In the world of recruitment and selection, a kind of arms race is currently under way between companies looking for suitable staff and jobseekers looking for a suitable job (or assignment). Both on the demand side of the labour market (companies, employers, recruiters) and on the supply side (whether actively seeking work or not), many are now being assisted by A.I., or more precisely: by Large Language Models, which have nowadays become the most important manifestation of A.I.
‘The recruitment and selection sector itself is enthusiastically joining in with all that A.I. activity.’
The recruitment and selection sector is enthusiastically participating in this. They advise companies to use A.I. to improve their job advertisements, and automate their own matching processes with A.I.. Jobseekers, in turn, scrape vacancies from the internet almost entirely automatically, have them assessed for their suitability, create a targeted covering letter and CV, and submit their application via A.I. For those less adept at doing this themselves, there is a forest of A.I. firms offering their services for this purpose. Companies are faced with an enormous number of responses to vacancies and are therefore increasingly using A.I. for the initial selection too. Various companies, incidentally, also offer their services for that.
Does it make things better?
The question is: does this ultimately make things better? And the question behind that is: what exactly do we want to improve? From a labour economics perspective, it is about ‘more’, ‘better’ and ‘faster’ in creating ‘productive matches’. In plain English: more people end up in a better workplace for them more quickly. And more companies find staff better suited to them more quickly. That is what it is all about, after all: the recruitment and selection process is designed to help solve, or at least alleviate, the information problems in the labour market.
And do not forget, all that A.I. force also consumes vast amounts of energy…
But it is highly questionable whether all that A.I. force is currently really contributing to that. And do not forget, all that A.I. force also consumes vast amounts of energy, which, despite efforts in the energy transition, is still not available without limit and causes climate damage (because it is generated using fossil fuels). Is the energy we put into A.I. for recruitment and selection, literally, well directed? I fear not, but the attentive reader will have understood that by now. I believe the same thing is happening here as in earlier waves of automation involving work centred on ‘interaction between people’.
Progress?
People mainly like to use technology to engage in more ‘interactions’, but that does not necessarily improve the quality of those interactions. Sixty years ago, academics wrote letters to one another and corresponded in that way with editors of academic journals and their fellow researchers. The arrival of word processing on PCs and email made that process enormously ‘efficient’. Letters no longer had to be dictated to a secretary (m/f/x) and then sent via the typing pool and post room. Much of this communication work has therefore ‘shifted’ to the academic themselves.
‘Posting a vacancy and responding to it is becoming increasingly easy. But…’
And because it became so much easier, they also started doing it more. In my view, the result has been that academics have indeed published ever more articles and have far more interactions with colleagues, but this has not necessarily accelerated progress in science. Something similar has also been happening, for years, in the world of recruitment and selection. With all the ICT solutions, it is becoming ever easier to publish a vacancy. From the labour supply side, it is also becoming ever easier to search for vacancies and respond to them.
Seeing the wood for the trees
But jobseekers still cannot see the wood for the trees, because more trees are constantly being added and it is not always clear whether they are genuinely new trees. And more importantly: once you, as a jobseeker, have found a suitable ‘tree’, you then convince the person who posted the tree of your own suitability for that vacancy in an automated way. In doing so, you mainly, and perhaps only, show that you are very good at the game of applying for jobs, of recruitment and selection. What you do not show, and what is also very difficult to show, is whether you are very good at the work that needs doing in that vacancy.
A.I. should also contribute to the quality of interaction, not just the quantity.
So all that A.I. and that ICT-driven arms race does lead to many more ‘interactions’, but those interactions still do not convey the information needed to arrive at a ‘productive match’. If A.I. is to count as a solution at all, then A.I. should contribute to the quality of interaction between people, not just the quantity. Disclaimer: this piece was written without A.I. (assistance). Some of the pessimism about A.I. in it certainly stems from the author’s own search for a new position, and the frustrations associated with it.
About the author

Also read
- Ronald Dekker: ‘Can the information problem in the labour market actually be solved?’
- Is Thomas Piketty the new John Maynard Keynes (and will we soon be working much less)?
- Ronald Dekker on ‘practical thinking’ and ‘impractical doing’
- How the 4-day working week can also be a magnet for councils