Jack Dorsey cut 40% of Block's staff and called it AI — Wall Street isn't buying it
Block laid off 4,000 employees claiming 50% AI efficiency gains. Wharton's Ethan Mollick calls it implausible. The stock rose 24%. What's really going on.
Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR: Block announced the elimination of roughly 4,000 jobs — 40% of its workforce — in late February 2026. CEO Jack Dorsey credited AI efficiency gains of over 50% as the justification. Wharton professor Ethan Mollick, one of the most credible voices in academic AI research, called that claim implausible. The stock rose 24% on the announcement. What you are watching is a live case study in "AI-washing": the practice of dressing up a financial restructuring in the language of technological transformation to satisfy Wall Street.
On March 1, 2026, Block — the fintech company formerly known as Square, co-founded by Jack Dorsey — confirmed it was laying off approximately 4,000 employees, representing roughly 40% of its total workforce.
The cuts span Cash App, Square, and corporate functions. The announcement came via an internal memo from Dorsey that was shared publicly, consistent with his preference for radical transparency and his long-running practice of communicating directly through social channels.
The headline number is not subtle. Forty percent is not a trim. It is a structural reshaping of the company. For context:
| Metric | Detail |
|---|---|
| Employees before cuts | ~10,000 |
| Jobs eliminated | ~4,000 |
| Percentage cut | ~40% |
| Official reason | AI-driven efficiency gains |
| Stock reaction | +24% |
In his memo, Dorsey framed the cuts not as a response to financial pressure but as the natural consequence of AI making human labor redundant at scale. The specific figure he cited: Block had achieved greater than 50% efficiency gains through AI deployment across engineering, operations, and other functions.
That claim became the story.
When a CEO announces a 50%+ efficiency gain from AI in a single cycle and uses it to justify eliminating 4,000 jobs, the first question any serious analyst should ask is: where is the evidence?
Ethan Mollick is a professor at Wharton and one of the most cited researchers on AI's actual productivity effects. His work involves running controlled experiments on AI tool adoption — the kind of rigorous measurement that corporate press releases do not typically commission. When Mollick was asked about Dorsey's claim, his assessment was direct: implausible.
That word carries weight coming from him. Mollick is not an AI skeptic. He uses Claude and GPT models extensively in his own work. He has published findings showing meaningful, measurable productivity gains — in the range of 20% to 40% — for specific knowledge tasks like writing, coding, and research synthesis. What he has not found, and what the broader research literature does not support, is a company-wide 50%+ efficiency gain that would justify eliminating nearly half your workforce in a single restructuring.
The problem with Dorsey's claim is not that AI cannot improve productivity. It is that the specific arithmetic does not hold.
If AI gave Block a 50%+ efficiency gain across the board, that would imply the same output requires roughly half the labor input. But Block is not operating at the same output level. Cash App's user growth has been slowing. Square's merchant base has seen increased competitive pressure from Stripe, Toast, and others. If your efficiency gains were so dramatic, you would be reinvesting the capacity savings into growth, not announcing a 40% headcount reduction.
The simpler explanation: Block needed to cut costs. AI provides a culturally acceptable, forward-looking frame for what is fundamentally a margin improvement exercise.
"The question isn't whether AI improves productivity — it does. The question is whether a 50% company-wide efficiency gain is plausible given what we know about how AI adoption actually works in organizations. The evidence doesn't support that." — Ethan Mollick, Wharton School
AI-washing borrows its structure from greenwashing: the practice of exaggerating environmental credentials to appeal to socially conscious investors and consumers. The mechanics are the same. Take a real underlying trend. Claim credit for it aggressively. Use the language of transformation to obscure what is actually happening.
In the AI context, AI-washing typically manifests in one of three ways:
1. Retroactive attribution. A company makes a business decision — cutting costs, consolidating operations, exiting a market — and then attributes it to AI strategy after the fact. The AI did not drive the decision. The decision was made on financial grounds, and AI is the explanatory frame applied post hoc.
2. Vague capability claims. A company says it has "deployed AI across operations" or "integrated AI into our core workflows" without specifying what was automated, at what scale, with what measured outcome. These claims are unfalsifiable by design.
3. Efficiency theater. A company announces productivity gains from AI tools that are technically true in isolated cases but extrapolated far beyond what the data supports. If your legal team uses AI to draft contracts 30% faster, that is not a company-wide 50% efficiency gain. It is a departmental improvement in one workflow.
Block's announcement has elements of all three. The 50% figure is specific enough to sound credible, but there has been no third-party verification, no before-and-after measurement methodology disclosed, and no explanation of how efficiency gains in software engineering translate to eliminating 4,000 jobs across a company that also runs hardware operations, customer support, and financial compliance functions.
