AI Isn’t Going to Change Everything — It’s Going to Separate Winners from Losers

Imagine that tomorrow all large companies double their productivity thanks to Artificial Intelligence. Record profits. GDP goes up. Headlines celebrate. And yet, something feels off: people on the street don’t feel it in their wallets. Stores have fewer customers. Restaurants, fewer full tables. The economy grows on paper but hollows out from within.

That’s exactly what James Van Geelen, founder of Citrini Research and the most-read finance writer on Substack, called “Ghost GDP” — ghost GDP — in an essay published in February that went viral. Millions of views in days and, according to several reports, it contributed to an 800-point drop in the Dow in a single session. His central argument: machines don’t buy pizza. If AI replaces white-collar workers, companies capture all the productivity gains, but those gains never return to the real economy because the displaced stop consuming. A tree that looks green on the outside but is rotting on the inside.

The real risk isn’t that AI destroys jobs. It’s that it destroys consumers.

It’s a thesis that deserves to be taken seriously, especially because there’s already something about today’s economy that feels a bit like that — companies report record profits, markets rise, but many people feel that their daily lives don’t reflect those numbers. In our view, the fear is overstated. But ignoring the question would be a mistake.

In February, Citadel Securities published a response direct to Citrini — and it’s worth knowing who Ken Griffin is: the same man who in January called AI “pure garbage” in Davos, and who in May publicly admitted he changed his mind after seeing the results inside his own firm — published a direct reply to Citrini, and his argument is empirical: if the collapse in employment were already happening, it should show up in job-search data. And it doesn’t. Job postings for software engineers on Indeed rose 11% since May of last year, precisely as the overall labor market fell 4%. Customer service: up 9%. Banking and finance: up 9%. Accounting: up 18%.

Citadel explains this with something economists call the Jevons Paradox, though the concept is pretty intuitive: when something becomes more efficient, total demand goes up, not down. The first time it was observed was with coal in the 19th century — improving the steam engine didn’t reduce coal consumption; it sent it soaring, because all of a sudden coal could do many more things. For knowledge work, the logic is the same: if an engineer can produce twice as much with AI, more companies will want to hire engineers, because now they can build things that were previously out of reach for them.

Gutenberg also scared the scribes

An analogy we find even more illustrative: Gutenberg’s printing press. When Gutenberg mechanized book production in the 15th century, hand-copyists disappeared — that specific job, yes, was eliminated. But what came next was not less work but an explosion in demand for readers, editors, translators, teachers, booksellers, and eventually massive universities. AI will probably eliminate certain tasks, just as the printing press eliminated copyists. What it will not eliminate is the need for people who know what to do with all that new knowledge.

Now, we don’t know whether this will be as transformative as the printing press or the Industrial Revolution. Although we think it will be, those comparisons are easy to make and hard to sustain. What we do know is that the historical pattern — technology that expands demand rather than destroys it — has repeated enough times that the burden of proof lies with those predicting collapse, not with those predicting adaptation.

And Citadel found something else worth mentioning: of the S&P 500 companies that mention AI in the context of their workforce, 8 out of 10 describe it as a complement to their employees, not as a replacement. Only 5% talk about AI as a direct substitute.

Software: where the fracture is already visible

Where the impact is already asymmetric and measurable today — and directly relevant for portfolios — is in the enterprise software universe.

The logic of the problem is simple: if AI allows companies to do more with fewer people, they need to buy fewer software licenses. And the business models of the largest software companies in the world were built exactly on that user count. Salesforce, Adobe, and ServiceNow fell as much as 40% to 60% from their highs, bottoming in April. It wasn’t because they reported bad results — it was because the market asked an uncomfortable question: if there are fewer employees, how many subscriptions stop getting bought?

Evolution Wealth Advisors

Our read is that this drop is related to uncertainty, not a death sentence. These are companies with huge customer bases, long contracts, and brands that have been built over decades. They’re not going to disappear. But they are required to reinvent themselves, and in that process there is real risk. Salesforce and ServiceNow are betting on becoming platforms that coordinate AI agents instead of simply selling user licenses. If that bet works — and we believe they have the pieces to make it work — this year’s drop will have been overdone and there’s an opportunity there. If they don’t execute well, the adjustment has only just begun. It’s exactly the kind of situation where it’s worth looking for opportunities and distinguishing between winners and losers.

The clearest contrast comes from Palantir, which reported in its last quarter revenue growth of 85% year over year — the highest in its history as a public company — with operating margins of 60%. The difference isn’t that Palantir is “more technological” than Salesforce. It’s that it charges based on the value it generates for its customers, not on how many users use it. When AI reduces the number of people at a company, Salesforce’s model suffers directly. Palantir’s does not.

What this means for portfolios

Three specific questions worth reviewing.

What kind of exposure do you have? An index like the S&P 500 or the Nasdaq already gives you access to the big winners in AI infrastructure — Nvidia, Microsoft, Alphabet, Amazon — without needing to bet on individual stocks. If you have concentrated exposure to traditional software, it’s worth reviewing.

How do the software companies in your portfolios charge? Software companies that charge by number of users face a structural risk if AI reduces the number of licenses sold. Those that charge based on value generated are better positioned.

What’s the real time horizon? The history of technology suggests that productivity gains accrue over years and decades, not quarters. The risk isn’t that technology fails—it’s that capital becomes too concentrated too quickly, while returns are distributed gradually.

Most people today are asking the wrong question. It’s not “is AI going to eliminate jobs?”—that answer depends on the sector, the country, the role, the year. The right question is simpler and more urgent: are your assets on the side of those who capture the value of this transformation, or on the side of those who absorb it as a cost? That distinction is already showing up in prices. And it will keep doing so.

Article written by Michael Marciano and Santiago Lossada

Disclaimer / Legal Notice:

EVOLUTION WEALTH ADVISORS LLC has prepared the Content. as a service to EVOLUTION WEALTH ADVISORS LLC. readers. It is provided only for general information, and it is not intended to address your requirements or to constitute any form of advice or recommendation (including but not limited to legal advice). EVOLUTION WEALTH ADVISORS LLC. has used reasonable efforts in collecting, preparing and providing quality information and material, but does not warrant or guarantee the accuracy, completeness, adequacy or currency of the information contained in the Content. Users of information on any EVOLUTION WEALTH ADVISORS LLC. letters do so at their own risk and acknowledge that this information should not be relied upon in making (or refraining from making) any specific investment or other business or personal decisions (professional advice should always be obtained prior to making any such decision). In no way should EVOLUTION WEALTH ADVISORS LLC. be liable to you for any claims relating in any way to your inability or failure to perform any kind of research or related work or to perform such research or related work properly or completely, even if assisted by EVOLUTION WEALTH ADVISORS LLC., your use of the Content or your reliance on the information contained in the Content.

Evolution Wealth Advisors LLC has prepared this presentation as a service to its clients and readers. This presentation provides general information and does not constitute any form of advice or recommendation (including, but not limited to, legal advice). Evolution Wealth Advisors LLC has made reasonable efforts to collect, prepare, and provide quality information and material, but does not guarantee the accuracy, completeness, or timeliness of the information contained in this letter. Users of any material sent by Evolution Wealth Advisors LLC do so at their own risk and acknowledge that this information should not be relied upon in making (or refraining from making) any specific investment or other business or personal decisions (professional advice should always be obtained before making such a decision). In no way shall EVOLUTION WEALTH ADVISORS LLC be responsible for any claim related in any way to your inability or failure to conduct research, or to conducting such research in an inadequate or incomplete manner, even if you have the assistance of EVOLUTION WEALTH ADVISORS LLC