8 Businesses for the AI Age: How to Build Something That Is Hard to Copy

If anyone can press a few buttons and build your product, what do you actually own?

That is the question behind a lot of the anxiety around AI right now. It matters whether you are an employee thinking about starting a company, a founder trying to build an asset, or an incumbent with billions in revenue wondering whether a frontier AI lab will move into your market.

As software, content, and answers become cheaper to produce, the ability to produce them becomes less of an advantage. But that does not mean businesses stop being valuable. It means we need to get more thoughtful about where the value lives.

I think about this through eight categories: atoms, permission, attention, timing, belonging, liquidity, unique data, and outcomes. These are not guarantees of immunity from AI. They are ways to build a business around something that a competitor cannot simply generate overnight.

Table of Contents

• 1. Atoms: Physical Businesses That Cannot Be Cloned Overnight
• 2. Permission: Licenses, Regulation, and Partnerships
• 3. Attention: Build a Media Machine Around a Niche
• 4. Timing: Use Early Entry to Accumulate Something Durable
• 5. Belonging: Paid Communities and Real-World Experiences
• 6. Liquidity: Marketplaces Still Have Plenty of Opportunity
• 7. Unique Data: Sell Access to Something Others Cannot Easily Get
• 8. Outcomes: Customers Paying for Results
• The Best Businesses Can Combine These Advantages

1. Atoms: Physical Businesses That Cannot Be Cloned Overnight

There are pixel businesses, like software, and there are atom businesses, which operate in the physical world.

A coffee shop needs to serve actual coffee. A plumber needs to show up and fix the problem. Generating a website for a plumbing company does not repair a leak.

That distinction sounds obvious, but it becomes more important as digital products get easier to copy. Physical operations create friction that software alone cannot remove. Someone still has to deliver the service.

Of course, physical businesses are harder to operate. That is part of the tradeoff. You are not getting the convenience of a purely digital business, but you are also not competing in a market where someone can reproduce your entire offering in an afternoon.

I also think there is a case that valuations for some of these businesses could rise as more AI and technology money looks for durable places to invest. That is a prediction, not a certainty. But the growing interest in so-called boring businesses makes sense to me.

The opportunity is not that physical businesses avoid AI. It is that AI cannot instantly duplicate their physical delivery.

2. Permission: Licenses, Regulation, and Partnerships

Another useful question: what does your business have permission to do that a new competitor does not?

Permission can come from a few places:

• Licenses: Credentials or approvals required to operate.
• Regulation: Rules that limit who can own or deliver a particular service.
• Partnerships: Relationships that provide access to customers or distribution.
• Preferred-vendor status: A position inside another company’s buying or referral process.

Imagine you run an agency and become a preferred vendor for Salesforce, Microsoft, or a medium-sized company. A competitor might copy your services page. They cannot automatically copy that relationship.

This is why being good at relationships still matters. Getting in front of the right people, earning their confidence, and putting a deal together can create an advantage that does not come bundled with an AI subscription.

Regulated professional services offer another way to think about permission. Ownership and qualification requirements vary by jurisdiction. In some places, owning a law firm requires being a lawyer; certain accounting or bookkeeping activities can also carry professional requirements.

Where the rules allow it, one possible approach is to partner with a qualified professional. They handle the professional practice, while you contribute marketing, growth, and AI systems.

The point is not to treat regulation as something to get around. It is to recognize that legitimate access, credentials, and relationships can be part of the business’s defense.

3. Attention: Build a Media Machine Around a Niche

Somehow, I still think attention is underrated.

People follow people. They also return to brands they trust. In a world where anyone can generate content, having an audience that actually cares about what you publish becomes more valuable, not less.

AI makes building that audience more practical. I used to hate creating short-form video. I enjoyed consuming it, but the editing process made producing it a chore. With an AI-assisted editing workflow, I can create on my schedule, and I have started enjoying it.

Over one recent week, that content generated hundreds of thousands of views, along with feedback that people found it useful. The interesting part is not just the reach. It is how much friction disappeared from the production process.

That same idea applies to:

• Newsletters.
• Niche meme pages.
• Founder-led content.
• A broader media company serving a specific audience.

