There’s a brand new acronym reshaping how employees take into consideration their careers: FOBO — the Worry of Changing into Out of date. In contrast to conventional job insecurity, FOBO isn’t about getting fired. It’s about changing into irrelevant. 4 in 10 employees now title AI-driven job loss as one in every of their main fears — a share that has practically doubled in a single 12 months, in keeping with KPMG. Sixty-three p.c say AI will make the office really feel much less human. Talent calls for in AI-exposed roles are shifting 66% sooner than they did only one 12 months in the past. In 2026, FOBO grew to become the defining psychological situation of the American office.
After Dario Amodei, CEO of Anthropic, claimed final 12 months that AI might get rid of 50% of entry-level white-collar positions inside 5 years, he was joined inside months by Microsoft AI CEO Mustafa Suleyman, who supplied an analogous outlook. Extra just lately, Senator Mark Warner (D-VA) mentioned that AI leaders themselves have been stunned and alarmed on the tempo of disruption, and they’re “literally consciously pulling back on their predictions because of the short-term economic disruption.” Warner put the brand new school grad unemployment at 35% inside two years.
These are the predictions feeding FOBO — and so they’re touchdown. An enormous new research from MIT needs to pump the brakes. Not on the concern — FOBO, it seems, is pointing in roughly the suitable course — however on the timeline. And the timeline, it seems, modifications the whole lot.
Researchers at MIT FutureTech revealed findings this week displaying that AI’s march by way of the labor market seems to be far much less like a sudden disaster and much more like a gradual, rising flood — severe and accelerating, however not the in a single day apocalypse that has dominated headlines and government anxiousness for the previous two years.
“Rather than arriving in crashing waves that transform a certain set of tasks at a time,” the researchers write, “progress typically resembles a rising tide, with widespread gains across many tasks simultaneously.”
The research, titled “Crashing Waves vs. Rising Tides,” is without doubt one of the most complete empirical examinations of AI’s real-world activity efficiency up to now. The staff of 9 researchers led by Matthias Mertens and Neil Thompson collected greater than 17,000 evaluations of LLM outputs from domain-expert employees throughout greater than 3,000 labor market duties drawn from the U.S. Division of Labor’s O*NET classification system. These duties spanned the whole lot from authorized evaluation to meals preparation, administration to laptop science. Greater than 40 AI fashions have been examined, starting from GPT-3.5 Turbo to GPT-5, Claude Opus 4.1, Gemini 2.5 Professional, and DeepSeek R1.
For anybody gripped by FOBO, the core query the researchers requested can be essentially the most unsettling one: Can AI full these duties effectively sufficient {that a} supervisor would settle for the output with none edits? The reply is already sure — steadily.
Throughout all fashions and job classes examined, AI efficiently accomplished roughly 50% to 75% of text-based labor market duties at a minimally acceptable high quality degree. That’s not a future projection. That’s right now. Extra particularly, the research discovered that by the third quarter of 2024, frontier AI fashions have been already hitting a 50% success fee on duties that take people a couple of full workday to finish.
The development trajectory is steep. Between the second quarter of 2024 and the third quarter of 2025, frontier fashions went from clearing a 50% success threshold on 3- to 4-hour duties to clearing the identical bar on duties that take people a complete week. Failure charges are halving roughly each two to a few years throughout the board, which interprets to annual positive aspects of 15 to 16 share factors in success charges.
Extrapolating these developments — and the researchers are cautious to notice this represents an optimistic, upper-bound state of affairs — AI techniques might full most text-based duties with 80% to 95% success charges by 2029 at a minimally ample high quality degree. For almost all of survey duties, which take a number of hours for a human to finish, the projected 2029 success fee approaches 90%.
MIT doesn’t use the phrase however that is FOBO, calibrated. The concern isn’t irrational — it’s untimely. The water is rising. However the MIT knowledge suggests the floorboards gained’t be underwater by subsequent Tuesday. The researchers’ most consequential line for anxious employees: “Workers are likely to have some visibility into these changes, rather than facing discontinuous jumps in AI-driven automation.” The rising tide provides you time to maneuver. The query is whether or not you’re shifting.
FOBO on the institutional degree
Right here’s the irony: whilst MIT paperwork AI’s sweeping functionality positive aspects, most corporations have but to deploy the instruments in any respect. FOBO isn’t only a private situation, then — it’s an organizational one. In response to Goldman Sachs economists Sarah Dong and Joseph Briggs, citing Census Bureau knowledge of their March 2026 AI Adoption Tracker, fewer than 19% of U.S. institutions have adopted AI. Goldman tasks that adoption will attain solely 22.3% over the subsequent six months.
Compounding that paralysis: solely about one-third of employees say their employer is offering sufficient AI coaching, steerage, or reskilling alternatives — down practically 10 share factors from 2024, in keeping with analysis from workforce nonprofit JFF. Most corporations are leaving employees to handle FOBO alone, with out the infrastructure that will truly resolve it.
That hole has a measurable value. Enterprise employees who do use AI are recapturing 40 to 60 minutes per day, in keeping with OpenAI enterprise knowledge from December 2025, and 75% say they’ll now full duties they beforehand couldn’t do in any respect.
“We continue to observe large impacts on labor productivity in the limited areas where generative AI has been deployed,” Goldman’s economists wrote. “Academic studies imply a 23% average uplift to productivity, while company anecdotes imply slightly larger efficiency gains of around 33%.”
Put merely: the businesses utilizing AI are pulling forward. And the mathematics is unforgiving. Throughout a staff of fifty, that 40-to-60-minute each day time saving interprets to 33 to 50 hours of recovered productiveness each single day. The race is on, then, however many corporations are nonetheless strapping on their trainers and ready for the whistle to blow.
