About half of American jobs could be replaced by artificial intelligence, according to a report by the McKinsey Global Institute.

The American consultancy’s analysis found that robots and AI agents could automate more than half of US work hours, both mental and cognitive, using technology that is available today, if companies redesigned how they did things.

Most of the roles at risk — about 40 per cent — involve the kinds of drafting, processing information and routine reasoning that AI agents can do. Hiring is slowing in some such jobs, such as among paralegals, administrative and office support workers and programmers, the research found.

Similarly, dangerous, physical jobs, in warehouses or operating machines, are most likely to be replaced by robots, McKinsey said.

Conversely, a third of US jobs would be difficult to replace with AI because they have uniquely human attributes, such as nursing, the analysis found. Some 70 per cent of the tasks performed by carers and other healthcare workers require the kind of physical presence, empathy, care and dexterity that machines cannot replicate.

Building maintenance and repair work which demands flexibility, judgment and thinking on the job, often in unpredictable environments, is similarly unlikely to be automated.

The barrier to adoption came from policy choices and spending on implementation and development, McKinsey said. The paper, Agents, Robots and Us, emphasised that changing workflows, rather than tinkering with individual tasks, could generate $2.9 trillion a year in economic value by 2030.

Klarna CEO Sebastian Siemiatkowski (left) and company Chairman Michael Moritz (center) meet with specialist Peter Giacchi (right) before their IPO begins trading on the New York Stock Exchange.

Klarna, which floated in New York in September, has said that AI is helping it to grow its revenue without increasing the number of employees

RICHARD DREW/AP

Human skills will not be dispensed with as AI takes hold, the research found, but instead people will work in “partnership” with bots. Workers will change how and when they do things, spending less time on jobs such as preparing documents and more time interpreting results and telling AI systems what to do.

A third of American jobs in professions such as teaching will use AI in a hybrid way to help with work, generate content and deal with administrative burdens, leaving humans to interpret results, make judgments and guide decisions.

It is not all doom and gloom. Researchers argued that the rise of AI was creating a wave of new roles, from product managers who co-ordinate AI systems to people who ensure safety, because artificial intelligence still depended on humans to supervise, tune, test, and physically support the systems that make it run.

Over the past few years, companies and governments have been trying to harness AI to increase efficiency and productivity. Some organisations have started to report a positive impact on their bottom line and a need for fewer workers.

Trump’s big problem is not Epstein — it’s the AI bubble

Klarna, the recently listed fintech, has said it will continue to grow its revenue, while keeping its number of employees flat, because it is using AI in a variety of functions. Clifford Chance, the law firm, announced last week that it was cutting the number of its business services roles through the increased use of the technology, echoing a similar move by BT, the telecoms company.

A recent Stanford Report found that early-career workers aged between 22 and 25 in the most AI-exposed occupations have experienced a 13 per cent decline in employment, but more experienced workers in the same fields have not seen equivalent job losses.

The McKinsey report follows one from Microsoft last month which found that translators, sales representatives and financial advisers would be among the first victims of the AI revolution. Nurses, ship engineers, plasterers and water treatment operators were considered to be the safest.