AI is changing recruitment, and much of that change is welcome. It can map a talent market in hours, surface candidates no one would have found by hand, and remove days of administrative work from a search. Used well, data makes hiring faster, fairer and better informed.
But there is a growing assumption that the recruiter is next to be automated. We disagree, and not out of self-interest. Hiring is a decision made by people, about people, under uncertainty. It depends on trust, timing and reading what someone means rather than what they say.
Data can tell you who could do the job. It takes a person to understand who should, and whether they will stay. That is why our approach at Leonid is data-powered, but human-led.
The clearest way to make the case is to look at the moments in a candidate's journey where a good outcome depends on human judgement. None of them can be handed to an algorithm.
1. Hearing what isn't said
An AI system processes words. A recruiter hears tone. When a candidate says "I'm fine" but sounds less certain than last week, the right response is to stop and ask what has changed. That single question can surface a concern that would otherwise derail a hire weeks later.
The same is true after an interview. "It went well" is the most common answer we hear.
The useful information is often in the pause before a question about the future line manager. A role can be the right career move on paper and still be wrong for the person taking it. Only someone who knows the candidate will notice the difference.
2. Finding the real reason someone would move
Much of our work is retained search for passive candidates: people who are performing well, content in their current roles and are not looking. When they consider a move, the first reason given may not be entirely accurate, for a multitude of reasons. To get to the bottom of it requires both time and trust.
Salary is often where the conversation starts. With time and trust, it can turn out to be a promotion that went elsewhere, a promise that was not kept or contribution that went unrecognised. No screening tool will uncover that, because candidates do not share it with a screening tool. They will only share it with someone they trust.
Ambition works the same way. A candidate may say they want the bigger title, when what they want is more flexibility and less pressure. Helping someone say what they actually want can change the whole brief, and it has a direct effect on retention.
3. Seeing potential the candidate can't see
Algorithms rank what is written on a CV. Candidates routinely undersell what they have done, describing a significant achievement as "just part of the job". A recruiter recognises the evidence of leadership, judgement or resilience behind it, and helps the candidate talk about it with confidence.
That matters most when confidence is fragile. A highly capable candidate may arrive at interview still knocked by a recent rejection. Spotting that, and preparing them for it, is often the difference between someone who performs to their potential and someone who misses their chance.
4. Knowing when to challenge and when to listen
When a strong candidate says "I'm not ready to move", the response depends on what that means. Sometimes it is a firm no that deserves respect. Sometimes it is uncertainty that a well-informed conversation can resolve. Judging which, and how hard to push, is a matter of relationship and experience.
The same judgement applies at offer stage. Nerves are normal when a big decision arrives. But sometimes unease points to something real that the candidate has noticed and cannot yet put into words. Helping them tell the difference, without steering them towards the convenient answer, is advisory work.
5. Protecting people when the news is bad
An automated rejection is efficient. It is also a missed opportunity. Being honest about why a candidate was unsuccessful, while protecting their confidence and giving them something specific to work on, builds a relationship that lasts beyond one search. Many of the people we speak to today will be candidates, or clients, again in future.
6. Noticing the wobble after "yes"
Acceptance is not the end of the process. When replies become shorter, or a candidate stops mentioning their start date, that could indicate an issue. A system logs the delay…but a recruiter picks up the phone and makes space for the conversation the candidate has been avoiding.
7. Putting the candidate's interests first
This is the clearest test of all. Sometimes, based on everything a candidate has said, the right advice is: "I don't think you should accept this."
That advice can cost a fee. An algorithm optimised for placements would never give it. A trusted advisor must, because a hire that is wrong for the candidate is ultimately wrong for the client too. Trust in recruitment becomes tangible when a candidate sees their recruiter is prepared to lose a placement for the right reasons.
Data-powered, human-led in practice
None of this is an argument against technology. It is an argument for using it in the right place. We use data and structure to remove friction, so our people can spend their time where judgement counts.
Our Magic Month methodology gives every search a clear, pre-defined timeframe and holds us accountable to it. LeonidLive video interviews give hiring managers an early view of a shortlisted candidate's communication style, saving hours of screening. Both make the process faster. Neither replaces the conversations described above. They create more time for them.
These moments are irreplaceable by technology. They are deeply important to the candidate. And they matter just as much to any client looking to hire the best talent, because the cost of a hire that does not last is far higher than the cost of a conversation.