Section I -- The Product That Works Best as an Alibi
In the first quarter of 2026, 78,557 technology workers were shown the door. Nearly half of those terminations -- approximately 37,638 positions -- were publicly attributed to artificial intelligence. The machines, employers explained, had arrived. The humans were no longer necessary.
There is a problem with this explanation. A survey of nearly 6,000 senior business executives conducted by the National Bureau of Economic Research found that nine in ten report AI has had zero impact on employment or productivity at their firms over the past three years. Zero. Not modest. Not emerging. Nothing measurable.
Let those two facts sit next to each other. Forty-eight percent of tech layoffs blamed on a technology that ninety percent of executives say has not affected anything. The Bureau has reviewed many accounting discrepancies in its time, but rarely one this elegant.
AI is the perfect excuse because nobody can prove it didn't.
Section II -- The Private Ledger
The gap between what executives say in public and what they plan in private has always been a reliable indicator of institutional honesty. In this case, the indicator is performing above expectations.
A Duke University survey of 750 chief financial officers, conducted in partnership with the Federal Reserve Banks of Atlanta and Richmond and published as an NBER working paper, projects AI-driven job cuts at approximately 502,000 in 2026. That figure is roughly nine times the approximately 55,000 AI-related cuts reported in 2025.
Nine times. Not a rounding error. Not a gradual increase. A ninefold escalation in a single year.
Only 44 percent of those CFOs said they plan any AI-related cuts at all, which means the cuts are concentrated among fewer than half the firms surveyed. The study's co-author, John Graham, director of the Duke CFO Survey, characterised this as follows: it is "not the doomsday job scenario."
The Bureau appreciates this commitment to measured language. Half a million projected terminations. Not doomsday. Just Tuesday.
Meanwhile, the same research found a productivity paradox so consistent it has its own historical name. Economist Robert Solow identified it in 1987: computers appeared to be everywhere except in the productivity statistics. Nearly four decades later, AI has inherited the tradition. Companies have invested billions. Executives report seeing potential rather than results. As Graham noted, the gains are "not really showing up yet in revenue."
The technology cannot demonstrably improve productivity. It can, however, demonstrably reduce headcount. The Bureau notes that one of these is a business outcome and the other is a press release.
BUREAU NOTE: The Bureau wishes to clarify its terminology. When a company fires employees because it hired too many during the pandemic, that is a "correction." When a company fires the same employees and mentions AI in the announcement, that is "transformation." The distinction is important for investor relations purposes and for no other reason.
Section III -- The Case Study
On 26 February 2026, Block CEO Jack Dorsey announced the elimination of more than 4,000 positions -- approximately 40 percent of the company's workforce -- reducing headcount from over 10,000 to under 6,000. The stated cause: artificial intelligence.
Dorsey was explicit that the business was not struggling. Gross profit, he noted, continues to grow. This was not a company in distress making painful cuts. This was a healthy company attributing mass termination to a tool.
Dorsey went further: he predicted that "the majority of companies will reach the same conclusion and make similar structural changes" within a year. He framed this as leadership. The Bureau recognises it as something else: the establishment of a permission structure.
The Darden School of Business published an analysis by Omar Garriott of the Batten Institute examining precisely this dynamic. Garriott's argument is straightforward: Dorsey's announcement did not merely describe a decision. It licensed a pattern. One CEO cites AI. Other CEOs cite the CEO who cited AI. Eventually, the justification becomes conventional wisdom. Nobody needs to demonstrate that the AI actually works. The social proof is sufficient.
Here is what the AI actually does at Block, according to current and former employees who spoke to reporters: roughly 95 percent of AI-generated code changes still require human modification. The AI tools, employees stated, cannot yet operate independently in regulated areas like banking and money transfers -- which is to say, the entirety of Block's core business.
Forty percent of the workforce was removed. The replacement requires human supervision 95 percent of the time. It cannot perform the company's primary function. But the business is strong. Gross profit continues to grow.
Section IV -- The Manufacturers Testify
Normally, when a product is used to justify a decision, the people who make the product are the last to question the decision. In this case, they were among the first.
Sam Altman, CEO of OpenAI -- the company whose models are most frequently cited as the reason companies no longer need employees -- spoke at the India AI Impact Summit in February 2026. His assessment was notably direct: "There's some AI washing where people are blaming AI for layoffs that they would otherwise do."
