The Ledger Is Blind

Notes from an evening about AI's hidden costs that turned into something else...

Last month I spent an evening in downtown San Francisco, adjacent to the Databricks summit, at a salon called After Abundance: AI and the New Technical Debt, hosted by The Luminary Societies and led by Rebecca Cidnie Kaykas-Wolff. A few dozen executives, founders, investors, authors, consultants, designers, and operators filled the room. The featured guests were Ben DeBow, a data economist who has spent thirty years watching technology budgets and whose book, End of Abundance in Tech, argues we lost something when we stopped designing under constraint, and Scott Brinker, the longtime martech leader now working as an analyst. The stated topic was tech debt. The actual topic, as the night revealed, was what our accounting systems refuse to see.

Ben opened with the part nobody wants to say out loud: tokens are money, and the industry has worked hard to make sure we don’t feel them as money. I’ll confess my own guilt here. When I set the pricing for my own firm’s advisory services, the tokens never entered the math. I get it. The obfuscation works, even on people who think about this for a living.

The receipts that circulated through the room were remarkable. An AI agent that quietly costs five dollars every time it renames a file, and nobody knows. A company that burned through a year’s token budget in four months. An enterprise nine figures deep in tokens alone in six months, unable to say what the spend actually bought. One consultant who had spoken with some seventy-five Fortune 500 CIOs in the prior month reported that most simply do not know what their AI is costing them, or whose workload caused the spike.

And the bill compounds. Ben’s most sobering point was about time: the average application runs for fifteen years. The code being generated at unprecedented speed right now, much of it by people who have never heard the words “data lifecycle,” will still be running, and still costing, long after the excitement fades. One participant cited projections of a billion new applications by 2028. If we have a tech debt problem today, imagine it then.

Then came the turn, and it’s the reason I’m still thinking about this night a month later.

Ben put it plainly: people appear on the balance sheet as cost. The value they create appears nowhere. We published research at Team Flow Institute in December that landed on this exact fault line.

And the whole room understood the implication at once, because we are living inside it. It is easy to cut people, because their cost is the most legible number in the enterprise, while their judgment, their context, their relationships, and their care are invisible to the ledger entirely. An investor in the room described the first crude math of this wave: this person costs this much a year, the AI costs this much a day, done. Then she described what happened next: some firms discovered the tokens were running ten times the cost of the junior engineers they’d cut, and quietly started rehiring. The measuring stick was wrong. It was always wrong.

The human evidence kept arriving. A venture builder described young workers teaching themselves AI on nights and weekends, unpaid, because the unspoken message is adapt or be cut. A compensation leader at a major tech company described AI usage being written into performance reviews, determining pay and promotion, with no good way to measure it and no answer to the question of why beyond everyone else is doing it.

John Hagel, who has spent decades inside these rooms, distilled it to the two questions senior executives actually ask him in private: how quickly can I automate, and how many jobs can I eliminate? And Charlene Li named the vacuum above it all: leaders have largely gone absent, handing AI to the technology function while every person in their organization is quietly asking what is going to happen to me, and hearing nothing back. That silence is removing psychological safety, and it is directly undermining the high-performance teams these same leaders say they want, because no one does their best work while bracing for the axe.

Here is my read, and it’s the lens I brought into the room, so I’ll own it. There were no villains in any of these stories. The CFO cutting headcount, the board demanding an AI strategy, the young worker grinding through a weekend course out of fear: every one of them is responding rationally to the incentives in front of them. That is what makes this moment dangerous. You cannot shame a system out of doing what it is built to do. Our organizations optimize what the ledger can see and liquidate what it cannot, quarter after quarter, and right now the ledger can see the salary and cannot see the human. What our ledgers cannot see, our organizations will not protect.

The choice underneath all of it is older than AI. Use this technology to need people less, extract the difference, and call it margin. Or use it to make people more capable and share the value those more capable people create. Extraction or co-elevation. AI didn’t create that choice. It put it on every leader’s desk at once, with a timer running.

We are building toward the second answer, an effort called the H-Corp, a standard for organizations that commit to human flourishing rather than replacement, so I admit I hear every conversation through that lens now. But what struck me hardest that night is that the room didn’t need my lens. A data economist, a martech analyst, an investor, a comp leader, and a consultant to CEOs all arrived, from completely different directions, at the same conclusion: the crisis isn’t the technology, and it isn’t even the cost. It’s that we have no honest way to count what people are worth, and we are making irreversible decisions with the blind side of the ledger.

What would it take to change that? The people in that room have suggestions at the ready:

• Ben suggests assigning value to people, not just cost.

• John Hagel wants scalable learning to replace scalable efficiency as the operating model of the enterprise.

• Charlene Li wants leaders to show up and answer the question their people are actually asking.

• I want all of that, plus the instrumentation to make it stick.

There is no excuse for not understanding, planning, and managing AI training and adoption in a way that measures the results.

The evening was called After Abundance, and Ben's book is titled End of Abundance in Tech. Nobody in that room planned the echo, but it kept the whole night honest. Abundance was never infinite. It was just unmeasured. So as the counting begins in earnest, here is what I want to know: when your organization finally starts pricing what it has been treating as free, will the people show up as a cost to be trimmed, or as the value that made any of it work?