There’s a new AI question every small business person is asking:
Do I hire someone, or build it myself?
My colleague Rand Larsen runs Scalepath, the SMB owner community I started and he acquired. He just put together a report on what the members are building at their companies.
There’s one story I thought was really cool, about a guy named Charles Miller who owns a commercial demolition company in Colorado Springs.
When they started getting federal work, like military bases and airports, the amount of paperwork just exploded. It was more than he could handle, and he was desperate for an assistant.
But first he tried to build something himself.
Now, he’s got a Mac Mini sitting in his office closet named Peggy Olson. (Yes, like Mad Men.)
He fed her (it?) his business processes. Now she has her own email address, can draft (but not send) stuff for him, runs his calendar, and protects his golf time.
Then he loaded in textbooks on estimating demolition jobs. Now she works every bid alongside his human estimator and they compare the two before anything goes out.
Now he says his team forgets she’s not a real person.
Here’s what I like about this: general admin work is a great fit for AI. But making it an expert can give your business an edge.
And building stuff yourself is totally doable.
But it’s not the same for everyone.
One, Charles worked in software before he bought a demolitions company. So he’s the kind of guy that tinkers with this stuff for fun. Which leads me to…
Two: he’s the guy who fixes the thing when it breaks at 11pm on a Friday. And AI things break. A lot.
So I think the rule of thumb is:
Don't build anything your business depends on unless YOU’RE ready to fix it at 11pm.
I’m also a computer guy. (My degree is in computer science.)
So I'm building constantly right now. OpenClaw, Hermes, agents running across most of my businesses. It's the most fun I've had in years.
But none of it is mission critical yet. If it all went dark tonight, nobody would call me. That's on purpose.
Below: my take on finding that balance in your own business.
I’m proud to host Compound Conference again this year at Sundance.
It’s the event for operators who are compounding — businesses, capital, skills.
Feb 8–11, 2027 at Sundance Resort in Utah. Tickets are all-inclusive, space is limited.
Get your ticket now — or book a 15 minute call if you have questions.
A non-technical AI guide worth reading
Rand’s whole report has a ton of great insights from different community members.
A few of the takeaways from the report, specifically around buying vs building tools yourself:
If it's just for you, and you can live with it breaking, build it yourself.
If it's going to be used by employees or customers, and a mistake would be expensive or embarrassing, don't be the one building it for the first time on your own.
Some owners build mission-critical stuff themselves anyway and it works. They're also the ones fixing it at 11pm.
If you have the margin and ongoing need, a dedicated person beats a one-off project. The knowledge compounds instead of walking out the door.
Whatever you pick, expect to revisit it. The cost of building something yourself is dropping fast enough that a decision made a year ago may not be the right one today.
Getting genuinely good at AI may be one of the highest-leverage uses of your time.
The whole guide is free. Great, straightforward reading that isn’t trying to sell you on an AI product.
Grab it from Rand’s site here.
Here’s to Peggy Olson.
Michael
P.S. Table of contents is below. The AEO section was a relief. Turns out it’s not nearly as complicated as I thought.
—
Let me help you...
The Low-Risk Business → 30 years of building, turned into one playbook
Hire With Near → Build high-performing remote teams for 70% less
Bedrock → Validate the numbers before you buy a business
Compound Conference → where business owners figure out their next move