Ridgebeam
Field Notes

How to Automate Construction Proposals Without Giving Away Your Moat

A proposal is the letter that tells a client what your firm will do for them and what it will cost.

We were building a system to automate construction proposals for a Bay Area building consulting firm. While we tested it, one proposal looked perfect on the computer screen. The client had asked for one service, and the screen showed one service.

Then someone on our team read the letter the system had written. It offered the client every service the firm sells. Not the one they asked for. All of them.

A second letter had the opposite problem. Three services were picked on the screen, and the letter only listed one.

The screen said both proposals were fine. The letters said otherwise. The only way to catch it was a person who knew exactly what that letter should say.

That knowledge is what this article is about. Knowing exactly how your firm does its work, clearly enough to spot when something is wrong, is the most valuable thing you own. It's your moat: the thing that keeps a competitor from doing what you do.

Anyone can buy AI tools. Nobody can buy how you bid.

Every firm you compete with can sign up for the same AI tools you can, from the same companies, at the same price. A tool on its own doesn't set you apart.

Your data doesn't either. Every firm has years of jobs, bids and cost history sitting in its software, and at most firms it just sits there.

Your moat is how your firm turns all of that into a proposal that wins. Which services go together. Which exclusions always go in. What wording your principal will never sign. Which jobs get a second look before they go out the door.

That's the part you would never hand a competitor. At most firms it isn't written down anywhere. It lives in the head of whoever has written the most proposals.

Your way of doing things is the most valuable thing you own, and right now it goes home at 5pm.

What it takes to turn your process into a system.

The firm in this story came to us because writing a proposal took them about two weeks. They knew AI could help. They also knew that pointing a chatbot at their proposals wouldn't do it. They needed someone who could take the way their firm works and build a system around it.

That's the work we do, and it's hard work. Building with AI is engineering, the same way building a home is construction. It takes a plan, the right structure, and a lot of care.

We started with their process. Their team had built a master template for each kind of work they propose: the version of the letter that holds everything they could ever offer, in the words they want to use. We sat with them and wrote down the rules behind how that master changes for each job.

Then we built the system. When a new job comes in, it takes the details, marks up the master template for that job, and shows every change on the page. The person reviewing can see exactly what was kept, what was struck, and what was filled in.

Getting there took real architecture and screens their team would want to use. It took rounds of back and forth with their people, and weeks of testing, fixing and tuning. The two mistakes at the top of this article were found and fixed in that stretch, which is exactly what testing is for.

The result takes their proposal process from about two weeks to about two days.

Here's what that looks like on a Monday. An estimator opens three proposals the system drafted from last week's new jobs. Each one is marked against the master, with a note on anything unusual. She reads them, fixes one allowance, and sends them on for sign-off. The proposals go out that week instead of two weeks later.

That's the moat at work. The firm's know-how is written down, it's still theirs, and it's doing the job every day.

Why the people who check the work stay in the system.

Speed was never the only goal. This firm already had a rule that three people sign off before a proposal goes out: the person who owns the client, the technical lead, and a principal.

Partway through the build, there was a moment when dropping that step would have made the system simpler. We kept it, and those two letters at the top are the reason why.

A screen checks that the boxes are filled in. A person who knows the work reads the letter and says that's not the job we sold. Both matter, and only one of them knows what right looks like.

The system does the drafting. The people who know the work do the approving. The time saved comes from never starting at a blank page. It never comes from skipping the people who catch what a screen can't.

Write your process down this week, even if you never automate it.

Everything above started with one step, and you can take it yourself without hiring anyone.

Pull your last three proposals or bids. Sit down with your best estimator and go through each one, line by line. Circle every choice somebody made from memory: why this exclusion, why that allowance, why this wording, why this job got priced higher. Next to each circle, write the reason in one sentence.

At the bottom of the page, add one more line: who checks a bid before it goes out, and what they look for.

You'll end up with a page or two of rules, and that page is your moat in writing. It helps even if you never use AI. A new estimator can learn from it. Your best estimator can take a vacation. And the way your firm wins work no longer lives in one person's head.

Once your moat is written down, protect it.

A written-down process is worth guarding. That isn't a reason to keep it in someone's head. It's a reason to know where it lives.

Ask anyone who will work with it, including us, two questions. Where will it live? And what will you do with it?

For the firm in this story, the answers are plain. Their system runs on its own server. Their own proposal software stays the master record, and ours is a working copy. We keep the answers behind each proposal for 12 months and then delete them, keeping only the finished document and a record that it was made. If they ever want a shorter window, their policy wins.

We tell every firm the same thing: we do not train any AI model on your data.

You can run one check this week. Ask your team what they paste into free AI tools. If the answer includes bids, pricing or client names, some of your moat is already sitting somewhere you didn't choose. That's not a reason to panic. It's a reason to set one clear rule, and our AI policy field guide shows where to start.

Your moat is how you work. Ours is building around it.

The AI tools are for sale to everyone. The way your firm bids belongs to you, and once it's written down and built into your system, nobody can copy it.

That's also Ridgebeam's moat. Out-of-the-box software is built for thousands of firms, so it asks every one of them to work the same way. If you bend your process to fit it, you give up the thing that makes you hard to compete with.

We work the other way around. We sit down with you, learn how your firm runs, and design the system around that. Your process stays yours, and your moat stays intact. The only thing that changes is how fast your work gets done.

You build custom for your clients. So do we.

Frequently asked questions

Can AI write construction proposals on its own? It can draft them. A person who knows the work should approve every one, because a proposal can look right on screen and still be wrong on paper.

How much time can automating construction proposals save? It depends on your process. For the firm in this article, proposals went from about two weeks to about two days.

What do I need before I automate construction proposals? A master template for each kind of work you propose, and written rules for how your team changes it for each job. Most firms have the first and have never written down the second.

Will automated proposals make my firm sound generic? Not if the system is built on your master template and your rules. Then it writes your proposals faster, and they still sound like you.

Is it safe to put my bid data into AI? It depends on where the data lives and what the provider does with it, so ask both questions first. We do not train any AI model on your data.

See how it works. How It Works walks through what we build and in what order. Talk it through. Book a free 30-minute discovery call. If we can solve your problems, we'll tell you. If not, we'll tell you that too.

Book a discovery call

Field Notes, in your inbox

One useful article for custom home builders and remodelers, one or two times a week. No filler.

No spam. Unsubscribe any time.

By subscribing, you agree to our Privacy Policy.