Watch the extraction output closely. This is where things go wrong first. In one test run, the AI returned a real budget figure but tacked its own editorial judgment onto it in parentheses, and that aside carried straight into the generated proposal. A real client would see the AI's opinion about their budget sitting right next to the number.

Pro tip: If the extraction step keeps producing judgment calls instead of clean data, swap AI by Zapier out for Claude, ChatGPT, or Gemini in that step. Their structured-output modules tend to hold to a schema more consistently, which matters most for a field like budget that should only ever contain a number or "not discussed."

We also triggered the Zap with a note from a meeting that wasn't a sales call. It fired anyway, the extraction came back incomplete, and the Zap stalled instead of producing a usable proposal. A workflow like this needs a way to recognize when it's been handed the wrong kind of meeting. See Going Further below.

However you generate the fields, keep the last step set to Create Draft, never Send Email. Before you send anything yourself, check:

• Recipient: is the email address actually the prospect's, not a placeholder or a teammate's?
• Scope and needs: does key_needs reflect what they actually said, not a guess?
• Timeline: is decision_timeline real, or did the AI invent a date?
• Price: is budget_range a clean number you'd put in front of a client, with no editorial aside attached?

That review is the whole point of drafting instead of sending. The Zap gets you from call to reviewable proposal in minutes; you're still the one who decides it's ready to go.