AI Makes It Easier to Leverage Cognitive Capital

Can Cognitive Capital Close Private Equity’s Growth Gap?

Rising multiples, lifted by low interest rates, powered more than half of buyout returns. That era is over. The growth now has to come from what a company knows and how fast it acts on it.

For years, private equity’s best partner was the Federal Reserve. Low interest rates lifted valuations, and rising multiples did much of the work. Bain & Company’s Global Private Equity Report 2026 puts it plainly: low rates “led to steadily rising multiples, which powered over 50% of all buyout returns.”

That lever is gone. Bain’s illustrative U.S. example, built on industry benchmarks, shows how much harder the job has become. In a typical 2015 buyout, a sponsor borrowed about half the purchase price at roughly 6%, paid 10 times EBITDA and sold at 12.5 times. Five percent annual EBITDA growth was enough to return 2.5 times the money over five years, which works out to an internal rate of return of about 20%. Today the sponsor borrows closer to 36% of the price at 8%, pays about 14 times EBITDA and can expect to sell at about 15. To earn the same 2.5 times, Bain says EBITDA now has to grow 10% to 12% a year. Its shorthand is “12 is the new 5.”

So where do the extra five to seven points come from? You cannot borrow them, and you cannot count on the next buyer to pay a higher multiple. You have to build them. I’d like to explore the idea that cognitive capital may be part of the solution.

In 2018, Miles Everson and I argued in strategy+business that companies compete with six forms of capital. Three are familiar: financial, human and natural. Three exploded when the world digitized and they matter most now. Behavior capital is the collection and modeling of data that tracks what customers, employees and machines actually do. Network capital is the set of connection points a company can use to develop and carry out its strategy – especially when it’s the customers’ networks, as in Meta. Cognitive capital is the set of algorithms, captured processes, decision rules and data that together create value. In every firm there is a mix of human and digital versions of cognitive capital. The better companies are always surfing the frontier of "structuring" decisions, developing codified insights from repetitive analyses, and knowledge engineering around those cognitive activities that create distinctive customer experiences and results. We called the second group BeCoN capital, because the three are most effective when used together.

As many pundits, including Brian Arthur have pointed out: BeCoN assets behave differently from the traditional three. A factory wears out. A truck fleet depreciates. A pricing algorithm can improve each time it runs, and it improves faster when it is fed more behavior data and linked to more parameters. In 2018, five of the most highly capitalized U.S. companies, Alphabet, Amazon, Apple, Facebook and Microsoft, which we called bionic, together accounted for about 13% of the capitalization of the U.S. stock market.

In 2018, cognitive capital was expensive. Building it meant hiring scarce machine-learning engineers and collecting years of proprietary data. Only the largest technology companies could do it at scale.

Generative AI changed the cost curve. Stanford’s 2025 AI Index reports that the cheapest model scoring at GPT-3.5’s level on a standard benchmark fell in price from $20 per million tokens in November 2022 to 7 cents in October 2024, a drop of roughly 280 times in under two years. A mid-sized distributor, a regional insurer or a portfolio company can now turn its underwriting rules, pricing judgment and service playbooks into software that makes decisions. Eight years ago that was out of reach.

That is the new leverage point. Financial leverage multiplied the return on equity. Cognitive leverage multiplies the output of the people and processes a company already has.

Consider two companies that have built it. JPMorgan Chase budgeted about $18 billion for technology in 2025 and now treats AI spending as core infrastructure, alongside data centers, payment systems and risk controls. DoorDash was built on cognitive capital from the start. Its dispatch, pricing and matching systems learn from the behavior of billions of orders. In 2025 it handled about 3.17 billion orders and generated $2.78 billion in adjusted EBITDA, more than 45% above the $1.9 billion it reported for 2024. Part of that growth came from the Deliveroo acquisition, which DoorDash counts from October 2, 2025. Excluding Deliveroo, DoorDash’s fourth-quarter orders still grew 20% ( DoorDash ). Either way, the growth is well above the 10% to 12% a buyout now needs.

Now consider two companies whose revenue still scales with people and subscriptions. Gartner sells research largely through annual subscription contracts. Its shares fell 47.9% in 2025. Revenue growth slowed to 2.7% in the third quarter and 2.2% in the fourth, and fourth-quarter contract value grew just 1% in constant currency. Two forces were at work. Cuts to U.S. federal spending reduced renewals, and some clients tried AI research tools in place of advisory work. Gartner’s CEO noted that contract value grew 6% in the third quarter excluding the federal business, so the AI effect is real but not the whole story ( Gartner ).

Accenture is the harder case, and it complicates the simple version of this argument. It cut more than 11,000 roles in a single quarter in 2025 and took $615 million in business optimization costs in the fourth quarter of its fiscal 2025. It then guided to growth of 2% to 5% for fiscal 2026 and delivered 5% in local currency, with the fourth quarter up 7%. Its fiscal 2027 outlook is 3% to 6%, and headcount is back to about 814,000 ( Accenture ). People-based models can keep growing for now. What the cuts show is a cost structure under pressure. When software can do part of what a consultant or analyst does, a business that bills for the person can lose pricing power, and the owner has to decide who captures the gain.

Three questions for anyone who owns a business

Whether you run a fund or own a company outright, you face the same three moments.

At entry, ask whether cognitive capital is part of your core value thesis or an option you hope to exercise later. If it is not in the model, it will not be in the 100-day plan.

During ownership, ask whether you are building it and capturing value from it. Count the decisions that now run on software, and track what they do to margin and growth.

At exit, ask whether you can show the next buyer an asset that will keep compounding. A buyer will pay for a pricing engine with three years of results behind it. A buyer will not pay for a pilot.

The decade of easy money taught investors to look for leverage on the balance sheet. The next decade will reward the investors who find it in what a company knows.