Self-Funding the Machine – The Brutal Math Behind Tech’s New A.I. Layoff Strategy

Early in 2023, Eric Vaughan called an all-hands meeting with his worldwide remote workforce and gave a message that wasn’t wrapped in the customary corporate assurances. “We’re all in this together” was absent.No, “our people are our greatest asset.” Rather, the CEO of the enterprise software company IgniteTech informed his staff that artificial intelligence would now be the focus of every project, priority, and working hour. He then replaced almost eighty percent of the attendees in that meeting over the course of the next year. Recently, he was questioned about it. Without hesitation, he declared that he would do it again.

The operating logic of what is currently happening in the technology industry is contained in that story in a more focused and honest form than most companies are willing to publicly express, so it is worth reading slowly. Although those explanations have been convenient, the layoffs that hit the tech industry in 2025 and 2026 are not primarily related to post-pandemic overcorrection or economic downturns.

They are increasingly focused on a particular financial calculation: the money saved by reducing the number of human employees is being discreetly but purposefully transferred into the AI infrastructure that will perform the tasks that those humans previously performed. In a very real sense, the machine is financing itself. The employees who are being let go are covering the cost of their own replacements.

Before the layoffs started, Vaughan committed 20% of IgniteTech’s payroll to a mass AI learning initiative. Every Monday was designated as AI Monday, with no budget meetings, no customer calls, and only AI projects in every department, including marketing and sales. He called in outside specialists. He paid for AI tools and prompt-engineering training for staff members.

According to him, it didn’t work. It wasn’t a lack of technology, but rather the fact that belief was more difficult to produce than skill. He said, sounding genuinely surprised, that the technical staff was the most resistant, more interested in cataloging what AI couldn’t do than investigating its potential. Contrary to expectations, the sales and marketing staff leaned in. The engineers dug in their heels. Then, methodically, they vanished.

CEO Jack Dorsey specifically cited AI efficiency as the reason behind Block’s announcement of the layoffs of about 4,000 jobs, or nearly 40% of its workforce. More than 12,000 positions will be eliminated by TCS in 2025, with mid- and senior-level employees most at risk. These figures are not insignificant, and the individuals behind them are real people. They are experts who have spent years honing their skills in systems that are now being transferred, function by function, to tools that don’t need health insurance, don’t need a salary, and don’t oppose AI Monday. The engineers who developed or implemented these systems are now getting the same termination notices that the systems were designed to optimize, which is a particularly cruel dynamic that is difficult to ignore.

Although the phrase “self-funding the machine” sounds metaphorical, it actually describes the financial mechanics quite literally. A company can free up $100 million annually by eliminating 500 engineers at an average fully-loaded cost of $200,000 annually. Over the same time period, the cost of a serious enterprise AI deployment—the kind that can significantly absorb a large portion of that engineering work—is much lower. The transition is fully funded by the delta.

Investors are aware of this math, which is why announcements of headcount reductions combined with commitments to invest in AI have performed well on earnings calls, unlike layoff news alone. The market seems to have determined that this trade is not only acceptable but also admirable from a strategic standpoint. It remains to be seen if this consensus will hold as the magnitude of displacement becomes more difficult to overlook.

Because Vaughan didn’t conceal the rationale behind cozy euphemisms, the IgniteTech case merits careful analysis. He made no mention of “rightsizing,” “streamlining,” or “focusing on core competencies.” He stated bluntly that it was more difficult to change people’s minds than to acquire new abilities, and that those who refused to change were replaced. It’s uncommon to be so direct. The majority of businesses using the same playbook do so with much more cautious language and much less openness about the true motivations behind the decisions. However, the result is functionally the same: human capital is treated as a variable cost and reduced proportionately to what AI can absorb, with the money that is freed up going back into the infrastructure that is absorbing.

It’s still unclear how far this logic can go before it encounters structural constraints, such as those related to regulations, the true limit of what current AI systems can reliably accomplish, or the organizational reality that someone still needs to have a thorough understanding of the business in order to meaningfully direct the machines. A hint of that can be found in Vaughan’s own story: in order to make the transition successful, he needed enough human buy-in, and when he didn’t get it, he replaced those individuals rather than realizing that the strategy itself might need to be modified.

That’s one interpretation of the circumstances. Another possibility is that he was just impatient with the timeline and correct about the direction. Whether or not the experiment was intended, the tech industry as a whole is conducting the same experiment at a scale that will eventually yield an answer, and the truth is most likely somewhere in between those readings.