Pretend for a moment that it is the late 1800s and your name is Theodore Ashcroft and you own a large textile manufacturing company. You have a factory, a workforce, and a healthy customer base. Someone walks in and tells you electricity is going to transform your business. What's your first reaction?

• Bah humbug.
• Why do I need that?
• Bring it on. 

Every one of those responses is plausible. And every one of them is exactly what business owners are saying right now about AI.

Electricity wasn't just a new product. It was the second industrial revolution. It changed how businesses ran, created industries nobody had imagined, and left behind the owners who ignored it. AI is doing the same thing. The digital revolution, the internet, was the third industrial revolution. AI is the fourth. AI is going to transform your business. The real question is what that transformation looks like and whether you'll be part of it.

AI adoption is faster than anything we've seen

Two years after the internet arrived, adoption was around 20%. Two years into AI adoption, it is north of 40%. That speed isn't just due to AI being more useful. The digital revolution already built the highway. People have computers, phones, and broadband. There's nothing left to install before you can use it.

I recently sat on a panel where someone made the case that AI should be thought of as a utility, like electricity, rather than a tool. No business will survive long-term without it, the same way no business survives without power today. That reframe changed how I think about it. Nobody talks about "using electricity" as a skill. We talk about what we power with it.

The 6 levels of AI adoption

My journey accelerated last July when a banker introduced me to Derek Weber. Derek runs a 35-person marketing company and has built well over 100 custom AI agents for his team. He taught me there are six levels of experiencing AI, and most business owners are stuck on the first two.

Level one is the non-user. They don't have time. They're skeptics. I personally know someone who, on principle, refuses to use AI. If that's you, the only way through is curiosity. Suspend judgment and start.

Level two is the supercharged search engine. 

This is where most people land. A recent Gallup poll found that about half of U.S. employees use AI at work in some capacity. Most are using it as a better search engine, a better Google. To move past this level, you have to start asking AI to help you think, not just look things up.

Level three is the executive assistant. This is where impact shows up. I write a 300-word newsletter article to our employees once a month. It used to take me 45 to 60 minutes. Now it takes five. Not an exaggeration. Five minutes. The danger here is that speed feels like the win, and many leaders stop. To move past it, you have to look at your business in thin slices and ask what work AI could do, not just what work AI could speed up.

Level four is the strategic partner. At our company Mosby Building Arts, every one of our 12 managers has a Mosby Business Advisor. It knows our company inside out: where we're going, where we've been, how we perform financially, who our target audience is, how we make money, where the pitfalls are, who reports to whom, what each manager's strengths and weaknesses are. This data is at their fingertips.

Level five is automation, where AI runs across a series of steps without a human touching every one. We're not fully there yet. Most companies aren't.

Level six is whatever comes after. Nobody knows yet.

Don't automate chaos

One of the biggest lessons I've picked up is that if your processes are messy, AI will scale the mess. Automating a broken handoff just gives you a faster broken handoff. Clean the structure first.

Here are two examples from Mosby. The first is a scope-of-work reviewer for our 13 salespeople. Before a scope goes to a client, the salesperson uploads it, and a custom AI specialist checks it for grammar problems, inconsistencies, and conflicting information. Simple, consistent, fast.

The second is on our roadmap: a handoff tool for our eight designers, who currently create scopes of work from scratch and pass them to two estimators. Eight designers means eight different formats hitting two estimators. The estimators have to switch hats every time, putting on their "Laura hat" or their "Shannon hat" to figure out how each designer organizes information. A handoff tool will standardize the format and pull a lot of friction out of the process.

Building the Mosby Business Advisor

The most common question I get is how we built the advisor. The short answer is that Derek developed an operating system, a 30-page document that lives in the file and knowledge section of a custom Claude project. The operating system tells Claude how to behave: don't hallucinate, don't make stuff up, be 90% confident or ask follow-up questions, use informal language, give numbered lists rather than bullets.

From there, I went through a series of guided prompts. Core values. Mission and vision. Strategic goals. Where we make money. Target market. Management team personalities, strengths, weaknesses. I spent about 70 hours on it. I uploaded our employee manual, financials, and employee reviews. The output was seven files, roughly 30 to 50 pages each, that live in the custom Claude project alongside the operating system.

Now, when our marketing director, Rebecca, sits down to build a plan, the advisor knows our go-to-market strategy, our target audience, and our financial reality. It gives her insights grounded in our actual business rather than generic advice.

The work slop problem

One concern I often hear is that AI puts out a lot of low-quality work. That's real. We handle it with shared guidelines for our management team. The first rule:

You're responsible for the output. AI may write your article in five minutes, but you still have to read it carefully.

Make sure there's no BS. Make sure it doesn't say anything you don't want to say. That's a behavior you have to train across the organization.

The second piece is parameters. You can tell AI what you do and don't want, and build those rules into the knowledge section so it follows them every time. When our business advisor provides a list, for example, it's supposed to be numbered, not bullet pointed. You can put guardrails on almost anything if you take the time.

What AI adoption costs

Our subscriptions run about $450 a month. I'm also paying $500 a month to be in Derek's learning cohort, which has been worth the price of admission. Add the 70-80 hours I spent building the business advisor, plus the time spent on everything else we've built. Call it around $500 per month going forward, growing as we add users, plus a real-time upfront investment.

What to do Monday morning to improve your AI adoption

Three takeaways for remodelers who understand that they need to get better at understanding and implementing AI into their workflow:

1. Identify where you and your employees are on the six levels. You can't move forward if you don't know where you're starting.
2. Shift your mindset from "can I do this?" to "how can I do this?" Your business is a black box, and there are things you can focus on if you look hard enough.
3. Find your champions. You can't be the only one driving this in your company. I have 135 employees. I need other people who care about it as much as I do.

And before you automate anything, clean the process and data first. Otherwise you're just scaling chaos.

I'm not an expert. I'm on a journey, just like you. There is a lot I don’t know, but here’s one thing I do know: The cost of standing still is higher than most of us realize.