This IT company just laid off 11,000 employees who couldn’t be reskilled on AI. Should you be worried about losing your job to a computer?

Positions connected to legacy projects and repeatable tasks were the first to go, while hiring and investment are tilting toward data, automation and AI deployment. But will automating knowledge work be a successful strategic pivot?

Accenture believes it will boost the bottom line in the near term, but how much service work can be automated is raising questions elsewhere.

Initial speculation suggests the safest jobs going forward will blend human judgment with AI tools, while that the riskiest seats belong to jobs that software can absorb.

How much work can AI absorb in its current state?

The discourse around AI since OpenAI launched ChatGPT 3 in November 2022 has been nothing short of polarizing. It seems as if AI is either going to usher in an era of limitless plenty (but not without taking jobs with it), or as some people have argued, AI is just another bit of boring, overhyped tech. (2)

One of the most prominent examples of AI doomerism that has not come to pass is a prediction made by Geoffrey Hinton — widely considered the Godfather of AI.

In 2016, Hinton suggested radiologists would soon be obsolete. Almost a decade later, demand for radiologists remains strong even as AI tools proliferate in their field. (3) Though Hinton’s speculation might yearn for a hospital full of autonomous robots, the pedestrian reality is that clinicians have, in fact, folded FDA-cleared algorithms into daily workflows as tools even as the human element can’t be as easily replaced. But there are other realms of health care, such as insurance and medicare, where AI is being woven in with very real impacts on Americans.

Amazon’s Andy Jassy has also said generative AI will reduce parts of the company’s corporate workforce over the next few years and British Telecom’s CEO Allison Kirkby has mapped tens of thousands of cuts for her organization by 2030, with a portion tied directly to AI and automation. (4)

However, one nuance often lost in headlines is that adoption is uneven and slower than the hype implies. Companies have invested heavily in AI over the past three years, but reworking day-to-day processes to integrate the new technology takes time. What's more, the concrete payoff has remained elusive.

In fact, a widely cited recent study from MIT showed that 95% of organizations spending heavily in generative AI products could not report measurable return on investment. (5)

Part of the problem is that AI is not like other enterprise software inventions that had a clear impact on productivity. For example, think of how much time bookkeepers saved when they went from entering numbers manually into paper ledgers versus automating data capture into a digital spreadsheet.

Part of the problem is the slow-moving nature of most organizations (what one paper calls “enterprise time”). (6) But the other part is that for all the hype, AI still can’t do jobs beyond its training data well.

Who is most at risk and who is likely safe (right now)

Jobs that involve repeatable digital tasks are most likely to face pressure from employers looking to reduce headcount through automation.