Out went what Armstrong calls “pure managers.” In came “player-coaches,” flat hierarchies capped at five layers, and “AI-native pods” that could include one-person teams directing agents that do the work of engineers, designers, and product managers.

“We are not just reducing headcount and cutting costs,” Armstrong wrote on X. “We’re fundamentally changing how we operate: rebuilding Coinbase as an intelligence, with humans around the edge aligning it.”

Layoffs economywide are relatively low. But inside tech, the AI layoff is the story of 2026. And Armstrong put his finger on something real: The org chart of the American corporation is under a kind of pressure that most companies have not yet figured out how to relieve.

A sweeping new index from KPMG, built from surveys of 300 C-suite leaders, analysis of earnings calls from 177 publicly traded companies, and hard capital data across six industry groups, puts numbers to what many executives are quietly living. The verdict: 81% of executives said boards have raised expectations for their organizations’ adaptability. The organizations beneath them, in most cases, are not ready to meet it.

Atif Zaim, KPMG’s deputy chair and U.S. managing principal, frames it as a factory floor reckoning.

“When electricity first came about … it wasn’t until much later that folks said, you can reorganize the entire factory—proximity to the boiler or proximity to the river or the water wheel is no longer important,” he told Fortune. Most large companies, he suggests, are still standing next to the boiler, acting as if they are still in the steam era when electricity is rewiring the factory floor.

Coinbase’s restructuring is one of the most explicit public expressions yet of this dynamic. By capping its hierarchy at five layers and increasing its employee-to-manager ratio to 15-to-1, Armstrong is making a structural bet the old command-and-control machine cannot survive the AI era. He is not alone: Meta’s new applied engineering team runs at a 50-to-1 ratio, and the broader “megamanager” trend has pushed the average manager’s span of control to 12.1 employees, up from 10.9 in 2024, according to Gallup.

Courtesy of KPMG U.S.

The KPMG data bears this out in sometimes uncomfortable detail. Only 30% of executives say their organization’s structures, roles, and processes can reconfigure quickly as business needs change. Only 24% identified more dynamic talent deployment as a key change made over the last year. The C-suite has embraced the language of transformation. The org chart hasn’t moved.

In a separate interview with Fortune that predated the KPMG index, Zak Kidd, an economist and cofounder of the organizational feedback startup Ask Humans, said he thinks the AI reorg will go further than most executives are prepared to admit. In his view, the management layer that most Fortune 500 org charts are built around is not just being challenged. It is structurally threatened.

“The future organization is just equity holders and essential workers with LLMs in between,” he said, adding agents will play a large role. “There’s really no need for the management function of human beings if large language models can do the discernment.”

The disconnect between executive expectation and organizational reality shows up in the numbers CFOs are watching most closely. A new survey by the Federal Reserve Bank of Richmond, cited by Apollo chief economist Torsten Slok, finds CFOs expect AI’s biggest impacts to fall on decision speed and accuracy, output per worker, and time spent on high-value tasks—in that order. Total employment, by their own projections, remains essentially unchanged. In other words, the C-suite expects AI to rewire how work happens inside the existing structure—without substantially reorganizing the structure itself. Kidd’s argument is this is exactly the wrong way to think about it.

The spending tells the real story

Follow the money and the culture war becomes concrete. Across every industry group in the KPMG index, executives report that increasing investment in new technology was the top action they took last year. They are nearly twice as likely to increase tech spending as to invest in employee training. Less than half say they find technology “very effective” at improving adaptability.

The math of that trade-off has a human cost. Four out of six industry groups in the index recorded year-over-year declines in hiring. Consumer retail shed headcount at a 7.9% rate. Health care, already strained, dropped 5.6%. Companies are simultaneously demanding more adaptability from their workforces and making those workforces smaller. The result: 46% of executives report burnout and change fatigue as an unintended consequence of their adaptability efforts.

The macro data is beginning to validate what the KPMG index measures at the organizational level. An AI disruption tracker published on April 13 by Morgan Stanley economists finds AI’s impact on the labor market is “narrow, early-stage, and largely visible in micro data rather than aggregate outcomes”—but the micro signals are pointed in one direction. Young workers in high-AI-exposure occupations have seen unemployment rise more sharply since 2023, and they are taking longer to find new jobs. Meanwhile, company earnings calls are increasingly referencing AI in the context of labor, with mentions of displacement rising faster than mentions of job creation. The macro picture looks stable. Beneath the surface, the adjustment is already underway.