I would not think about this as creating one page and hoping it grows. I would think about creating a repeatable media operation around a niche or micro niche.

What does that audience care about? What do they consistently need help understanding? How can AI amplify your ability to serve them?

Not all media is defensible. Generic, interchangeable content is exactly the kind of thing AI makes abundant. The goal is to create a trustworthy brand people choose for a particular subject.

You do not necessarily need to be the face of that brand. But having a real human associated with it can help. The machine supports the relationship; it is not a substitute for one.

4. Timing: Use Early Entry to Accumulate Something Durable

I keep seeing trends emerge in smaller and smaller niches. Sometimes I am shocked by how few businesses exist to serve the people participating in them.

That is an opportunity, especially if you build a system for noticing those changes rather than waiting for an idea to appear.

You can use something like Idea Browser or a personal AI agent to monitor trends in the markets you understand. The important thing is to connect discovery to action.

A useful sequence is:

1. Monitor a niche: Look for changing interests, demand, and unmet needs.
2. Identify where you can add value: A trend alone is not a reason to start a business.
3. Build an initial offering: It could be software, a physical product, a media company, or a marketplace.
4. Accumulate assets: Learn from customers and develop data, insights, relationships, and trust.

Here is the counterintuitive part: your first product might not be particularly AI-proof. It might be very easy to copy.

But being early can give you something more durable than the product itself. You begin understanding the niche before other people do. You build a brand. You gather people around the business. You earn loyalty.

Timing matters when you convert an early start into an accumulating advantage. Being first is useful, but collecting insight and trust is the more interesting part.

5. Belonging: Paid Communities and Real-World Experiences

People are more connected than ever, yet many feel disconnected. That creates a real opportunity to build businesses around belonging.

I am not talking about opening a group chat and calling it a community. A strong community usually has a shared interest, a goal, and a reason for members to make progress together.

That could take the form of:

• A paid community hosted on Circle or Skool.
• A weekly virtual session focused on accountability.
• A recurring local meetup.
• A fully in-person experience built around a specific group.

People may be paying for accountability, progress, entertainment, or the experience of being around others who understand what they are trying to accomplish.

My friend Sam Parr’s company, Hampton, is a useful reference point. As I understand it, it brings established entrepreneurs together through local groups and meetups across North America. The interesting question is: what is the equivalent for another specific audience?

What group would spend hundreds or thousands of dollars to be around the right peers, with the right host helping them move forward?

The host’s work is not just organizing a calendar. It is curating the group and helping members along their journey.

We recently held an event in Nashville for founders interested in using agents to build their businesses. Demand was strong enough that we were not entirely sure how to meet it. That reinforced how much appetite exists for getting the right people together in person.

The hard questions are practical:

• Who exactly is the community for?
• What goal brings members together?
• How will you deliver a genuinely good experience?
• What price makes the value obvious?

AI can support the operation. But the relationships between members are a different kind of asset from generated information.

6. Liquidity: Marketplaces Still Have Plenty of Opportunity

People tell me software is finished. Marketplaces are one of the first counterexamples I bring up.

A marketplace is simply a place that connects supply and demand. The website or app is the interface. The harder thing to build is the activity on both sides.

That is liquidity: having relevant buyers and sellers show up so that transactions can happen.

A friend recently launched a marketplace for high-end collectible cars, with curated vehicles and a recurring auction format. He already had a large Instagram audience, which gave him a way to bring attention into the marketplace.

That combination is interesting: an existing audience, a defined niche, curation, and a particular transaction mechanic.

Cars & Bids is another useful example of a valuable marketplace in the car category. The broader point is that marketplaces can be incredibly sticky because assembling the participants is difficult.

A competitor can copy the interface. That does not mean the buyers and sellers move over.

Two Ways to Build the Supply of Attention

You can build an audience yourself and then turn some of that attention into marketplace activity. Or you can partner with a creator or media company that already has an audience and help build its marketplace.

Either way, I would ask:

• Which niche is underserved?
• Where are people struggling to find the other side of a transaction?
• Could a different buying or selling mechanic make the experience better?

