How we turn meeting notes into first-draft proposals

The problem

I hate writing proposals!

Every good sales call leaves three jobs behind it: write up the notes, send a follow-up, and write the proposal.

The proposal was the slow one. It meant writing it from scratch, then laying it out by hand in a design tool so it matched our proposal design. A single proposal could take half a day. And the longer it waited, the more of the detail from the call had faded.

What we built

Two connected steps: one turns the call into notes, the other turns the notes into a first-draft proposal that’s ready for me to edit and fine-tune by hand.

Step 1: from call to notes

1. The call is recorded and transcribed with Otter, and the transcript is saved to Dropbox.
2. An n8n workflow watches that folder. When a new transcript arrives, it sends it to Claude.
3. Claude writes three things: the meeting notes, a draft of the follow-up email, and a tidied-up copy of the transcript.

Step 2: from notes to proposal

1. The notes are saved as structured data (JSON), either to a folder or sent to our own web address that accepts them.
2. A simple web portal shows the notes in a readable form, so we can correct anything the AI got wrong or add what wasn’t said on the call.
3. AI then researches and writes the first draft of the proposal. It compares what the client needs with the notes from the call, the prospect’s website and our similar past projects. It knows how we work, the services we offer and how we structure our pricing.
4. The draft comes out laid out in our own proposal design, not as plain text that still needs formatting.
5. We edit it, and only then does it go to the client.

Why the notes are structured data

Plain notes are fine for reading but awkward for a computer to build on. Saving them as JSON means each part of the conversation (what the client wants, what’s in the way, deadlines, actions) sits in its own labelled field. The proposal step can then put each one in the right place, and the same notes could feed other things later, like a CRM.

What the AI doesn’t do

It doesn’t send anything. Every set of notes can be checked in the portal, and every proposal is read and edited by us before it goes out, including the pricing. The AI gets us to a strong first draft quickly; the judgement about what to recommend stays with us.

What it saves

We estimate it cuts the time it takes to write a proposal by about 75%. A proposal that used to take half a day now takes around an hour, most of it spent editing and fine-tuning. It’s an estimate, not a measurement.

The notes and follow-up are also ready soon after the call, while the detail is still fresh, and nobody spends time rebuilding a layout by hand.

Could this work for you?

If you sell through calls and then send a quote, proposal or scope of work, yes. Trades and builders, agencies, consultants and B2B suppliers all have the same three jobs after a call.

A version for your business would use your services, your pricing rules and your own template, whether that’s a Word document, a Google Doc, a PDF or a web page. It works with recordings from Zoom, Microsoft Teams or Google Meet as well as Otter.