Retraining workers for the AI world

Last month, JPMorgan Chase announced that all its new hires would receive training in the use of artificial intelligence — promising that such a move would help staff to eliminate repetitive “no joy work” and boost productivity and revenues.

“This year, everyone coming in here will have prompt engineering training to get them ready for the AI of the future,” said Mary Erdoes, chief executive of the US bank’s Asset and Wealth Management business.

She was referring to the process of writing the most effective text ‘prompts’, which are required to generate the desired response from an AI application. And Erdoes is not the only business leader to see this need. Prompt engineering is emerging as one of the in-demand skills in workplaces where AI is taking over and automating tasks, or acting as an aid to workers — for example, through chatbots such as OpenAI’s ChatGPT. 

In governments, globally, productivity improvements from generative AI could be worth $1.75tn annually by 2033, according to Marc Warner, co-founder and chief executive of Faculty, a London-based company providing AI software, consulting and services. 

But reaping these benefits will require careful investment in retraining for an AI world.

“It’s not AI that will replace humans, it’s humans who can work with it [who] will replace [those] who can’t,” says Khariton Matveev, a tech entrepreneur. His advice is: “View AI as a co-worker, don’t avoid it, but look more for cases of implementation in your field.”

According to one study, nearly a fifth of US workers could have at least 50 per cent of their tasks affected by the introduction of large language models, such as OpenAI’s GPT-4. Matveev believes this AI adoption will be more about “partially replacing some job elements than entirely replacing professions”. He suggests that information-processing jobs — such as translator, researcher, or designer — are at higher risk than professions requiring physical strength. 

In preparation for the new technology, he recommends that workers complete a course in prompt engineering, and try to experiment with AI tools by integrating them into their daily lives. 

Christian Rebernik, co-chief executive and co-founder of Tomorrow University of Applied Sciences, says there is a more “urgent need” to retrain for AI in certain sectors than in others. Sectors where the need is great include healthcare, climate change mitigation, and cyber security — as cybercriminals are increasingly using AI to carry out sophisticated attacks. 

There are also AI-specific roles that will become increasingly important as the technology is more deeply embedded in our everyday lives, such as data-labeller or annotator — people who can help to train AI algorithms by, for example, clarifying what a particular image depicts.

“The rise of Generative AI brings a number of opportunities for retraining — mapping to the AI pipeline that goes from data collection and labelling, to model creation and training, and finally application and feedback,” says Dev Nag, chief executive of QueryPal, an AI assistant for businesses. “Domain experts who can help at the beginning and end of this pipeline — [advising on] which data to collect, how to label it, and how to apply it — will continue to add enormous value.”

To those most at risk of having their roles replaced, but who remain tech savvy, Matveev says: “You can secure yourself, if you are a top-tier expert who can help teach AI in your field . . . [you can] earn from training models on data set created by you.” 

But, with all these changes, the skills required to stay ahead go beyond the purely technical. 

To be most effective as a prompt engineer in a particular sector, for example, will require deep knowledge of that sector. “Our experience so far suggests . . . that there is a strong correlation between subject matter expertise and the ability to create the best prompts,” explains James Longster, partner in the technology and commercial transactions department at UK law firm Travers Smith.

He cites the legal sector: if different staff members are given the task of using AI to extract information from a contract, the experienced lawyers will tend to outperform their non-legal counterparts in creating the prompts that extract the best results.

And all employees who work with clients will need to know how to wield AI in ways that retain their clients’ trust and confidence. Nag points to the enduring role of financial advisers who must “balance the risk versus reward judgment for individual investors, so that [AI] algorithms don’t push outside of the target risk envelope”.

Matveev agrees. “As AI takes on more “hard” components — like data, analysis, execution — the human role will shift to better understanding needs, what do clients really want and what should we do,” he says.

Others also mention the importance of interpersonal skills and ethics as AI is adopted in the workplace. “Skills such as emotional intelligence — recognising and regulating one’s emotions — and social intelligence — understanding and influencing the emotions of others in social situations — will ensure AI integration remains human-centred,” argues Rebernik.  

Above all, staff will need to be “flexible and adaptable”, says Stanford University professor and AI specialist Erik Brynjolfsson — as some skills may quickly become redundant as the technology moves so fast. 

“Prompt-engineering skills were hailed as the important new skill to learn,” he notes. “But they are already being eclipsed, as [large language models] learn to write better prompts than humans.”