Selion Perspective
Read in Translation
For everyone on either side of a senior appointment, now that every application reads well.
A well-made application once told its reader something about the person who wrote it. A clear letter suggested a clear thinker. A CV that answered the specification point by point looked like the work of someone who had read it closely and recognised themselves in it.
That reading has stopped working. In advertised hiring, the specification now tends to come back almost word for word, every requirement met, each phrase returned in roughly the order it was sent. The documents are fluent and well organised, and hard to tell apart.
Somewhere behind them is a person. The paper says less and less about who that person is.
AI belongs in working life now, and in the writing of applications too. Used well, it sharpens a person’s own draft and brings order to their thinking, and candidates use it as anyone would. Once the tool does the writing itself, fluency is a given and stops telling the reader very much, and the work of judgement moves somewhere else. For anyone who reads careers for a living, that is where this year’s more interesting questions have gathered.
What changed on the page
The clearest evidence so far comes from the Netherlands. Researchers at Tilburg University asked students in their final semester to write cover letters for real entry-level roles at Philips, PwC, Rabobank and VodafoneZiggo, and gave half of them access to ChatGPT. Recruiters from the four companies scored the letters without knowing which had been written with help. The tool improved them, the weaker ones most of all. The gains sat in the introductions and closings, the most standard parts of any application letter. What the recruiters cared about most was why the applicant wanted the job and how clearly the letter was written, and on both counts the tool made no measurable difference. Invitations to interview did not rise. Applicants had asked the tool for help with their motivation again and again.
A second experiment, with some four hundred recruiters, found that they could not tell which letters had been written with AI. They did no better than chance. Told which of the strong letters had been written unaided, they valued those letters more. The study has since been accepted by the Journal of Labor Economics.
It looked at graduates, and a senior appointment rests on far more than a letter. The mechanism is the same wherever a page stands in for a person. A specification mirrored back says less about who did the mirroring, and an application’s English may no longer be the applicant’s.
Where the weight moves
Retained executive search has always run in a different order. The first contact is a conversation, and the career record follows it, one piece of evidence among several. That sequence counts for more than it used to. The weight of a decision moves to the part no tool can sit through on a candidate’s behalf: the conversation, allowed to go somewhere unplanned, held more than once and in more than one setting.
European law has reached a related view by its own route. The EU’s AI Act classifies systems used to recruit and select people as high-risk, with requirements for human oversight, and those obligations apply from December 2027.
Judgement that tests itself
Judgement has weak points of its own.
Every assessment is a snapshot, a Momentaufnahme in the German sense: one person, on one day, in one setting, seen by a reader with a history of their own.
It is only as good as that reader’s ability to set their own leanings aside. Psychometric assessment has earned its place in senior selection, and its findings become most useful when read against a career and against what surfaces over several conversations.
Interviews deserve the same humility. A major re-examination of selection research, co-written in 2022 by a professor at Singapore Management University, found that the familiar methods predict performance less well than had long been believed. Structured interviews came out on top, and structure matters because first impressions are persuasive. Structure establishes what needs to be examined consistently. It need not prevent an experienced interviewer from following an answer into territory that was not anticipated.
In a field study of employment interviews, interviewers who had formed a favourable view of a candidate beforehand spent more of the conversation selling the company and less of it finding things out. Another study found that the more experienced and more confident interviewers reached their decisions faster, which is reason enough for experience to keep checking itself. Familiarity plays its part as well. A candidate who seems like someone the interviewer already knows tends to be read more generously, or less.
Judgement earns trust when it tests itself.
Before settling on a view of a candidate, the strongest assessors gather the evidence and treat their own first impression as a working hypothesis, most of all when it is a favourable one. Where several people assess the same candidate, each forms a view alone and sets it down before the discussion begins.
Intuition belongs here too, and its place is at the end. In the mid-1950s a young lieutenant named Daniel Kahneman, who would later win the Nobel Prize in economics, was asked to set up the Israeli army’s interviews for combat units. He had the interviewers rate a handful of traits one at a time, and they disliked it. As a compromise he let them close their eyes when they had finished and give an overall judgement. That last intuition proved as valid as the ratings, and added something of its own. Decades later he still gave the same advice: judge the parts first, and let intuition speak only after that.
Listening
An account polished for the page is also a defended one, and defences come down slowly. What lowers them is attention. In a series of experiments, researchers in Israel found that people who were listened to closely and without judgement grew less anxious about how they were coming across, and began to reflect more openly: more aware of the contradictions in their own views, and more willing to admit them. The studies dealt with opinions, and the same thing happens across an interview table.
Listening of that kind is slower than it looks.
It leaves a silence alone long enough for the second answer to follow, the one answer that was not prepared, and it follows what was actually said, wherever that leads, while the list of questions waits. Then the account gains texture. Someone admits they do not know. The people who worked for them stop being “my team” and become individuals, each described more particular. A detail surfaces that was not mentioned earlier. The listening protects the assessor too, because the first impression keeps meeting new evidence while the conversation goes on.
The translation
What a strong application sounds like depends on where its author learned to present themselves. Some conventions state achievements directly and claim them as one’s own. In others the same achievements are framed through the team or the circumstances, and a reader raised on directness may take understatement for a lack of substance. Each language has its own rules for what may be said outright and what is left for the listener to hear.
A shortlist for a regional role brings several such conventions into one process. The consultant reads each candidate from a convention of their own, then translates them for a client who reads from another. Before it reaches the CHRO or CEO, a shortlist often passes through more hands, a regional HR lead, say, or the line manager who will work with the appointee. Each is another reading, open to the first impressions and the pull of familiarity described earlier, and each is a point where something of the person can come through intact, or go missing. The strongest processes make sure the fullest reading of a candidate travels with them all the way to the people who decide.
The CHRO or CEO then translates once more, for the board and for the organisation that will receive the appointment.
This is where judgement is made or broken, and it is the part of the process that loses most when it is delegated.
Writing tools add a layer of their own. In an experiment at Cornell, writers from two countries completed short writing tasks with and without AI suggestions. With the suggestions, the non-Western writers’ prose moved toward Western styles, and some of its cultural texture fell away. The tools tend to pull toward one register, so the page can now arrive in a voice that is not quite its author’s. The differences have not gone anywhere. They return in the first conversation.
Some of those differences only surface in person. Where a role requires an executive to lead and negotiate in English, that ability needs to be understood early. It belongs in the first conversations, and long before a client meeting is arranged. A gap discovered at the introduction is costly for everyone and hardest on the candidate. Then there is everything the body says unasked. Silence runs longer in some places than in others. A pause before answering can look like hesitation to someone raised to answer fast. A courteous yes can be good manners and nothing more. Each is a convention, to be translated before it is judged.
The same translation happens inside organisations during annual talent and succession reviews. Managers present their people for promotion decisions to committees that read from conventions of their own, and a modest account of a strong year can be heard as a modest year. Committees that hear many such accounts learn to ask what was left unsaid.
Careful selection now looks patient from the outside. It forms several views, each on its own, and tests first impressions against the evidence. It takes the candidate through each translation to the people who decide, losing as little as it can on the way. In cross-border search this has always been the centre of the work.
The candidates with most to gain are those whose substance was never in the polish. On paper, they were always the easiest to miss.