Block shares rose 24% on the announcement. That number tells you exactly how investors are processing AI-washing narratives right now: they are pricing the story, not the evidence.
This is not irrational on its face. When a company cuts 40% of its workforce, you are looking at a significant reduction in operating expenses regardless of the stated reason. Cash App has high gross margins. Square's software revenue is sticky. Strip out $400-500 million in annual labor costs and the free cash flow math changes substantially. The stock was pricing the restructuring benefits, dressed in an AI narrative that made the restructuring seem like strategic clarity rather than financial necessity.
There is also a momentum dynamic. In the current market environment, any announcement that links job cuts to AI is processed as a signal that the company is "getting ahead of the curve." It validates the investment thesis that AI is deflationary for labor costs. Investors who hold that thesis buy the story. The stock goes up. The narrative is reinforced.
The risk, which most of those buyers are not pricing, is reputational and regulatory. If the 50% efficiency claim is shown to be materially overstated — and it will be stress-tested as earnings come in — the credibility gap becomes a liability. Customers who worry about service quality, regulators who scrutinize AI claims in public filings, and the employees who remain at Block after watching 4,000 colleagues lose their jobs to a number that does not hold up: all of these are costs that do not show up in a same-day +24% move.
It would be wrong to suggest that companies are not achieving real productivity gains from AI tools. They are. The research is clear on this, even if the scale of gains varies widely by function, tool quality, and implementation rigor.
Here is what the current evidence actually supports:
| Function | Documented AI Productivity Gain | Source |
|---|---|---|
| Software coding (GitHub Copilot) | 26-55% faster task completion | GitHub/Microsoft research, 2023-2025 |
| Legal document review | 20-40% time reduction | Various law firm studies |
| Customer support resolution | 14-34% faster resolution | Salesforce, Intercom data |
| Marketing copy generation | 30-50% faster first drafts | Various agency benchmarks |
| Financial analysis | 20-35% time savings on specific tasks | JPMorgan, Goldman internal studies |
These are real. They are also task-specific, context-dependent, and require significant change management to capture at the organizational level. They do not aggregate into a clean 50%+ company-wide efficiency gain that justifies eliminating half a workforce.
The legitimate version of what Block might be doing: using AI tools to allow remaining engineers to ship features faster, reducing the marginal cost of customer service interactions, automating certain compliance checks. Those are defensible claims. The 50% company-wide figure is not.
The distinction matters because it determines whether what Dorsey announced is a well-reasoned restructuring with honest framing, or a well-timed use of AI language to minimize investor scrutiny of a financially motivated headcount reduction.
The Securities and Exchange Commission has been paying attention to AI-washing since at least 2023, when it began sending comment letters to public companies that made material AI capability claims in their filings without adequate substantiation.
In 2024, the SEC settled its first AI-washing cases, bringing enforcement actions against two investment advisers — Delphia and Global Predictions — for claiming to use AI in ways that were either false or materially misleading. The companies paid combined penalties of $400,000. Small amounts, but a clear signal.
For a company the size of Block, the SEC exposure is more significant. If Dorsey's 50% efficiency claim was communicated in a way that influenced the market — and a 24% single-day stock move is strong evidence of market influence — then the standard for substantiation goes up considerably.
The regulatory framework does not require that a company's AI claims be correct. It requires that they be made in good faith, with a reasonable basis, and without material omission of contrary evidence. If Block's internal data does not support the 50% figure, and if that figure was a material reason investors bought the stock on March 1, 2026, then the company has a problem that the +24% does not fix.
The broader dynamic is that companies have been racing to include AI language in earnings calls and press releases because the market rewards it. Between 2023 and 2025, the frequency of "AI" mentions in S&P 500 earnings call transcripts roughly tripled. Not all of those mentions represent real capability. Some fraction of them are what the SEC is increasingly treating as a disclosure risk.
Block's situation is a stress test for how aggressively the agency pursues that risk in a restructuring context.
Block is not the first company to reach for AI as the explanatory frame for a workforce reduction, and it will not be the last.
The pattern is visible across industries:
IBM announced in early 2023 that it was pausing hiring for roughly 7,800 roles that could be replaced by AI over five years. IBM had been declining in headcount for years before that announcement. The AI framing repositioned an ongoing attrition story as a forward-looking transformation narrative. The stock moved up.
Klarna made headlines in 2024 claiming that its AI assistant handled the work of 700 customer service agents. The underlying number was real — call volumes handled by the AI tool had increased substantially. What was missing from the headline was that Klarna had also been building toward this capacity for years and had cut staffing proactively in anticipation. The AI did real work. The framing compressed a multi-year process into a single dramatic data point.
Duolingo announced in early 2024 that it was cutting contract workers as it "leaned into AI." The company had been building AI-assisted content generation tools for its language lessons. Some roles genuinely were made redundant by those tools. But the blanket framing — "we're replacing people with AI" — overstated how clean the causal line was.
