If you search for “AI business ideas,” you’ll quickly find yourself drowning in an ocean of identical suggestions. AI customer support. AI marketing. AI sales assistants. AI legal research. AI tutors. AI agents. Millions of entrepreneurs are asking ChatGPT exactly the same question, and unsurprisingly, they receive exactly the same answers.
That’s not where fortunes are made. The real opportunity isn’t finding another application for AI. It’s discovering problems that nobody even realized were AI problems in the first place.
Every major technological revolution follows the same pattern. In the beginning, everyone copies existing ideas using the new technology. The first automobiles looked like horse carriages without horses. Early television simply filmed radio shows. The internet started by digitizing newspapers. Only later do people realize that the technology changes the rules completely.
AI is at exactly that stage today. The mistake is starting with the technology instead of the problem.
Stop looking for AI applications
Most people ask, “Where can I use AI?” That question leads directly into crowded markets because everyone starts from the same premise. A much better question is:
“Where is human thinking still expensive?”
Throughout history, intelligence has been scarce. Reading reports, comparing documents, remembering thousands of details, explaining regulations, planning projects, coordinating teams — these activities required people because computers simply couldn’t do them. Today they increasingly can. Instead of searching for AI companies, search for expensive thinking.
Look for organizational scar tissue
Companies rarely operate the way they would design themselves from scratch. Processes accumulate over decades like scar tissue. Someone writes a report. Someone else reformats it. Finance extracts the numbers. Management creates another summary. Executives ask questions. Analysts search the original spreadsheets again.
Nobody designed this process because it creates value. It exists because understanding information has always been expensive. When intelligence becomes cheap, many of these layers simply stop making sense. The most valuable AI companies may not automate individual tasks—they eliminate entire workflows.
Search where software historically failed
Traditional software demanded structure: Databases love tables. ERP systems love forms.
Businesses, unfortunately, don’t.
The most valuable information still lives inside emails, PDFs, meeting transcripts, engineering drawings, phone calls, handwritten notes, contracts and regulations. For decades, companies accepted this as “human work.” Large language models don’t require perfect structure. Suddenly, software can work directly with the messy information that organizations produce naturally. Entire categories of software become possible.
Find professions that are really reasoning engines
Some professions don’t primarily produce things. They connect knowledge. Insurance brokers combine policies with customer needs. Compliance officers connect regulations with business operations. Patent examiners connect inventions with prior art. Building inspectors connect regulations with construction. Technical writers connect engineering with users. Many knowledge-intensive jobs are essentially reasoning middleware between different worlds.AI doesn’t necessarily replace these professions. It amplifies them. And every amplification creates opportunities for new products.
Go where nobody wants to go
AI founders love software, finance, healthcare, and education. Meanwhile, industries like wastewater treatment, municipal procurement, warehouse leasing, industrial valves, cemetery administration, or recycling rarely appear in startup pitches. They should. These sectors are full of regulations, documentation, repetitive decision-making, and decades-old software. They’re also largely ignored. History suggests that neglected industries often produce the biggest opportunities because competition is low while inefficiency is high.
Use randomness as a business tool
Creativity isn’t always about inspiration; sometimes it’s about forcing unlikely combinations. Pick three unrelated industries at random. Suppose you get vineyards, orchestras, and insurance. Now force yourself to answer:
“What would AI make radically easier here?”
Most answers will be terrible. One may be brilliant. Randomness is surprisingly effective because it prevents your brain from returning to familiar territory.
Build a map instead of chasing ideas
Rather than waiting for inspiration, build a systematic search process:
Imagine a matrix. On one axis, list industries: construction, logistics, agriculture, manufacturing, government, retail. On the other axis, list cognitive work: reading, writing, planning, explaining, coordinating, compliance, knowledge transfer, decision-making. Every cell becomes a question.
Construction × Coordinating: “Where does coordination break down between crews during daily sequencing?”
Logistics × Decision‑Making: “How do dispatchers make routing decisions under shifting constraints?”
Agriculture × Planning: “How do farmers adjust plans when weather disrupts expected cycles?”
Manufacturing × Compliance: “How do operators interpret compliance rules during real-time production?”
Instead of one idea, you suddenly have hundreds of hypotheses worth exploring. The entrepreneurs who consistently discover valuable businesses aren’t necessarily more creative. They simply ask better questions.
The products that shouldn’t exist
Don’t ask:
What can AI automate?
Ask:
What product could literally not have existed before 2024?
The most interesting AI businesses won’t be faster versions of today’s software. They will be products that literally couldn’t exist before large language models. Software that continuously reads every new regulation and tells a company exactly how tomorrow’s operations must change. Engineering systems that remember every design decision ever made inside an organization. Procurement platforms that negotiate with suppliers while considering historical contracts, company policies, and current market conditions. Personal knowledge systems that remember every meeting, document, and conversation you’ve ever had.
These are not features. They are entirely new categories.
The next fortunes won’t belong to AI experts
I suspect that the next wave of AI businesses won’t come from people who know the most about AI. They’ll come from people who deeply understand obscure industries and recognize where “expensive thinking” is quietly embedded in everyday work. The AI itself may become a commodity, while the real competitive advantage lies in identifying those hidden reasoning bottlenecks and designing products that remove them rather than simply adding AI features.
Where to go from here
Steve Jobs – “Start with the customer experience and work backwards to the technology” (WWDC 1997)
The Nature of the Firm (Wikipedia overview)
The Nature of the Firm (original paper)
Sam Altman – The Intelligence Age
Andrej Karpathy – “The hottest new programming language is English”
Jobs to Be Done theory (Strategyn overview)
Clay Christensen Institute – Jobs to Be Done
McKinsey – The Economic Potential of Generative AI
Generative AI at Work (Brynjolfsson, Li & Raymond, NBER)
IDEO Design Thinking Resources
Paul Graham – Ideas for Startups