The niche does not have to be new. It just has to have unmet demand.

AI Can Reduce the Operating Burden

Historically, marketplace management could be intense. Support tickets, coordination, and a large number of participants created a lot of operational work.

AI agents can now handle parts of that workload. I think that opens the possibility for a team of one to five people to support a marketplace doing many millions in gross merchandise value, or GMV.

GMV is the value of transactions flowing through the marketplace, not the marketplace’s own revenue. The important point is the potential operating leverage: a small team can support more activity than it could have before.

Use AI to make the marketplace easier to run while building the liquidity that makes it hard to replace.

7. Unique Data: Sell Access to Something Others Cannot Easily Get

Unique data creates two related opportunities: selling data to AI labs and selling access to useful information directly to customers.

Data for AI Models

AI labs are hungry for types of data they do not already have. Coding-related data has been one example, with companies such as Mercor and micro1 participating in the broader market for model-training expertise and data.

At a high level, the business question is straightforward: what valuable data can you access or create that the buyer cannot easily obtain elsewhere?

Then comes the second question: how do you package and sell it?

The value is not simply that something is called a dataset. The value is its usefulness and the difficulty of obtaining an equivalent.

Curated Directories and Databases

The other route is a curated website, directory, or app that people pay to access.

I expect more of these businesses to emerge because infinite content creates a need for trusted selection. LLMs can produce endless answers. Social feeds can deliver endless posts. Neither automatically gives you a collection you trust.

Restaurant recommendations are a good example. I often find reviews polarized between love and hate, and I am skeptical of how some review systems get gamed. I would like a source of thoughtful, consistently useful recommendations rather than another pile of star ratings.

That could become a curated database someone pays for. It is an illustration, not a claim that a consumer restaurant app is an easy business. Consumer businesses can be riskier and harder to make work.

I also see opportunities in enterprise, small-business, and prosumer markets. The underlying question stays the same: can you create or curate information that is difficult to get elsewhere and valuable enough to pay for?

You might monetize that through recurring access to an app, through sales to AI labs, or through both.

8. Outcomes: Customers Paying for Results

The final category is automated AI services, where customers pay for completed work rather than access to another tool.

Imagine a creative agency powered by multiple models and agents that produces logos or landing pages. The customer wants the result. Your system coordinates the work that delivers it.

Another approach is to study what people already buy on Fiverr or Upwork. Those platforms reveal demand for specific services. There may be opportunities to build systems that identify relevant work, bid on it, and use agents to fulfill parts of the service.

But I would not rank outcomes as the strongest standalone defense on this list. Of these eight categories, it is probably the weakest by itself.

If everyone has access to similar models, an automated service can be copied too. The defense has to come from something surrounding the models:

• Context: A deeper understanding of the customer and the task.
• Data: Better inputs that competitors do not have.
• Systems: A workflow that consistently produces a better output.
• Permission: A relationship or agreement that others cannot easily reproduce.

The opportunity is not merely to have agents. It is to give those agents better data, insights, inspiration, and operating context.

Automated delivery becomes more defensible when it is attached to another advantage.

The Best Businesses Can Combine These Advantages

I would put physical businesses and permission-based businesses high on the list for defensibility. Outcomes would sit closer to the bottom on their own. But I would not turn the whole framework into a rigid ranking.

You do not have to choose one category and stay there.

• Attention plus belonging: Build a trusted niche audience, then create a paid community around a shared goal.
• Permission plus liquidity: Use a valuable partnership to help establish a marketplace.
• Unique data plus outcomes: Give an automated service better inputs than competitors can access.
• Timing plus brand: Enter a niche early and turn that head start into trust and customer insight.

That is the real exercise: think about what your business accumulates as it grows.

If the answer is only more code or more generic content, I would push further. Does it also build relationships, access, data, trust, community, or a physical operation that is hard to reproduce?

These eight categories are not an exhaustive map. They are starting points for finding value when the cost of creating digital things keeps falling.

I do not look at that world and conclude that we are done creating businesses or building assets. Sometimes the most chaotic periods create the most opportunity. You just have to be creative about what is worth owning.