AI is giving some graduates responsibilities once reserved for experienced staff. The dispute is whether this accelerates their apprenticeship or bypasses the work that teaches professional judgement.

Lotte Underwood, 25, is an AI associate at London PR agency Fight or Flight, whose dedicated team has built around 30 agents. She manages agents, checks outputs and joins client meetings to establish what would help. But she has yet to pitch directly to the media, a traditional PR skill she intends to acquire.

The case for this shift is substantial. Repetitive work is not automatically educational; automating it can let graduates test ideas and contribute sooner. At consultancy Valliance, Cian Clinton, 23, uses around 10 agents daily and says he handles responsibilities previously reached after four or five years. He argues that progression increasingly rewards critical thinking and understanding processes rather than accumulated knowledge.

Dillan Patel, 24, offers another answer to the learning objection. At engineering software company Stride, he began by developing a chatbot to capture a retiring colleague’s knowledge. Now an AI automation analyst, he says projects across the business are giving him valuable knowledge and experience. AI work need not mean isolation from the profession.

ArvatoConnect chief executive Debra Maxwell argues that broader graduate roles can build engagement and loyalty. One graduate uses AI for market research supporting contract bids. Yet the contact-centre outsourcing company has paused its usual graduate intake this year, planning an AI-focused return next year: redesigned opportunities do not necessarily mean uninterrupted access.

The strongest objection is that greater responsibility does not establish greater competence. Checking an agent’s briefing requires enough subject knowledge to recognise omissions and poor advice. PwC’s analysis of 2.4 million US entry-level jobs found those most influenced by AI were seven times more likely to demand traditionally senior skills. Roles with higher expectations grew 35% since 2019; other entry-level roles shrank 10%. That demonstrates changing demand, not successful training.

Portman Finance Group’s Simon Pratt offers a concrete alternative: a two-year programme with four six-month rotations alongside senior relationship managers, learning products, credit papers and customer relationships before running transactions. PwC’s Claire Reid likewise argues that baseline AI literacy must be coupled with sector knowledge and an understanding of clients’ needs.

Neither objection proves that traditional junior tasks must survive unchanged. Client contact, cross-business projects and supervised decisions can teach the trade differently. But removing old tasks requires replacing the learning they supplied. Employers promising accelerated progression must show how graduates acquire the judgement to challenge an agent, rather than merely the ability to operate it.