What these cases share: a real AI capability, a financial motivation to reduce headcount, and a communication strategy that leads with the technology story because the market responds better to "AI transformation" than to "cost reduction."
| Company | AI Claim | What Was Actually Happening |
|---|---|---|
| IBM | Pausing 7,800 hires due to AI | Years of declining headcount, AI framing repositioned as strategy |
| Klarna | AI replaced 700 agents | Multi-year automation project, compressed into single PR claim |
| Duolingo | Contract cuts due to AI lean | Some legitimate automation, some straightforward cost reduction |
| Block | 50%+ efficiency gains from AI | ~40% workforce cut, AI claim lacks third-party verification |
The next 90 days matter for two reasons.
First, Block's Q1 2026 earnings will be the first financial readout after the restructuring. If the 50% efficiency claim has any substance, you would expect to see it in operating leverage — meaningful improvement in revenue per employee, gross margin expansion, and faster product velocity. If the numbers do not move in that direction, the AI narrative loses credibility fast and the stock gives back its gains.
Second, the Block case is going to accelerate a conversation that investors, regulators, and employees are already having: what standard of evidence should be required before a public company can use AI efficiency claims to justify material workforce reductions?
Right now, there is no standard. Companies can say what they want in press releases, as long as they are willing to defend it if a regulator asks. Block raised the stakes by making a claim specific enough to be falsifiable. That is either bold confidence or poor legal judgment. We will find out which.
For the employees who lost their jobs: the framing does not change their situation. Four thousand people are out of work regardless of whether the AI narrative holds up. But it matters for the broader precedent being set about what companies owe their workers and their investors when they restructure — and whether "AI made you redundant" is a reason, an excuse, or just a story.
For investors watching other companies in fintech, tech, and beyond: the Block playbook is now visible. When you see a company announce significant layoffs alongside a specific, aggressive AI efficiency claim, ask what the evidence base is. If the company cannot answer that question, price accordingly.
Block announced the elimination of approximately 4,000 jobs in late February 2026, confirmed publicly on March 1, 2026. The cuts represent roughly 40% of the company's total workforce of around 10,000 employees. The reductions span Cash App, Square, and corporate functions across the company.
In an internal memo shared publicly, Dorsey stated that Block had achieved efficiency gains of greater than 50% through AI deployment across engineering and operations. He framed the workforce reduction not as a cost-cutting measure but as a structural consequence of AI making certain roles unnecessary. He did not disclose the specific measurement methodology behind the 50% figure.
Mollick, who researches AI productivity effects through controlled experiments, has documented meaningful AI-driven efficiency gains in specific knowledge work tasks — typically in the 20-40% range for functions like writing, coding, and research. A company-wide 50%+ gain that would justify eliminating half a workforce exceeds what current research supports, especially given that Block's slower growth trajectory does not align with a company that freed up massive productivity capacity and redeployed it aggressively.
AI-washing is the practice of overstating or retroactively attributing business decisions — particularly cost-cutting — to AI capabilities in order to frame them as forward-looking strategy rather than financial necessity. Legitimate AI strategy involves documented capability deployment, measurable outcome tracking, and transparency about what was automated and at what scale. The distinguishing feature of AI-washing is that the AI narrative is applied to justify a decision that was made on other grounds, without evidence to support the specific claims made.
Potentially. The SEC has been increasing scrutiny of AI-related claims in public company communications since 2023, and brought its first AI-washing enforcement actions in 2024. If Block's 50% efficiency claim influenced investors — and a 24% single-day stock move suggests it did — then the standard for substantiation rises. The SEC does not require that companies be right about their AI claims, but it does require a reasonable factual basis and material completeness. If the claim lacks internal evidentiary support, there is regulatory exposure.
Markets priced the financial restructuring benefits, not the AI narrative's accuracy. Eliminating 40% of your workforce represents a substantial reduction in operating costs, which improves the free cash flow outlook regardless of what is driving the cuts. The AI framing made the cuts look like strategic clarity rather than distress. Investors bought the story. That does not mean the story is true — it means the market was willing to pay for the version of Block that emerges from a leaner cost structure, AI transformation narrative included.
Microsoft expands Dragon Copilot into a unified clinical AI platform at HIMSS 2026, promising 50%+ documentation cuts and deep Epic/Cerner integration by end of March.
GSMA launches Open Telco AI at MWC 2026 with AT&T, Ericsson, Nokia, Vodafone and AMD to fix the 84% GenAI failure rate plaguing telecom network operations.
Google's March 2026 Pixel Feature Drop expands Circle to Search globally, adds Gemini real-time visual understanding, AR overlays, and shopping integration to Pixel 7a and above.