Teach First is a UK charity that drops graduates into education jobs. It’s generally considered to be a good organisation, one that I know from experience has helped lots of smart people build a career in teaching. The group has historically selected candidates based on written assessments, but earlier this month it said it was accelerating a plan to switch towards face-to-face interviews.
The reason? University graduates are using AI in applications, which predominantly take the form of written assignments where anyone can put ChatGPT to work. Patrick Dempsey from Teach First said the charity had seen around a 30% increase in applications so far this year on the same period in 2024, a development he primarily attributes to large language models:
“The shift from written assessment to task-based assessment is something we feel the need to accelerate…there are instances where people are leaving the tail end of a ChatGPT message in an application answer, and of course they get rejected.”
Other accounts paint a similar picture, with graduate employment specialist Bright Network reporting that the number of people using AI for job applications has risen from 38% last year to 50% this year. And why wouldn’t they? Applications are tedious at the best of times, never mind when the odds of success are long.
If everyone can write well enough to pass a hiring round or two, it follows that employers will change tack to focus on assessments where LLMs aren’t much use. While that might seem straightforward enough, we ought to remember that in-person tests are not exactly a direct replacement.
Those running the hiring process are no doubt aware that teachers are encouraged to use AI in the classroom, so in-person interviews already proceed on the basis that successful candidates will use LLMs when they get the job.
A pivot to face-to-face tests might simply ignore that reality, or it might just reflect the lack of compelling alternatives. But whether or not the move aims to approximate written tests via in-person assessment doesn’t really matter. After all, the nature of face-to-face tests means they also probe for verbal and social competencies.
That’s because the people who succeed will be the ones who interview best, who get on well with those asking the questions in a way that makes them think they’ll be suited to the job. I think about this as the vibes premium: the increase in value placed on subjective traits — charisma, manner, confidence, aesthetic, speech, and presence — as it becomes tougher to use technical measures to distinguish between candidates.
The idea is related to but different from ‘interpersonal skills’ that are also likely to appreciate in value in the age of AI. However, our vibes premium is less ‘problem-solving’ or ‘leadership’ and more simply ‘making people want me around’. There are lots of ways to make that happen, but they generally deal with personality traits and behaviours even softer than these ‘soft skills’.
Applying for everything
It’s rough out there for graduates. Clearly good jobs have always been difficult to land, but I suspect competition has been stiffening for quite some time as barriers to entry have collapsed. More people apply to graduate opportunities thanks to an expansion in higher education, online application portals, social media job openings, and international mobility. Now candidates can use AI tools that let anyone sound polished, which in turn raises the standards of applications and makes it harder for any one person to land a job.
On the other side of the line, a few friends have told me they no longer plan to hire because their respective leadership teams are convinced AI can do the tasks that they might want a junior person to take on (though it is worth saying the picture is more complicated than that).
Whatever the case, a recent survey reported that vacancies for graduate jobs, apprenticeships, internships and junior jobs with no degree requirement dropped by 32% since the launch of ChatGPT in November 2022. The same poll says that entry level jobs now account for 25% of the market in the UK, down from 28.9% in 2022.
We have two forces at play, both of which may be driven at least in part by the emergence of powerful AI systems:
Graduate job availability is falling. This might be because managers are automating the work or doing it themselves with AI, though it is also possible there are other labour market effects at play that are responsible.
Graduates are using AI in applications, so writing samples converge around the same level of quality. This compression may happen elsewhere, but it’s especially influential for entry level roles where differentiating factors are harder to come by.
In some ways the emergence of powerful AI systems is deeply humanistic. In entertainment, I expect good taste and human-led curation to become more valuable in a world overflowing with slop. I also imagine the same is true for the workplace, where technical skills become less relevant versus the qualities that only humans possess.
To be clear, my view is that AI raises the floor for technical skills but doesn’t necessarily eliminate the ceiling. A sharp candidate who knows the domain, has original ideas, and uses LLMs well can still outpace someone who just pastes in a prompt. The best performing candidates might even find themselves slowed down by the technology.
Even so, we’re interested in entry level jobs and graduate roles. These are the places where more people now get passable technical skills than ever before, which means employers need to look elsewhere to distinguish between candidates.
Regardless of how good the models get in the near future, I don’t see them fully displacing humans across the board. Part of that is about allowing employers to maintain a clear locus of responsibility at work, so if an AI makes a mistake firms can point to someone holding the bag.
But it’s also because there are some roles where we want a human in the loop, even if we don’t strictly need one. People go to a doctor for reassurance as well as care. They want empathy, explanation, and the sense that someone is emotionally attuned to their issue. The same goes for teachers, therapists, lawyers, and countless other jobs whose value partly flows from the human touch.
Alas, that is probably little comfort to graduates finding it hard to land a job. Even if today’s impact is overblown, young people will be those who experience the economic impact of LLMs first because it’s easier for the technology to help low skilled people become average rather than the excellent to become brilliant.
This sounds worrisome, but it does mean that they are likely to be the first to respond to the new demands of employers. Things that centre the human, things that let them play well with others, and things that mean their bosses want them around.
New signals
That leaves us in an interesting place, one where technical proficiencies like writing are no longer reliable indicators of skill or effort. Firms of all stripes are hungry for new kinds of signals that they can use to sort people, and so we get a hard pivot to face-to-face interviews.
This is a type of assessment less interested in output than in character, a trend that is likely to become more popular as firms look specifically for traits that only human beings can provide. It’s a manifestation of a kind of neo-humanism premised on the idea that the more technically proficient machine outputs become, the more we value the ineffable.
Neo-humanism returns hiring to something closer to the old apprenticeship model, where being able to work alongside someone mattered more than being highly credentialed. And at the other end of the spectrum, we should remember that elite institutions like Oxbridge never stopped interviewing for some courses.
There's something appealing about jobs going to people who can actually collaborate and communicate rather than those who simply know their way around a marking scheme. But a world where success depends on being likeable in a fifteen minute conversation is going to be a brutal one for people to navigate, where the premium on hard work may not be what it once was.
Maybe the future pans out differently, though I wouldn’t be so sure. As technical skills become table stakes, knowledge work will be increasingly defined by the stuff that makes us human. Put another way: given enough time, AI will make personality hires of us all.
I found this compelling and insightful thanks!
These hiring needs seem concerningly at odds with the apparent rise anxiety amongst graduate-aged people; I expect anxiety would be non-conducive to having agency and communicating solidly. I think education will need a fast re-orientation towards teaching agency as a skill, much less faith put in passing tests and being OK
Great post! I do, however believe that these soft skills have always been important to the hiring decision. This may be optimistic but I do think, entry level roles will (a) transform for higher levels of complexity and (b) become democratized as the risk to hire a candidate who ins’t there yet ‘technically’ dramatically falls, allowing companies to screen for potential ceiling rather than a minimum floor.