I'm always surprised there aren't more accidents on the road.
There's a traffic circle in my neighbourhood that I drive through almost every day. Half the people seem to have no idea how it works. Everyone's coming and going, trying to figure out who has the right of way, and somehow we all get through it.
It's like some form of managed chaos.
I've been thinking about that quite a bit lately because I've spent the last three weeks trying to organize all the information associated with our construction business.
And I've come to a similar conclusion.
I'm surprised any business survives at all.
It's Amazing How Much We Have to Coordinate
Think about everything that happens inside a business.
Accounting information lives in one system. Payroll and time sheets live somewhere else. Customer information is in another system. Project documents, emails, photos, proposals, invoices, and employee communications are scattered across all sorts of different places.
Somehow, all this information has to come together for a business to function.
And if you're the owner, you're expected to make decisions based on it.
Our construction business is probably better organized than many because we've built most of our own systems. We're not dependent on dozens of different applications to run our daily operations.
But even with our own software, making sure all this information is connected and understood is a significant amount of work.
I've been reminded of that recently by the Microsoft email outage and some changes at QuickBooks.
It's remarkable how much businesses have come to depend on technology they don't control.
But I think something important is beginning to change.
I Connected ChatGPT to Our Business
Over the past few weeks, I've connected both ChatGPT and Grok to our company's systems.
I've made selected, sanitized datasets available so ChatGPT can examine business information without being given customer names, email addresses, property addresses, and other identifying details.
The idea is that I can have a conversation with AI about what's actually happening inside our business.
Not a hypothetical business.
Our business.
I can ask how things are progressing from week to week. I can ask what areas might need additional support, what patterns are developing, or what questions I ought to be asking.
And what's surprised me is that I've learned a few things about our business that I hadn't considered before.
I've been doing this for a long time.
You'd think I'd know most of what there is to know about how our company operates.
But that's the thing about being a business owner. You don't know what you don't know.
And it's often difficult to find someone who has enough context to help you see what you're missing.
Your employees understand their respective responsibilities. Your friends might be good sounding boards. Your accountant knows your financial information.
But very few people have a complete picture of what's going on.
AI creates an opportunity to bring that information together and start asking better questions.
There's a Difference Between Using AI and Connecting AI
We're already using AI in several parts of our business.
We use it to help extract job tasks from construction drawings, prepare safety paperwork, and develop safety meeting agendas using information about incidents and near misses.
We've also connected Grok to help with more action-oriented tasks, including emails and social media.
But what I've been focused on recently is something different.
I'm trying to create a digital picture of our entire business.
A digital story might be a better description.
I want AI to understand how a lead becomes an opportunity, how we estimate and win work, how an employee's time clock entry becomes a payroll record, and how work completed in the field ultimately results in an invoice.
That's much more useful than simply having AI read an email or summarize a document.
Think about our daily construction logs.
At the end of each day, we receive information from job sites about what happened, who was there, what work was performed, and sometimes photographs documenting the work.
That information is important by itself.
But what if the system could also recognize that a change order might be required? Or that a job had reached a point where an invoice should be prepared?
That's where I think things become interesting.
The information already exists. The opportunity is to connect it to the decisions and activities that should follow.
I've Seen This Change Before
I was an entrepreneur when businesses were first getting connected to the internet.
I remember wiring offices with Cat 5 cable, installing routers, building websites, and trying to figure out how to move information from one place to another.
When I was running a mortgage brokerage, one of our competitive advantages was our ability to digitize paperwork.
People used to bring in physical pay stubs, employment letters, and bank statements. We built a process to turn that paperwork into digital packages that lenders could access online.
It seems fairly ordinary now.
But at the time, it was a meaningful improvement in how the work was performed.
I feel like we're at a similar point with AI.
Businesses have accumulated all kinds of software designed to make them more efficient. Most owners probably don't even realize how much they're spending every month on software subscriptions.
Now there's an opportunity to connect that information and automate some of the work that happens between those systems.
But it takes more than subscribing to an AI service.
You have to help AI understand your business.
Your Information Might Be Worth More Than You Think
There's another part of this exercise that I'm particularly interested in.
I'm working on putting together a data manifest that could eventually allow us to license selected business information to companies developing AI systems.
My thinking is that specialized business data could become increasingly valuable.
There is a tremendous amount of publicly available information that AI models have already been trained on.
But there are many industries where the most interesting information isn't publicly available.
Take our commercial demolition and asbestos abatement business.
There are all sorts of details about how work is estimated, scheduled, performed, documented, and invoiced that aren't sitting in publicly accessible books and websites.
What might that information be worth to someone trying to build a more capable AI system?
I don't know yet.
That's what I'm trying to find out.
There are plenty of questions about data quality, privacy, ownership, customer confidentiality, and whether anyone will actually pay for it.
But I think it's worth exploring.
At a minimum, organizing our information should make our own business more useful and efficient.
If we eventually find a way to license some of that information, that's an additional opportunity.
Maybe More Businesses Will Survive
Businesses fail for all sorts of reasons.
Ultimately, many run out of money.
But I've often thought that running out of money is the consequence of something else going wrong.
Most business owners are very good at whatever product or service they provide.
What they don't necessarily know is how to manage all the other moving parts required to run a successful business.
That's where I think AI could be extraordinarily useful.
Imagine having something that understands how your company works and can help you identify problems before they become serious.
Not something that takes over the business.
Something that helps you run it.
I don't think AI necessarily eliminates all the work people do. In many cases, I think it changes the nature of the work.
Just as a construction worker can operate a machine rather than doing everything by hand, people in offices may increasingly supervise, review, and approve work prepared by AI.
And if that gives business owners more information, better decisions, and greater capacity without continually adding overhead, I think that could be a very good thing.
I'm still early in this process.
I'm working through how to structure the information, help AI understand our workflows, and eventually explore whether the resulting datasets have commercial value.
There's a lot to learn.
But I'm enjoying the exercise, and I'll keep sharing what I discover.
Because the more I look at how businesses operate, the more remarkable I find it that they work at all.
Maybe AI will help change that.
Maybe we'll eventually understand our businesses well enough that their survival doesn't seem quite so miraculous.
Watch or listen
Watch on YouTube