AI-Powered Modular Proposal Engine Development -- 2

At the moment, my quoting process is actually remarkably efficient *in principle.*

I visit a customer, record the consultation, upload the recording into ChatGPT, and ChatGPT turns it into a proposal. I then make a few tweaks, copy it into ServiceM8 and send it.

It works.

In fact, it should work surprisingly well.

The problem is that it still feels like a contractor’s quotation. I spend far too much time tweaking wording, punctuation and formatting, trying to make it look polished. It’s still very text-heavy, and before I know it I’ve wasted a surprising amount of time that I’d much rather spend visiting the next customer. And it still looks sucky.

So I’m not really trying to improve the proposal.

I’m trying to improve the entire process.

What I think we’re actually trying to build is a modular proposal engine.

The challenge isn’t getting ChatGPT to write.

It already does that remarkably well.

The challenge is creating a system where ChatGPT writes specifically for a professionally designed proposal template instead of generating a generic document that has to be forced into one afterwards.

We already have a designer creating the visual side.

Chris is brilliant at that and will work closely with you, so you don’t need to worry about making anything look pretty. Your challenge is making the clever bit happen behind the scenes.

The vision is fairly simple.

Chris creates a library of beautifully designed pages—for example Premium Fencing, Budget Fencing, Paving, Clearance, Gates, Artificial Grass and so on.

The layouts, branding and photography are already built.

Only the words change.

After a consultation, I simply upload the transcript into ChatGPT and tell it something like:

“Include one Premium Fence page, one Budget Fence page, one Clearance page and one Paving page. Finish with the detailed specification.”

ChatGPT then writes specifically for those pages, knowing exactly how much text belongs on each one.

From my point of view, the workflow needs to be almost effortless.

Ideally, I upload the transcript, click a button, and finished PDF ready wording appears.

Then, I will manually add measurements, quantities and pricing at the end because those come from my tape measure, not AI, and they are to important to leave to chance.

I then press a button and it spits out a pdf I can send to them and attach to hubspot.

Everything else should happen automatically.

The less I have to change or modify about the process, the better.

Initially, I’d also keep Version 1 simple.

Rather than trying to generate AI “after” images at this stage of a minimum viable product, there will simply be professionally chosen photography that Chris has already built into each page that doesn’t change.

Premium fence pages showcase premium fences.

Budget fence pages showcase budget fences.

Paving pages showcase paving.

The important thing is that every proposal still feels personal because the words are personal.

It naturally includes the customer’s name and references details from our conversation—dogs, children, privacy, future plans or whatever else was discussed—so it feels written specifically for them rather than copied from a template.

The proposal would then split into two sections.

The front is inspirational as above.

Beautiful pages.

Professional photography.

Easy to read.

The back contains the detailed specification, measurements, materials and pricing.

That’s the section my installation teams will actually build from, and it’s also the definitive record of exactly what’s been agreed.

Ultimately, I don’t think we’re building a quoting system.

We’re building a modular proposal engine that combines AI, design and automation into a workflow that’s incredibly fast for me to use while producing something that feels genuinely premium for the customer.

I’d love to hear your thoughts.

Is it achievable?

I thought of Canva as ChatGPT does seem to interact with it but would you approach it differently?

Come at it from what we want to achieve and work back. I only say chatGPT as it’s what I’ve used up to now. if there is something else that is chew proof and can be used on a mobile im cool with that.

don’t be limited by my imagined process. Just tell me how you will get it done

And if it’s something you’d enjoy building, I’d be really interested in receiving an estimate for creating the underlying system that makes it all work.

P.s - this will all be used with a mobile, in the field.