LISTEN
Listen to this perspective
Written by Oleg Cohen. Companion narration uses the Brian voice.
Last week I sat in the audience at a conference for the middle market and listened to two CEOs talk about AI. Both run privately held companies, one of them around a billion dollars in revenue. Neither runs a technology business. These are operating companies: people, sites, customers, margins.
The first CEO said something I’ve been thinking about since. His company had written an AI policy, but not the kind you’d expect. They wrote it, in his words, to mitigate what might already be happening, not to encourage it to happen. Employees were ahead of management. He summarized the attitude of his younger staff in one line: you don’t have to use it, but I’m going to.
Asked whether the company was building, buying, or partnering, he said all of the above. Asked how they were deciding, he said they were interviewing AI consultants to find out whether their strategy was any good. The other CEO leaned into the microphone and said, let me know when you find a good one.
The room laughed. I didn’t, quite. I opened the first essay in this series with that line, and I keep coming back to it. That exchange is the state of AI in most of the American economy.
Not behind. Unserved.
It would be easy to read that scene as two companies that are behind. They aren’t. Everyone in that room has the chatbot. Many have licensed a serious agentic platform and use it, in practice, as a slightly better chatbot. Adoption isn’t the problem. Adoption happened on its own, from the bottom up, before leadership noticed, which is why the first policy was written to contain it.
What hasn’t happened is the thing that creates enterprise value: turning intelligence into capabilities the company owns. Not a summarizer for the sales team. Something that changes how the company estimates, schedules, prices, staffs, sources, or serves, and keeps changing it as the business learns. That requires someone who understands the business deeply enough to know what to build, and someone who can build it. The CEO on that panel was looking for exactly that person. His peer had been looking too.
In The Missing Owner I asked who inside a company should own AI. In the middle market, the harder question is who outside it could.
Who the AI economy is built for
Look at who is being served. The Fortune 1000 have AI organizations, consulting relationships, and the undivided attention of every vendor with a sales team. Startups are AI-native by birth; they don’t adopt intelligence, they’re made of it. Between those two groups sits most of the economy: family-owned, founder-led, and private-equity-backed companies from a few million to a few billion in revenue. Manufacturers, distributors, contractors, logistics operators, professional services firms, multi-site healthcare, dealer groups. Knowledge-intensive, operations-intensive, and, on this question, alone.
This isn’t new. The middle has always been underserved, and for a structural reason. Venture-backed companies play upmarket by default. Large contracts, expansion revenue, low churn, procurement to sell through, logos that raise the next round. The middle market is the opposite on every axis, so the only things that ever reach it are products that install themselves. That’s how the middle got Intuit, Shopify, and the chatbot. A tool installs itself. A capability doesn’t. Capability looks like services, and services don’t return the fund.
AI was supposed to change this, and it half did. The cost of building software has collapsed. What used to be a two-year integration is now months. But the part that got cheaper is the code. The part that’s still expensive is the judgment: knowing the business well enough to decide which capabilities are worth owning, and designing them so they compound. Nobody has figured out how to deliver that at middle-market prices. The cost curve moved. The gap didn’t.
The most human criterion, and its limit
The first CEO’s decision rule for AI investments deserves respect. He asks whether an application adds joy to the human experience of work or detracts from it. Create flow for teams, he argued, and the financial results follow. His fellow panelist framed it as helping people work in their zone of genius.
I believe them. I also think it’s a criterion for choosing tools, and choosing tools is the part that’s already solved. It says nothing about which capabilities the company should own, which is the part nobody is offering them. It’s the thoughtful, humane version of the individual-productivity trap I described in Busy With AI: everyone is faster, and nothing about the company has changed.
Why it matters
Here is what makes this more than an oversight. The middle is where intelligence would pay the most. In a large enterprise, knowledge is already documented, systematized, and diluted across layers. In a billion-dollar family company, the knowledge that runs the business lives in a dozen heads and thirty years of habit. That’s precisely the material that can be captured, structured, and put to work across every decision the company makes. It’s a better fit for AI than any Fortune 500 back office. And it’s the group with no path to doing it.
These companies employ much of the country. They’re the supply chain the Fortune 500 depends on. A great many of them are owned by private equity, which means the capability gap sits inside portfolios that will be marked and sold within a few years. What they’re missing isn’t intelligence, and it isn’t will. Both CEOs on that panel had plenty of each.
What they’re missing is a counterparty. Someone on the other side of the table who can turn the abundant thing into the scarce thing. The market that serves the top of the economy hasn’t noticed the middle, and the market that reaches the middle can only sell what installs itself.
Let me know when you find a good one.
Where I sit: I advise through Kainora and am building technology for exactly this gap, so I have an interest in it being recognized.