“In times of disruption, workers need more training and support, not less,” the KPMG report states flatly. Almost no one is providing it. While 57% of leaders say improving performance and efficiency was a top priority last year, fewer than 10% say developing stronger workforce training programs was among their primary objectives.

Pointing out that KPMG itself is more than 150 years old (Fortune is a relatively spry 96), Zaim added, “You’ve got layers upon layers of lore and culture and muscle memory that has been built into, ‘Yeah, this is how we do things.’ A lot of this stuff is not written down anywhere. It’s not like you can go and change the policy and suddenly things change.”

Here is where the white-collar culture war gets its most revealing data point. The industries most focused on innovation—most aggressive about new technologies, business model experimentation, and accelerating R&D—are not the most adaptable. The correlation is essentially zero.

Manufacturing and energy, which scores the highest strategic adaptability of any sector in the index at 71 out of 100, is not a sector known for radical reinvention. It adapts through disciplined scenario planning, centralized decision authority, and operational execution. Health care leads the overall index at 53—not because it innovates fastest, but because it scores consistently across culture, ecosystem, and strategy. TMT—the sector that most loudly evangelizes transformation—scores 41 on cultural adaptability, near the bottom.

Zaim invoked Kodak, a company that literally invented the digital camera and was destroyed by it anyway.

“Is there a risk that you are a leader, excellent in innovation, but somehow that diffusion, that adoption, that getting it into the rest of the organization didn’t happen because you didn’t have the culture for it?” The answer, the index suggests, is yes—and it is happening at scale, right now, inside some of the most sophisticated companies in the world.

Courtesy of Ask Humans

Kidd said he sees such failures as symptomatic of something deeper—a fundamental misunderstanding of what AI actually does to organizational intelligence. He said he was skeptical of consulting firms and corporations that try to draw bright lines between what humans can do and what AI cannot. “Human skill capability ceiling is fixed,” he said, “whereas on the other side of the curve, it’s not fixed.” Any boundary you draw today, he said, is just a “sandcastle.”

The companies investing heavily in innovation while neglecting culture, he argued, are making the same category error: assuming the gap between human and machine capability is stable when it is, in fact, closing faster than most org charts can process.

The answer, in aggregate, is no. Just 9% of executives—across all industries—identify increased psychological safety as one of the behaviors their organization changed most in the past year.

Delta Air Lines said it is replacing auxiliary power units (APUs), a type of engine, on its Airbus aircraft to address toxic fume incidents that can harm airline crew and passengers.

The airline confirmed to CBS News Thursday that its replacing engines on more than 300 of its A320s in an initiative that began in 2022. The airline did not comment on the cost of the undertaking, which is more than 90% complete.

Delta is replacing engines that can cause toxic fumes to leak into the aircraft cabin’s air supply, posing health and safety risks to crew members and passengers. 

Flight attendants have been calling attention to the issue for years, noting their own experiences with the dangerous toxins.

“I felt like I might die, and I thought, you know, I wonder what my mom is going to think — I went to work, and you know, I may not make it,” flight attendant Vanessa Woods told CBS News in 2016 of breathing in engine fumes on a flight she was crewing. 

The leakage of fumes most commonly occurs on the Airbus A320 family of narrowbody jets, a recent Wall Street Journal investigation found.

Breathing in the toxic engine emissions can lead to general nervous system problems, Dr. Robert Harrison, an occupational medicine specialist at the University of California San Francisco who has treated more than 100 flight crew members for exposure to fumes, told CBS News.

Delta on Thursday said it is also exploring new synthetic turbine oils from different manufacturers for its engines. 

Reached for comment on the issue, United CEO Scott Kirby recently told CBS News that it is a matter he has “personally been focused on for over a decade.”

United has a “proactive maintenance program” that tracks the plane engine’s oil consumption and oil pressure, and proactively replaces APU seals to prevent oil leaks and oil burns that would lead to fumes entering airplane cabins, Kirby said. The issue, he added, isn’t a concern on newer airplanes, like Boeing 787 Dreamliner, which have a different design.

