The last skill AI can't replace is the one nobody teaches

A million young people in the UK are not in employment, education or training. The Guardian reported this week on some of them: a former pub manager in Warrington who has applied for two thousand jobs without a single offer; an Oxford languages graduate who can't get past the first round; a 20-year-old in Peterborough sending out forty to fifty applications a month and hearing almost nothing back.

The easy explanation is that the economy is hard right now. That's true, but it's not the whole story.

The jobs disappearing fastest are the ones that were always task-based. Writing, data entry, basic research, content production, first-draft code. A digital marketing agency owner quoted in the same article was direct about it: he used to take on apprentices every year. Since adopting AI eighteen months ago, he hasn't hired one. The tasks those apprentices would have done (the boring, laborious work that was also how you learned) are now handled by software.

This is not a temporary blip. A 2025 McKinsey survey found that 51 percent of organisations reported generative AI was reducing their need for entry-level workers. Early-career workers in AI-exposed fields saw a 16 percent relative decline in employment between 2023 and 2025, while more experienced workers remained largely unaffected.

The entry-level layer is thinning. And it was always the way in.

What's actually growing

The same research that documents this displacement points consistently in one direction. McKinsey's European labour market modelling projects that demand for social and emotional skills will rise by 11 percent by 2030, while demand for roles where basic cognitive skills predominate is expected to fall by 14 percent. A 2024 study published in Nature Human Behaviour, analysing over 12 million job postings, found accelerated demand for interpersonal skills in the post-pandemic labour market. A separate labour economics paper found that occupations requiring greater interpersonal skills are meaningfully less exposed to automation than those built around task completion.

The pattern is consistent across sources. What AI does well is tasks with defined parameters. What it cannot do yet, is navigate the human complexity of a room. The persuasion that happens in real time. The reading of what someone actually needs, as opposed to what they said they needed. The authority that makes a person worth listening to when the stakes are high and the audience is sceptical.

These are not soft skills in the dismissive sense of that phrase. They are, increasingly, the hard economic edge.

The network problem

There is a second issue running alongside the AI one, and they interact in ways that make the picture worse for young people with less access to both.

The research on how jobs are actually filled is striking. Referred candidates account for roughly 7 percent of all applicants in the UK but between 30 and 50 percent of all hires.

A referred candidate is approximately four times more likely to receive an offer than someone applying through a job board. UK graduate vacancies dropped 8 percent in 2024/25 while average applications per vacancy reached 140.

What that means practically: if you are applying into the formal system, you are competing for a shrinking share of roles against a growing field of candidates. The jobs that don't appear on job boards: the roles filled through a conversation at an industry event, through a former colleague, through someone who saw a piece of writing and reached out, go to the people who are already in the right rooms.

Getting into those rooms requires exactly the skills the research says are growing in value. Presence. The ability to hold a conversation with someone you've just met and leave them wanting to continue it. The confidence to be visible in a professional context rather than retreating into a screen.

The inequality underneath

This is where the argument becomes uncomfortable, because those skills are not equally distributed.

A peer-reviewed systematic review of UK undergraduates from disadvantaged socioeconomic backgrounds found that employers consistently exclude talented candidates from working-class backgrounds through recruitment practices that reward social and cultural capital rather than objective ability. A 2026 report from the Association for Business Psychology found that communication is routinely assessed in ways that conflate it with class markers (vocabulary, presentation style, professional fluency) rather than actual communicative effectiveness. The Royal College of Speech and Language Therapists found that 88 percent of long-term unemployed young men have long-term speech, language and communication needs, with direct causal links to social disadvantage.

The child who grows up in an environment where confident self-expression is modelled, encouraged and rewarded arrives at the job market with advantages that have nothing to do with intelligence or work ethic. The child who doesn't has to find another way to close that gap.

The system, as currently designed, mostly doesn't help with this. The universal credit job coach telling an Oxford graduate to stop applying for her dream job and try a toy shop instead is not the exception. She is the system operating as designed.

What this suggests

I am not arguing that communication skills are a silver bullet for youth unemployment, or that the problem is simply one of personal development. The structural issues are concerning: AI is displacing entry-level work, employer costs are rising, transport and housing lock young people out of opportunity geographically, and the jobs that remain are increasingly concentrated in places and networks that are hard to access without already being inside them.

But I do think there is something important in the observation that the skills which are becoming harder to automate are also the ones that have always been most unevenly available. The ability to be present in a room. To speak with authority rather than anxiety. To build the kind of professional relationship that results in someone thinking of you when something comes up.

This was often assumed to be innate, something you either had or didn't. The evidence from actor training, and from the work I've spent twenty years doing with professional actors and with executives and lawyers who face the same fundamental challenge, is that it isn't. It's trainable. But training it requires getting to the physical reality of how someone is in a room, not the theory of how they should be.

The young people in that Guardian piece are not failing because they lack intelligence, qualifications or effort. They are failing because the thing the labour market is increasingly selecting for is something they were never given the chance to develop, and the current system has no mechanism for giving it to them now.

Luke Marquez is co-director of Rebellion Presence, where he trains executives and legal professionals in presence and communication using professional actor training technique.