What Building BidPilot Is Teaching Me
I started BidPilot with an idea that is probably too big for one person.
That is not false modesty. The goal is to help construction suppliers, steel building dealers, and subcontractors handle the messy path from an incoming quote request to a proposal, follow-up, and eventually a project.
In plain English: someone sends information about a job, the app reads it, pulls out the important parts, helps build a bill of materials, creates a proposal, and keeps the work moving instead of letting it die in an inbox.
That sounds clean when I write it in one sentence.
It is not clean when you try to build it.
The first thing this project has taught me is that software is mostly decisions. Not magic. Not just typing. Decisions.
What happens if the quote file is messy? What if the job name is missing? What if the AI pulls out something that looks right but is wrong? What does the user see first? What is safe to automate, and what still needs a human to check it?
As someone with no coding background, I used to think the hard part was “making the app.” Now I think the hard part is knowing what the app should do next.
Early on, I had to force myself into what I called the MVP happy path only. That means the smallest version that works when everything goes basically right.
Not every customer. Not every document. Not every strange edge case. Just one reasonable path:
Upload a quote.
Let the AI extract the useful information.
Turn that into a bill of materials.
Generate a proposal.
Create a project from it.
Even that “simple” path is a lot.
I have had to remind myself that the first version is not supposed to be impressive to everybody. It is supposed to prove that the core idea has legs.
Another thing I have learned is that AI feels powerful and unreliable at the same time.
When it works, it feels like watching a very fast assistant read through a pile of paperwork. When it gets something wrong, I remember very quickly that this cannot be treated like a calculator. It needs guardrails. It needs review. It needs plain steps where a real person can say, “Yes, that looks right,” or “No, fix that.”
That has been humbling.
I did not start this because I wanted to become technical. I started because I have spent enough time around business problems to recognize one that costs people time, money, and sanity.
Construction quoting has a lot of friction. Emails, attachments, spreadsheets, follow-ups, revisions, missed opportunities. There is a lot of valuable work trapped in busywork.
That is what pulled me in.
But the ambition can get ridiculous if I am not careful. “Autonomous construction revenue platform” sounds grand. Maybe too grand. Some days I look at those words and think, who exactly do I think I am?
Then I come back to the smaller version.
Can a supplier upload something and get a useful draft proposal faster than before?
Can the app help organize the job instead of adding another headache?
Can I make one workflow clear enough that a normal person would not need a manual?
That is the work.
If I were talking to someone like me — 50-ish, curious, non-technical, tempted to build something with AI — I would say this:
Do not start with the whole castle. Start with one hallway.
Pick one real problem. Pick one kind of user. Pick one path through the app. Let it be smaller than your ego wants it to be.
Also, expect to feel foolish. I certainly have. There are moments when the builder gives me something that works and I feel like a genius. Then ten minutes later I realize I do not understand why something broke.
That swing seems to be part of the process.
What keeps me going is not that I know exactly what I am doing. I do not. What keeps me going is that each small working piece teaches me the next question to ask.
And at this stage, that may be enough.
Small takeaway: build the smallest useful path, then let it teach you.
Keep reading
The Real Cost So Far
Building MyBidPilot has not just cost AI credits. The bigger bill has been time, focus, and learning where ambition gets expensive.
The One Path Decision
I wanted BidPilot to feel big, but I had to make one small design choice first: stop trying to handle every possible version of messy.
The Upload That Went Nowhere
MyBidPilot had a simple first promise: upload a quote and move forward. Then my “simple” path quietly broke.
The Day It Felt Real
For a while BidPilot was mostly ambition and anxiety. Then the first simple pipeline began to make sense, and I could finally see it.