Barry Biffle, CEO of Frontier Airlines, which operates an Airbus fleet, said that while the company is following the toxic fume issue, incidents occur very rarely.

Apple once considered replacing Google as its default search engine in private browsing mode on its products in favor of DuckDuckGo, according to recently unsealed testimony by the rival search CEO.

“Our take was that they were actually really interested in this,” DuckDuckGo CEO Gabriel Weinberg said, according to a transcript viewed by CNBC of the testimony in federal court last month. “The people we were talking to were generally DuckDuckGo users themselves interested in privacy.”

Weinberg, testifying as part of the Department of Justice’s antitrust case against Google, said DuckDuckGo and Apple had about 20 meetings and phone calls over the course of the negotiations, which lasted from about 2016 through 2019. Throughout that period, Weinberg said Apple’s contract with Google to be the default search engine on its Safari browser “was often the elephant in the room.”

The government is trying to prove that Google’s exclusive contracts with phone and browser makers unfairly locked out rivals from the general search market by depriving them of distribution. Google has denied that its actions violated antitrust law.

DuckDuckGo, which is privately held, makes a privacy-focused search engine that competes directly with Google, as well as other privacy products that seek to limit how websites can track consumers across the internet.

DuckDuckGo first got a response from Apple about its idea to become the default search engine in private browsing in 2016, Weinberg said. DuckDuckGo claims its search engine greatly reduces the amount of tracking that is still possible in other search engines, even while on private browsing mode.

In 2017, DuckDuckGo was able to secure a meeting with Craig Federighi, a senior vice president, at Apple’s headquarters in Cupertino, California to discuss its proposal. DuckDuckGo presented Apple executives with data about what Apple users expect from private browsing mode, which Weinberg said he’d thought “was pretty compelling.”

Weinberg’s team said they could “make DuckDuckGo the best search option on Apple devices for Apple users by integrating their content into search modules” including Apple News, Maps, Music and TV.

Weinberg said he left the meeting with the impression that “it went very well.”

“I’ve pitched lots of things to Apple over the years,” Weinberg testified. “If there’s no interest, their move is basically silence.”

DuckDuckGo executives returned to Cupertino the following summer for another meeting and presented visuals of how the product would look once its search engine was integrated into Apple services.

Weinberg said his “impression was that they were really serious” about the idea “potentially for the next year’s release.” He said that Adler asked them “to come back basically as soon as possible to brainstorm what privacy integrations could look like.”

Separate recently unsealed testimony from Apple’s side tells a somewhat different story.

John Giannandrea, Apple’s senior vice president of machine learning and artificial intelligence and a former Google executive, joined the company in 2018. He testified that he wasn’t aware of Apple considering the search default switch.

Still, he went on to describe discussions with other Apple executives about the potential drawbacks of such a proposal. Giannandrea worried DuckDuckGo’s “marketing about privacy is somewhat incongruent with the details,” since he thought the company would have to share some user information with Microsoft due to its arrangement to receive search information from Bing.

DuckDuckGo says in its privacy policy that it prevents “our hosting and content providers from creating a history of your searches and browsing.”

In September 2018, DuckDuckGo returned to Apple headquarters to discuss integration, Weinberg testified. Apple “expressed they were really considering this for the 2019 release,” and Weinberg then realized they still had some lingering concerns. In particular, Apple realized it need to figure out how to resolve issues tied to its Google contract, Weinberg testified.

Sometime after the 2018 holidays, DuckDuckGo received documentation from Apple showing what its revenue share would be if it were the default. DuckDuckGO estimated its market share “would increase multiple times over” just by becoming the default in private browsing mode.

By the summer of 2019, DuckDuckGo began to understand the partnership would not happen. Apple didn’t announce the integration during its Worldwide Developers Conference in June. Four months later, following a meeting, Weinberg’s takeaway was that the deal was “dead.”

DuckDuckGo had also pitched Samsung, Mozilla and Opera on being the default option in their private browsing modes, but was not able to reach a deal with any of them. The company eventually stopped pursuing this model because it concluded “That each of these companies’ Google contract was the key thing preventing us from getting a deal done with them.”

Subscribe to CNBC on YouTube.

WATCH: DuckDuckGo CEO would like to see a search preference menu