How companies need to rethink the training of young people using AI

Until recently, young people fresh out of secondary school or university would begin their careers as apprentices. Then the roles became more appealing and were described using English terms such as intern or trainee. After completing a proper learning programme – often involving ‘trial and error’ – they were promoted as soon as they overcame the initial learning hurdles and achieved their first results, demonstrating the progress they had made. This was generally a lengthy process, which could (and still can) last between 12 and 18 months.

They were working their way up the ladder. We’ve all done it in one way or another. I remember that in my day I had a rather privileged apprenticeship, shadowing a young researcher at the University of Naples – now a professor in Rome – who supervised me for months until I learnt to do what he asked of me. And at that time, the use of PCs was widespread, but not yet very advanced. It was an important step in my development, one I remember perfectly, especially as I was an Economics student who wanted to study Total Quality Management in Mechanical Engineering. If I hadn’t taken that step, I wouldn’t have learnt many of the things I now know and teach to young people straight away.

What has changed? With the arrival of Gen AI, which has the fairly clear task of generating content by accessing millions of pieces of information, the basic work – that is, the simplest tasks where making a mistake has no consequences – is beginning to be delegated to artificial intelligence. Its role is to provide access to an enormous amount of information that a human might be able to evaluate, but only after gaining access to the data (assuming they are able to find it) and after investing a huge number of working hours.

Until two years ago, the task of reading through a contract and requesting amendments was left to the latest person to join the team, who would make the necessary notes and then, following supervision, pass it on to the lawyer. Now, an LLM can carry out this task in a matter of minutes and cross-reference it with internal material if a RAG (Retrieval-Augmented Generation) system is in place. In just a few minutes, the system is able to highlight the parts requiring human supervision and refer the task back to the manager to check for hallucinations or bias.

More and more often, I hear clients tell me that the team now consists of two senior staff, two offshore staff, and either Claude or ChatGPT, which are now regarded as the fifth member of the team. The so-called interns or apprentices, on the other hand, have completely disappeared.