DailyDime

AI will make more digital work cheap and abundant. That may increase the value of businesses where the scarce part is still physical, local, regulated, or built on trust.

The more I use AI, the more I keep coming back to the same question as a small business owner: where is it actually safer to build? The International Labour Organization estimates that one in four jobs worldwide sits in an occupation with some exposure to generative AI.

That does not mean one in four jobs disappears. The ILO says transformation is more likely than replacement. But the pressure is not evenly distributed, and that matters if you are deciding where to spend the next 10 or 20 years of your working life.

AI Is Eating the Most Digital Work First

AI is strongest where the work already lives inside a computer. Clerical occupations remain the most exposed in the ILO’s 2025 analysis, while exposure has also risen in highly digitized jobs such as software, finance, media, and web development.

Real-world usage points in the same direction. Anthropic’s March 2026 Economic Index found that computer and mathematical tasks accounted for 35% of conversations on Claude.ai. Coding was still the single largest occupational category in its sample.

That is not a forecast that 35% of programmers are going away. Claude users are not the labor market, and usage is not the same thing as job loss. It is still a useful signal of where AI is already good enough to absorb meaningful chunks of work.

I have written before about the AI stack and the most common AI use cases. The pattern is hard to miss: words, code, analysis, research, support, and other digital tasks are getting cheaper fast. A lot of software is becoming a feature before our eyes. That is a rough place to build a moat.

The Physical World Never Went Away

The physical world did not spend the last 20 years asleep. We got electric vehicles, reusable rockets, better batteries, drones, connected appliances, advanced semiconductors, and much better robots. Calling the iPhone the last hardware breakthrough would be a fun line and a bad fact.

What is changing now is the amount of capital and intelligence moving back into physical systems. The International Federation of Robotics says 542,000 industrial robots were installed in 2024, more than double the number installed 10 years earlier. About 4.66 million industrial robots were operating worldwide by the end of 2024.

AI is also forcing a gigantic physical buildout of its own. The International Energy Agency says capital spending by just five large technology companies exceeded $400 billion in 2025. It expects that spending to jump another 75% in 2026, driven in large part by data centers and AI infrastructure.

There is some irony here. The technology that makes software feel weightless needs chips, power plants, transmission lines, cooling systems, fiber, warehouses, construction crews, and a truly heroic amount of air conditioning.

The Part Nobody Explains: AI Needs a Physical Economy

The strongest case for the physical world is not that AI cannot touch it. AI absolutely can. The better argument is that AI lowers the cost of coordination while increasing demand for things that still have to exist, move, break, get installed, or get repaired.

That combination is interesting for a small business owner. If software can handle more quoting, scheduling, customer support, bookkeeping, route planning, inventory analysis, and marketing, then the office layer gets cheaper. The field work does not disappear at the same speed.

This is the part of the creative destruction story I find most compelling. Technology rarely destroys demand for everything around it. It rearranges where the scarce value sits. In an AI-heavy economy, scarcity may move away from producing another digital output and toward executing reliably in the real world.

A Better Filter Than “AI-Proof”

I would not use “AI-proof” as the literal standard. No industry gets a lifetime exemption from technology, and robotics will keep improving. The better question is whether AI can replace the part of the business customers are actually paying for. That is also the distinction behind my earlier look at AI-proof careers and businesses.

The most attractive models, in my view, combine physical execution with one or more extra frictions: local density, licensing, regulation, recurring maintenance, emergency demand, trust, or messy environments that are expensive to automate. A chatbot can diagnose a lot. It still cannot crawl into your attic with a compressor.

The point is not that physical businesses are immune to AI. It is that the scarce part of the customer outcome may remain harder to commoditize.

Where I Would Look as a Small Business Owner

I would start with businesses where the valuable outcome happens in the real world and AI can strip cost out of everything around it. Electricians, plumbers, HVAC contractors, equipment maintenance firms, specialty manufacturers, auto repair shops, dental practices, home health providers, and certain logistics businesses all fit pieces of that description.

I would not buy a landscaping company just because a chatbot cannot mow a lawn. Physical alone is not a moat. Commodity physical work can still have ugly margins, labor headaches, low switching costs, and plenty of competition.

The better businesses usually have several advantages working together:

• The job has to happen on site or through a physical asset.
• Demand repeats, becomes urgent, or carries a real cost of failure.
• Licensing, specialized skill, equipment, or capital raises the entry barrier.
• Local density creates a routing, reputation, or response-time advantage.
• AI can lower the non-physical overhead without replacing the core service.

Robots Are the Caveat

Robotics is the obvious counterargument, and it is a real one. Factories already run millions of industrial robots, and AI will make robots more capable in warehouses, manufacturing plants, hospitals, farms, and eventually homes.

But factories are also the easiest physical environments to automate because the environment can be standardized around the machine. A residential service call is different. The house is different, the problem is different, the customer is standing there, and the thing you need is somehow always behind the one pipe nobody can reach.

So I would not bet on physicality by itself. I would bet on physicality plus variability, local context, regulation, trust, or economics that make full automation difficult to justify.

Use AI Everywhere Except the Last Mile

The opportunity is not to reject AI and retreat into analog businesses. That would be like responding to the internet by buying a fax machine company.

The better play is to own something AI cannot easily deliver by itself, then use AI aggressively inside the operation. Let it answer calls, draft quotes, schedule jobs, analyze margins, forecast inventory, follow up with customers, build SOPs, and handle the administrative sludge that makes small businesses harder to run than they need to be.

If you are evaluating a business today, ask two questions. Which parts of this company will AI make dramatically cheaper? And what part still requires a person, machine, license, location, or physical asset to deliver the outcome?

If the first answer improves your margins and the second answer protects your value, you may have something worth building.

The past 20 years rewarded companies that moved more of the world into software. The next 20 may reward entrepreneurs who use cheap intelligence to run the physical world better.

That does not mean software is dead. It means software is becoming more abundant, and abundance usually pushes value toward whatever remains scarce.

The model can write the quote. The compressor still has to come out of the attic.