Leadership1 publisher2 min readPublished
Workers see AI as far more likely to take colleagues' jobs than their own
Irrational Labs found 67% of surveyed workers expect AI to take peers' jobs within a year, but only 15% expect to lose their own. For managers, widespread worry about AI says little about whether any one employee sees a reason to prepare.
The Board Room · Leadership desk
What happened
- The Irrational Labs survey covered 150 employees, half of whom said they work in the technology industry.
- Over three years, 75% expected colleagues' jobs to be affected by AI, against 32% who feared for their own.
- The authors attribute the gap to optimism bias and to the better-than-average effect, the belief that one is harder to replace than one's peers.
Compiled by The Board RoomSomething wrong?How this is made
Why it matters
- decision A manager sizing an AI reskilling budget cannot count team-wide worry as proof that individual employees see a personal reason to change how they work.
- constraint Messages that frame AI change as a near-term threat to jobs meet the most self-exemption, because workers accept personal risk far more readily over three years than over one.
- constraint The evidence is a 150-person survey of expectations, so the claim that complacent staff are failing to prepare stays an inference until a team checks it against its own training records.
The gap is widest over one year and narrows as the horizon moves out. Within a year, respondents were 4.5 times more likely to predict job loss for peers than for themselves [3]. Over three years the ratio falls to 2.3 times, with 75% expecting peers' jobs to be affected and 32% fearing for their own [4]. Between the two horizons, expectation of personal loss rose 17 points while expectation for peers rose 8 [1].
Outside data supports the three-year figure. Pew surveyed 5,273 people and found about a third expect AI to reduce their own long-term job opportunities, with about as many saying it will not make much difference [5]. The one-year figure has a smaller base: 150 employees, half of them in technology [1]. The 15% is therefore about 22 people [3].
Irrational Labs, which studies how emotions turn into beliefs and behaviors [13], puts the gap down to two biases. Optimism bias assigns bad outcomes to someone else. The better-than-average effect supplies the reason: people believe they are more capable than their colleagues and so harder to replace [6]. In a 1981 study that has since been replicated, 93% of American participants rated their driving above the median [7]. "AI may take jobs. Just probably not ours," the authors wrote [8].
The board-deck version of this finding says staff are anxious about AI and will therefore welcome help adapting. The survey confirms the anxiety. It also shows that most of that anxiety is about colleagues. "Too little concern gives us no reason to adapt," the authors wrote [9]. They cite the Yerkes-Dodson law, which holds that performance improves as arousal rises, but only up to a point [10].
A skeptic would say beliefs are cheap and that a 150-person survey is thin. That objection is fair. The survey asked people what they expect, and it did not track whether anyone is retraining. The authors compare workers to people who say they fear California earthquakes and still buy a house near a fault line [12]. That comparison is about behavior, and their data does not test behavior. We do not know yet whether low personal expectation leads to inaction on any particular team.
This quarter, the practical choice is the timeline a change program uses. A reskilling push framed around losing jobs within a year asks staff to accept something most respondents rejected [2]. A three-year framing starts where about a third of workers already are [4][5]. The authors dismiss the blunt option, an all-caps email reading "USE AI MORE OR YOUR JOB MAY BE TAKEN", as "obviously not going to be effective" [11].
What to watch
- Behavioral data, such as training enrollment, broken down by how much personal job risk workers expect, which would test whether self-exemption turns into inaction.
- A split of the Irrational Labs results between tech and non-tech respondents, since half of the 150 said they work in technology.
- A larger sample asking the one-year own-versus-peer question, since the 15% figure rests on about 22 people.