If a piece of software can do your job from a laptop, it’s worth asking what can’t be done from a laptop at all. That question has been sitting with me for a while. Not in a panicked way. More in the way you notice the tide coming in and decide where to put your chair.
My answer, for what it’s worth: the safest income to build right now is small, physical, local, and boring. A snack machine in an apartment lobby. An ATM in a bar. A laundromat. A row of storage units. A car wash. None of these will make a great headline. All of them require something to happen in the physical world, at a specific address, with someone’s permission. That combination is hard to automate away.
I’m not a financial advisor, and nothing here is a promise about income. It’s how I think about where to build.
The copy test
Here’s the simplest way I know to judge whether a business is exposed to AI. Ask how someone would copy it.
- Could they copy it with a prompt? If your product is words, images, code, analysis or answers, increasingly yes. That doesn’t make those businesses worthless. It makes them crowded, and it pushes prices down.
- Could they copy it with money? A business that’s only a machine is copyable by anyone with the same budget. Buy the same machine, get the same result. Almost.
- Could they copy it with a relationship? This is the hard one. The property manager who said yes to you, the bar owner who wants the person they know servicing the ATM, the landlord who signed a ten-year lease on a laundromat location. Nobody gets those by typing.
The businesses I keep coming back to are the ones where the answers are “no,” “partly,” and “only by beating you on trust.” That’s a strong position for a small owner.
Where AI does touch these businesses (and it does)
I don’t want to oversell this. AI is showing up in physical businesses too, and anyone telling you otherwise hasn’t looked lately.
The obvious one is the AI cooler: a glass-door fridge with cameras that watches what you take and charges your card when you close the door. They cost more than a traditional machine and carry a monthly software fee, but they can sell fresh food and bigger items and often do more at the same site. Storage facilities rent units through kiosks with nobody on site. Car washes read your license plate and charge your membership. Laundromats take app payments and alert the owner when a machine goes down.
Notice what all of those have in common. They make one owner able to run more. None of them removes the need for somebody to own the location, restock the product, fix the broken thing, and keep the relationship. I’ve written before about using AI to 10x my writing; this is the same idea applied to a route. The tool multiplies the operator. It doesn’t replace the operator.
The honest catch is that tools also lower the barrier to entry. When starting gets easier, more people start, and good locations don’t multiply. The competition moves to the thing that can’t be automated: getting the yes.
The risk that sneaks in the side door
Here’s the part I think most “AI-proof business” lists miss. A vending machine in an office can’t be automated away. But the office workers who buy from it can. If a company cuts a third of its staff because software now does their work, that break room gets quieter, and the machine sells less.
So I’d think about AI exposure at two levels: the business itself, and the customers it depends on. The locations I’d favor are the ones where people have to be physically present no matter what: apartment buildings, hospitals, factories, warehouses, gyms, hotels. People still sleep somewhere, get sick somewhere, build things somewhere, and wash their clothes somewhere.
There’s also a slow trend worth knowing that has nothing to do with AI: cash use keeps drifting down. The Federal Reserve’s payment research found cash was about 14% of consumer payments by count in 2024. That doesn’t kill an ATM in the right bar, but I wouldn’t build a plan that assumes cash volume grows.
The real goal isn’t replacing your salary. It’s making it optional.
This is the part I care about most. Most people hear “start a vending business” and imagine quitting their job in six months. That’s not how it usually works, and I’d rather be honest about that.
The better frame is a freedom floor: the monthly amount that covers your family’s non-negotiable bills. Mortgage or rent, utilities, insurance, food, car. Say that’s $4,000. (If you haven’t calculated yours, I walk through it in how to calculate your financial freedom number.)
Now suppose a decent machine at a decent location nets somewhere around $200 to $350 a month after product, the location’s cut and fees. That’s a hypothetical range, and plenty of placements do worse. At $250 a machine, covering a $4,000 floor takes about 16 machines. That sounds like a lot. But look at the halfway points:
| Machines (at ~$250 net each) | Monthly net | What it covers in a $4,000 floor |
|---|---|---|
| 1 | ~$250 | A utility bill. Mostly proof. |
| 4 | ~$1,000 | A car payment and insurance |
| 8 | ~$2,000 | Half the floor. A layoff stops being a crisis. |
| 16 | ~$4,000 | The floor. The job becomes a choice. |
That middle row is the one I’d aim for first. Half the floor covered means a lost job is a hard season instead of a disaster. Everything past that is the job slowly becoming optional.
What the path looks like
An illustrative example, not a real person: Tom is a paralegal. His firm starts using AI for document review in 2025 and hires fewer juniors. Tom still has his job, but he can see where it’s heading.
He signs one apartment complex. The machine does about $1,100 a month, maybe $275 after everything. He doesn’t spend it. By the end of the first year he has four machines, deliberately in different kinds of places: two apartment buildings, a hospital break room, and a warehouse with three shifts. No offices. His combined net covers his car and insurance, and his first equipment loan is paid off early.
Year two is slower than he hoped. One location changes management and asks him to leave, and it takes two months to replace it. That’s normal; I’d plan on it. By the end of year two he’s at nine machines and one ATM in a sports bar. Roughly half his family’s floor is covered. He hasn’t quit. But the next time his firm announces a reorganization, he reads the email without his stomach dropping. That feeling is the product.
If this pushed you from curious to serious, the next read is AI-Resistant Businesses in 2026: Vending, ATMs, Laundromats, Storage and Car Washes — it’s written for people actually doing it.
What I’d do this month
Not “buy a machine.” That’s step four.
- Write down your freedom floor. The actual number.
- Pick the model your savings can support. For most people starting out, that’s vending or ATMs; laundromats, storage and car washes usually take six or seven figures.
- Make a list of twenty places near you where people have to show up in person.
- Go get one yes.
For step three, VendBuddy (full disclosure: it’s my company) will pull apartment complexes, gyms, hotels, laundromats and storage facilities in any ZIP along with who to contact, and you can buy credits one pack at a time if you’d rather not subscribe.
I’ve said it for years, most loudly in never rely on just one job: the point of building something isn’t to escape work. It’s to make sure no single decision made in someone else’s meeting can take away your family’s footing. AI just made that advice more urgent. My older thoughts on AI and the future of work are still worth a read next to this one.
