How We Cut Proposal Time from 2 Hours to 10 Minutes with AI

AI Automation & Workflow

We were spending 1 to 3 hours per proposal obsessing over formatting and design. A Typeform + Make.com + Claude pipeline now generates branded proposals in about 10 minutes. 105 hours recovered annually.

Every proposal was a production

We are a two-person agency. When a discovery call goes well, the last thing you want is a 3-hour formatting session standing between you and a signed deal.

But that is exactly what was happening. Every proposal turned into a design project. Margins had to be perfect. Section headers had to be consistent. The narrative had to sound polished without being generic. And because we care about how things look, we could not just slap text into a doc and send it.

The process looked something like this:

1. Finish the discovery call. Take notes.
2. Open a blank document or duplicate an old proposal.
3. Rewrite the problem statement for this specific prospect.
4. Write the solution narrative tailored to their situation.
5. Format deliverables, timeline, and pricing.
6. Obsess over design details for 30 to 60 minutes.
7. Finally send it, sometimes a full day later.

At 1 to 3 hours per proposal and roughly 5 a month, that is anywhere from 5 to 15 hours of production time. Not strategy time. Not sales time. Document formatting time.

"I do not need a dedicated calendar block to create a proposal anymore. I just go in, fill a form, and that is it." — Rahul Lalia, Founder, RSL/A

The worst part was the delay. A prospect finishes a call feeling excited about working together, and then waits a day or two for a proposal. Momentum dies. We were not losing deals because our work was bad. We were losing them because the paperwork was slow.

Typeform + Make.com + Claude + Google Docs

We needed a system that could take structured inputs from a discovery call and produce a branded, client-ready proposal without any manual formatting.


Step 1: Structured intake via Typeform


After every discovery call, we fill out a Typeform that captures:

• Prospect name, business name, phone, email
• One-line business summary
• The problem they described on the call
• The solution we proposed
• Specific deliverables
• Timeline
• Investment amount

This takes about 5 to 7 minutes. The key insight is that structuring the input is what makes the AI output good. If you dump unstructured notes into an LLM, you get generic copy. If you give it specific fields, you get specific proposals.

Step 2: Make.com orchestration

Make.com watches for new Typeform submissions and triggers instantly. It takes the structured form data and sends it to the next step. No manual intervention. No copy-pasting between tools.

Step 3: Claude generates the narrative

This is where it gets interesting. Claude receives the structured data and generates:

• A compelling proposal title
• A problem narrative that reflects what the prospect actually said
• A solution narrative explaining our approach
• Deliverables formatted as clear scope items
• A phased timeline with milestones

The output is structured JSON, not free-form text. That matters because it flows directly into the template without any cleanup.

We currently use Claude, though the AI model is swappable. The prompt engineering is what matters more than the model.

Step 4: Google Docs template population

The final step injects Claude's output into a branded Google Docs template. Same design, same margins, same section structure every time. The template is the design layer. AI is the content layer. They never interfere with each other.

INFO: Why we switched from PandaDocs We started with PandaDocs ($65/month) for its built-in e-signatures and payment collection. After two months, we switched to Google Docs. The only trade-off is sending a separate payment link, which saves $780/year. For agencies that need integrated payments and signatures on every proposal, PandaDocs is worth it. For us, the manual follow-up is a non-issue.

• Typeform
• Make.com: Free tier available
• Claude (Anthropic)
• Google Docs

The results

• ~10 min: Per Proposal
• 105 hrs: Saved/Year
• 92%: Time Reduction
• 2 hrs: Build Time

Before: 1 to 3 hours per proposal, inconsistent formatting, multi-day delivery delays.

After: about 10 minutes per proposal, identical branding every time, same-day delivery.

At roughly 5 proposals a month, that is about 105 hours recovered annually. The build took 2 hours. Break-even hit on the second proposal.

What actually changed

• **No more rabbit holes. **The template handles design. You never touch margins or fonts again.
• **Same-day proposals. **Prospects get proposals hours after the call, not days.
• **Brand consistency. **Every proposal looks like it came from the same agency, because it did.
• **Scalable. **We could handle 10x the proposal volume without adding any time.

The $780/year PandaDocs savings is a nice bonus, but the real value is the 105 hours. That is over 2.5 full work weeks returned to actual revenue-generating work.

How to build something like this

The framework is simple and repeatable for any service business that sends proposals:

1. **Create a structured intake form. **Capture the specific fields your proposals need. Do not use a blank text box.
2. **Connect to Make.com. **Set up a trigger that fires on new form submissions.
3. **Prompt an LLM with structured output. **Tell it exactly what to generate: title, narratives, scope items, timeline. Require JSON output.
4. **Build a branded template. **Lock the design once. Never touch it again. AI populates the content.
5. **Test with 3 to 5 real proposals before going live. **Tune the prompt based on output quality.

Total build time should be under 3 hours for anyone comfortable with Make.com and API integrations.

We build custom automation systems for agencies and service businesses. If you are spending hours on proposals, reports, or any repetitive document workflow, we can fix that.