FOBO with a nook workplace
The MIT knowledge lands at a second when company leaders are scrambling to get their arms round a expertise that, as one senior government put it, is “outpacing the ability for humans and businesses to adopt it.” Joe Depa, the worldwide chief innovation officer at EY, advised Fortune in a latest interview that “the technology is in many ways ready, but it’s taking some time for us to … take advantage of it.”
That strain is sharpened by a stark inside actuality at many corporations: 83% of executives — drawn from a survey of 500 enterprise leaders — say they lack the suitable knowledge infrastructure to completely leverage AI.
EY’s purchasers, based mostly on 4,500 surveys, say they nonetheless lack the suitable knowledge infrastructure to completely leverage AI. In different phrases, the expertise is racing forward whereas the organizational plumbing wanted to really use it lags far behind.
FOBO’s cruelest irony
That’s the place the “rising tide” framing gives some reassurance to the numerous corporations grappling with this dynamic. The MIT findings instantly problem analysis from METR, a distinguished AI security group, which has argued that AI capabilities surge abruptly for particular units of duties — a “crashing waves” mannequin that means employees might all of a sudden discover themselves out of date with little or no warning. “We find little evidence of crashing waves,” they wrote, “but substantial evidence that rising tides are the primary form of AI automation.”
The MIT knowledge, drawn from practical and consultant job duties slightly than stylized benchmarks, persistently reveals a flatter efficiency curve. AI doesn’t all of a sudden grasp a slender set of duties and go away the whole lot else untouched. As a substitute, it will get broadly, incrementally higher throughout practically all activity sorts and durations concurrently.
“Workers are likely to have some visibility into these changes,” the researchers write, “rather than facing discontinuous jumps in AI-driven automation.” Extra broadly, the projection of AI enchancment to a near-perfect automation degree by way of the subsequent three years, not the subsequent 18 months of doomsday situations, offers what the researchers name “a window for worker adjustment, particularly in tasks with low tolerance for errors.” Moreover, their estimates assume AI progress continues on the tempo seen over the past two years, that means it’s an upper-bound or significantly quick state of affairs. AI simply might not preserve evolving and advancing as quick because it has just lately.
That issues for the way corporations plan and the way employees put together. A crashing-wave mannequin calls for emergency triage; a rising-tide mannequin calls for strategic adaptation. The MIT researchers argue the latter is the extra correct body — although they’re emphatic that “gradualism is not inherently protective.”
There are significant variations by occupation. Authorized work had the bottom AI success fee among the many domains examined, at simply 47%. Set up, upkeep, and restore work — for text-based duties particularly — topped the chart at 73%. Administration duties got here in round 53%; healthcare practitioners at 66%; enterprise and monetary operations at 57%. In different phrases, no white-collar sector is immune, however some are significantly nearer to the inflection level than others.
Depa mentioned he sees this sorting occurring in actual time inside EY’s personal workforce, and people are appearing unpredictably, even unusually on the prospect of this unusual new work companion. The agency is the third-largest Microsoft Copilot consumer on the planet, he shared, and the adoption knowledge tells a generational story: junior workers are all in; senior leaders are lagging. “When I look at the breakdown,” he mentioned, “two of my junior levels — high adoption, right out of the gate … and then when you get to the more senior levels, that’s where the adoption starts to drop off.”
He described a very worrying cohort: expert, skilled employees who’re merely refusing to make use of AI instruments. “We’ve got some software engineers that are 10x, 20x more productive than last year using AI, like, they’re just killing it.” He mentioned he’s seen employees go from “mediocre” to essentially “at the top of their game” as soon as they grasp these new instruments. On the identical time, you might have others “that used to be really, really strong software developers that are somewhat resistant to using AI,” he mentioned. They’ve an perspective that they’ll do it higher, in order that they don’t want the software. “And they’ve gone from being top of their class to now bottom of the peer group, right. And those are the ones I worry about the most.”
The concern of changing into out of date, in different phrases, is accelerating the very end result that employees dread most. Left untreated, a severe case of FOBO turns into self-fulfilling.
These AI resisters, with super useful expertise and expertise which might be tremendous important, however productiveness lagging their peer group at 10x and even 20x, “at some point, those individuals would have to find a different role,” Depa mentioned. “And I think those are the ones that we’re trying to figure out.”
What’s nonetheless lacking from the AI-at-work story
The MIT staff is cautious to not oversell its personal findings. Excessive task-level success charges, they be aware, don’t routinely translate into job displacement. The “last-mile costs” of integrating AI into precise workflows — organizational friction, legal responsibility issues, the economics of deployment at smaller corporations — stay vital obstacles which might be poorly captured by any benchmark.
Close to-perfect AI efficiency on most duties additionally stays years past 2029. The flat logistic curve that makes the rising tide gradual additionally means the ultimate climb towards 99%-plus reliability is an extended one, a significant buffer for error-intolerant professions in regulation, drugs, and engineering.
“While progress is significant,” the researchers write, “widespread automation, particularly in domains with low tolerance for errors, may still be some distance away.”
The underside line is extra difficult than both the doomers or the dismissers need to admit. AI is already succesful, bettering quick, and headed for many of your inbox within the subsequent three to 5 years. However the transformation is more likely to arrive as a gradual, seen tide slightly than a sudden drowning, which suggests the window to adapt is actual, if not infinite. If you wish to adapt, that’s.
FOBO is rational. The MIT knowledge confirms it. However the antidote isn’t denial or paralysis — it’s precisely what the employees thriving inside EY are already doing: treating AI as a software, not a verdict. The window is open. The query is whether or not you’ll stroll by way of it.