The man who builds the AI says the AI is not doing what companies say it is doing.
Marc Andreessen, whose venture capital firm has invested billions in AI companies, was even more blunt. In March 2026, he stated that "essentially, every large company is overstaffed" -- by 25 percent at minimum, 50 percent on average, and as much as 75 percent in many cases. The recent layoff wave, in his assessment, reflects pandemic-era hiring excess, not AI capability.
The remedy, according to Andreessen: "Now they all have the silver bullet excuse: Ah, it's AI."
The Bureau would like to underscore the structural comedy of this arrangement. The person who sells the tool and the person who funds the tool have both confirmed, publicly and on the record, that the tool is being used primarily as a justification for decisions it did not cause. They continue to sell and fund the tool. The customers continue to cite it. Everybody involved knows the stated reason is not the real reason. The system continues to function.
BUREAU NOTE: The Bureau has filed this under "Alibi as a Service." We believe it represents a new category of enterprise software. The product does not need to work. It only needs to be plausible enough to cite in a press release. Current market penetration: 48 percent of Q1 tech layoffs.
Section V -- The Numbers That Don't Add Up
For those who prefer their absurdity quantified, the Bureau offers the following ledger.
In January 2026, the United States recorded 108,435 job cuts -- the highest monthly total since 2009. Of those, Challenger, Gray & Christmas determined that AI was explicitly cited as the reason in approximately 7,600 cases. Seven percent. Not forty-eight percent. Seven.
The NBER survey of 6,000 executives: 90 percent report zero employment impact from AI. The same executives predict, over the next three years, an average employment reduction of 0.7 percent. Less than one percent. Across three years.
Sixty-nine percent of firms actively use AI. The executives themselves use it an average of 1.5 hours per week. Ninety minutes. This is the technology that necessitated the elimination of 37,638 positions in a single quarter.
The GitHub data tells a complementary story. Only 30 percent of AI-suggested code is accepted by developers. Pull requests containing AI-assisted code carry 1.7 times more issues than those written by humans. Organisations report technical debt increases of 30 to 41 percent within six months of widespread AI tool adoption. Only 9 percent of developers believe AI code can be used without human oversight.
The Bureau has reviewed the specifications. The product increases errors, generates technical debt, requires constant supervision, and cannot operate in regulated environments. It has nonetheless been credited with eliminating tens of thousands of jobs. By any conventional measure, this is the most successful product launch in the history of corporate narrative management.
Section VI -- The Mechanism
The pattern the Bureau has documented here is not new. Only the alibi is.
Companies over-hired during the pandemic. Remote work expanded headcounts. Interest rates rose. Growth targets contracted. The workforce needed reducing. This is a story as old as public markets, and it typically generates a predictable news cycle: layoffs announced, stock rises briefly, analysts nod approvingly, the affected workers are described as "impacted" rather than "fired," and the cycle moves on.
What AI provides is not capability. It is narrative cover. The layoff framed as "AI transformation" receives a fundamentally different reception than the layoff framed as "we hired too many people in 2021." The first suggests vision. The second suggests poor planning. Both describe the same action. One of them moves the stock price in the right direction.
The Bureau has observed this mechanism across many eras and many industries. The specific justification changes -- automation, offshoring, synergy, digital transformation, now artificial intelligence. The underlying action does not. What changes is the word inserted into the press release where "cost reduction" used to be.
AI is the perfect excuse because nobody can prove it didn't. You cannot audit a counterfactual. You cannot demonstrate that the algorithm did not, in fact, learn to do the job of the person who was just escorted from the building. The burden of proof runs in only one direction, and it runs in the direction most convenient for the entity making the claim.
BUREAU NOTE: The Bureau reminds all personnel that the phrase "AI-driven workforce optimisation" and the phrase "we are firing people" describe identical outcomes. The former is preferred in all external communications. The latter is reserved for exit interviews and, apparently, NBER surveys.
Filed under: Alibi as a Service. The Bureau of Algorithmic Blame -- operational since the first executive discovered that blaming the machine is easier than blaming the spreadsheet. Your position has been optimised. Your compliance is appreciated and expected.