Once AI Can Build Almost Anything, These 8 Types of Businesses Become Even Harder to Copy

Greg Isenberg recently posted a question on X: If AI can build almost anything, what should you build?

He called this question the source of much of today's AI anxiety. Software, content, and answers are all getting cheaper, and building a product is no longer as hard as it used to be. But he didn't stop at anxiety; instead, he broke down 8 types of businesses that he believes will get stronger as AI gets stronger.

These 8 categories are:

Atoms, physical businesses. AI can write leases and calculate quotes, but people who crawl under a house on a Sunday to fix a water heater are still scarce. One commenter said that after buying and selling several rental property portfolios, the business they still trust most is one that 'comes with a water heater.'
Permission, licenses, regulation, and partnerships. The more a space is gated by licenses, credentials, and distribution channels, the less likely it is to be replaced by a product built over a weekend.
Attention, media machines. Content production is getting cheaper, but attention isn't increasing. People who can consistently attract attention still hold a strong card.
Timing, early trends. AI can write code, but it can't judge for you whether a trend is too early. Greg gave examples like Hampton, final.com, and Cars & Bids.
Belonging, paid communities and offline events. People can chat with AI, but they will still pay for 'who they know' and 'who they spend time with.'
Liquidity, marketplaces. Matching buyers and sellers isn't hard because of the page; it's hard because of liquidity.
Unique data, unique datasets, data brokerage, and curated directories. Models need data, but what's truly valuable is data that others don't have or can't easily organize.
Outcomes, charging for results. Customers don't care what model you used; they only care whether errors were caught and whether results were delivered.

Under this tweet, the questions and additions were very direct. One commenter said that their AI can build almost anything, but every time they ask it, it still gives them a to-do app. Someone else reminded people that the first step is still talking to customers. Another person said that AI makes building things cheaper, but it doesn't make people change how they work, and adoption processes are still hard.

Greg Moran added another angle: many older companies already have customer trust and years of data, and as AI gets stronger, those two things become more valuable. Krevaya argued that areas involving taste are harder to replace: code either runs or it doesn't, but a video can become mediocre in many ways along the way. Jobi and William cast their votes for on-site work and trust: AI can handle paperwork, but someone has to show up, and someone has to answer the phone.

Aedan's summary was short: Building things has become easier; getting people to care has not.

Greg's full podcast episode is on YouTube, titled 'If You Want To Create Wealth In The Age of AI, Watch This.'

You can also open the YouTube link directly.

If you could take only one sentence away from the comments, it might be this: AI has made 'building it' cheaper, but it hasn't made 'finding people,' 'getting people to trust you,' or 'getting people to change their habits' cheaper. In these 8 types of businesses, the truly hard parts are mostly not in the code.