From Discovery Call to Signed Proposal in 30 Minutes (Free AI Kit)

Most freelancers and consultants lose deals in the gap between a great discovery call and the proposal. The call ends on a high. Then the proposal takes three days because you have to reread your notes, remember what the client actually said about budget, and rebuild a scope from scratch. By the time it lands, the buyer's excitement has cooled and a competitor has replied.

This kit closes that gap. It chains four free tools into one workflow, so you finish the call, spend about 30 minutes, and send a clear, priced proposal before the client has left their desk. Everything below is free to start. The last step mentions our paid Client Work Pack, but you can do the whole thing without it.

What you need: Claude (free or paid) or ChatGPT, a meeting recorder, and a rough idea of your rates.

The steps, times, and directory pages are in the kit card above.

Step 1: Prep the call (5 minutes)

Good proposals start with good questions. If you walk in without knowing the client's budget range, decision process, and deadline, no AI can write a proposal that wins.

Anthropic's free Call Prep skill produces a short pre-call brief: who is attending, what you know about the company, prior conversations, and suggested discovery questions. It's part of Anthropic's open-source Knowledge Work Plugins (Apache 2.0), so you can install the Sales plugin in Claude or copy the skill folder into your own setup.

No Claude plugins? Paste this into any chat instead:

I have a discovery call with [name, role] at [company] about [project]. Here is their website and LinkedIn summary: [paste]. Give me a one-page brief: what they likely care about, three risks, and eight discovery questions that uncover budget, timeline, decision-makers, and what "success" means to them.

Print the questions or keep them on screen. The aim is to leave the call knowing four things: the problem, the outcome, the budget range, and who signs.

Step 2: Capture the call (no extra time)

Don't take notes by hand. You'll miss what matters and spend the call typing. Granola records your computer's audio and transcribes it without adding a bot to the meeting, which matters with clients who dislike "notetaker has joined" pop-ups. Zoom, Google Meet, and Teams have built-in transcripts that also work.

Two habits make the rest of the kit work better:

• Say the numbers out loud. "So we're looking at roughly $8 to 12k, starting in March, and you and Dana sign off?" Getting it into the transcript means the AI picks it up.
• Close with a recap. Spend the last two minutes summarizing what you heard and what happens next. That recap becomes the spine of your proposal.

Always tell the client you're recording, and follow your local consent laws.

Step 3: Summarize into decisions and needs (5 minutes)

A raw transcript is too noisy to write from. Run it through the Summarize Meeting pattern from Daniel Miessler's open-source Fabric project (MIT license). It's a system prompt that pulls out the summary, key decisions, action items, and open questions.

Use it one of two ways:

1. Copy the pattern's system.md into a Claude Project or custom GPT as its instructions, then paste each transcript.
2. If you use the command line, install Fabric and run fabric -p summarize_meeting < transcript.txt.

Then add one follow-up of your own:

From this summary, list: (1) the client's problem in their words, (2) the outcome they want, (3) budget and timeline signals, (4) anything they said they do NOT want, (5) open questions I still need answered.

Item 4 is the one most freelancers skip, and it's where scope creep starts. Write the exclusions down now.

Step 4: Draft the proposal (15 minutes)

This is the step that used to take days. Anthropic's Proposal Builder skill, part of Claude for Small Business, turns whatever came out of a discovery conversation (a transcript, a voice memo, notes, or an RFP) into a branded, costed proposal or estimate.

Give it three inputs:

• the summary and the five-point list from Step 3,
• your rate card or two past proposals you liked,
• your standard terms (payment schedule, revisions, what happens if the scope changes).

Then ask for a structure that buyers can approve quickly:

1. The problem, in their words. Quote the client back to themselves.
2. The outcome. What's different when you're done.
3. Scope. Deliverables as a numbered list.
4. Out of scope. Pulled from Step 3, item 4.
5. Timeline. Milestones, not a wall of dates.
6. Investment. Two or three options (good, better, best) work well.
7. Next step. One clear action, like "Reply 'approved' and I'll send the agreement."

Read every line before you send it. The AI drafts and you decide. Check the prices against your real costs and remove anything you can't deliver.

Using ChatGPT instead? Paste the same inputs and the seven-part outline into a chat, or save them as instructions in a custom GPT so you never retype them.

Want it pre-built? Our Client Work Pack includes a proposal and SOW generator in three formats (a Claude skill, a custom GPT setup, and a copy-paste prompt), plus onboarding and scope-creep reply templates for after the deal is signed.

Step 5: Send it with a short cover email (5 minutes)

The proposal does the persuading, and the email should just get it opened. Use Fabric's Create Formal Email pattern with notes like:

Thank them for the call, reference the one thing they cared about most, attach the proposal, offer two times for a 15-minute walkthrough, and ask for a decision by Friday.

Keep it under 120 words. Then put a reminder in your calendar for two business days out.

Why this kit works

• Speed wins deals. A same-day proposal signals that you'll deliver the work the same way.
• The client's words go back into the proposal. Buyers sign documents that sound like what they told you.
• Exclusions are on paper from day one. That heads off most scope fights before they start.
• Every item is free to start. The skills and prompts are open source (Apache 2.0 and MIT), and Granola has a free tier.

Common mistakes

• Pasting confidential client data into tools your client hasn't approved. Check your agreements and use business plans with data controls where needed.
• Sending the first draft unread. AI will confidently include deliverables you never discussed.
• One price only. Options let the client choose how to say yes instead of whether to.

Your 30-minute checklist

• Run Call Prep, print the questions
• Record the call and say the numbers out loud
• Summarize, then list problem, outcome, budget, exclusions, open questions
• Draft with Proposal Builder using your rate card and terms
• Review every line, then send with a short cover email
• Set a follow-up reminder for two days later

Browse all of these items, plus 240 more free skills, prompts, and agents, in the YourAIKit directory.

YourAIKit products are workflow tools. They are not legal, tax, or financial advice.