BlueCollarIntel

Inspiration

Here is how a good business usually dies in America. The owner turns sixty-eight. The work is still there, the crews still go out, the phone still rings. But there is no successor, and no buyer ever calls, because no buyer knows the company exists. One Friday the owner locks the door and sells the trucks. The town loses its only heating contractor, and a dozen families lose their paycheck. The business did not fail. It was never found.

Six million American businesses will change hands by 2035, carrying five trillion dollars of value. The buyers exist: private equity, search funds, corporate acquirers, lenders, brokers. The records exist too, scattered across hundreds of official sources that were never built to be read together. What does not exist is the connection between them. We built it.

What it does

BlueCollarIntel is the intelligence platform for the handover of the real economy. It reads more than 171 official sources across the United States and Canada and resolves them into one authoritative record per real business: 107 fields covering identity, ownership, licensing, scale, compliance and reputation, and whether anything around its ownership has started to move. On top of the record sit 54+ signal types spanning market intelligence, company intelligence and intent intelligence, so a subscriber sees what the market misses the day it happens.

A buyer asks for what they want the way they would say it out loud: which HVAC businesses in Phoenix are big enough to matter, have few competitors nearby, are well run, and have an owner who might be ready to move? Gemini turns the question into a living rule, the platform answers with the slice of the market that fits, and every company carries the reason it surfaced and the evidence behind each fact.

Every fact on the platform traces to its source and date. Modeled values carry their confidence and are labelled as modeled. The bar is simple: nothing appears on screen that could not survive an investment committee, which is why acquirers and lenders can act on it directly.

How we built it

The whole product runs on Google Cloud: nine Cloud Run services, Cloud Scheduler driving the around-the-clock routines, and Vertex AI serving Gemini. Gemini reads filings and documents into structured facts, turns plain-English questions into live rules, reads every company against its peers and its history, and drafts the alerts, briefs and reports customers act on.

The AI does not just answer questions; it operates the company. It runs its own collection schedule across the source estate, validates each new batch against a quality bar and publishes only when it clears, resolves records that belong to the same business and merges them, and watches everything it has published. In August it validated a full state registry feed end to end and published it to customers with no human in the loop. Every decision it makes is written down: what it saw, what it decided, how sure it was, and which person approved, changed or rejected it.

We built the company the same way. There are four of us. Every decision is written in plain language, and AI agents build to the written standard, test their work against it, and surface what they cannot prove. First code was written 16 July 2026; 224 commits later it is a production system reading a continent.

Challenges we ran into

Resolving millions of filings into one record per real business is the kind of problem where almost right is worse than useless: two records that are really one company must merge, and two that merely look alike must never. We taught the system to act where the evidence is strong and hold the rest for review. Proving absence was harder than proving presence: a source that lists only active licences says nothing about a business it does not mention, and the platform has to know the difference. And serving intelligence that people act on with real money meant building the proof into the pipeline itself, so every answer ships with its evidence or does not ship at all.

Traction

Live at blucollarintel.com. Paid access opened a few weeks ago: nine buyers have paid for trials, and one of the country's largest skilled-trades acquirers began onboarding this week. Teams subscribe at $900, $2,500 or $5,000 a month, depending on how much of the firm runs on it.

Why Entrepreneurship & Job Creation

Every completed handover keeps a payroll where it is, creates a new owner-operator, and lets someone retire with something to show for thirty years. It also opens the profession itself: deal search used to need a research staff, and one person with BlueCollarIntel can cover a region, so searchers, brokers and lenders who could never afford a team can do this work at all. Better visibility, better capital decisions, stronger businesses.

Built With

• celery
• cloud-run
• cloud-scheduler
• django
• gemini
• google-cloud
• maplibre
• next.js
• postgis
• postgresql
• python
• react
• stripe
• tailwindcss
• three.js
• typescript
• vertex-ai