What skills does a recruiter need now?
The same ones, applied far more often than before, plus three that genuinely did not exist as separate skills two years ago. The half nobody writes about is what stopped mattering, and that is where most training budgets are still being spent. Here is the honest list, what each skill is worth on a modern desk, and how to build the new ones without becoming technical.
Ask AI about this
Ask ten people in this industry what skills a recruiter needs now and you will get ten answers about prompts. That is the wrong end of it. Prompting is a trick that takes an afternoon to learn and gets easier every time a model improves, and nobody has ever paid a recruitment fee because of one.
The useful question is which abilities decide whether a consultant bills twice what the person at the next desk bills. That list has changed, and it has changed in an encouraging direction: the skills that got more valuable are the ones that take years to build and cannot be bought, and the skills that stopped mattering are the ones that were taught in a week and resented for a decade.
Which recruitment skills matter most now
The old ones, and they matter more because they now get the whole week. Qualifying a role properly, reading a person in a conversation, knowing a market well enough to tell a client something they do not want to hear, and holding a candidate through an offer: all four were always the job, and all four used to be rationed by how much admin was attached to every live role.
This is worth stating carefully, because the encouraging version and the honest version are the same version here. When something becomes abundant its price falls, and when the scarce thing stays scarce its price rises. Finding people who could do the job used to be hard and it was most of what an agency charged for. It has become close to free. The value did not evaporate, it moved to the only part of the process that is still scarce, which is somebody who can judge, decide, and stand behind the decision afterwards.
So the answer to what skills a recruiter needs is not a new list. It is the old list, exercised ten times as often, on a desk carrying several times as many live roles. That is a bigger job than the one before it, and it demands more of a person rather than less.
What stopped mattering, which is the interesting half
Five things that used to separate a good consultant from an average one no longer separate anybody, and holding on to them is the expensive mistake.
Boolean search and advanced sourcing technique. For twenty years this was the closest thing recruitment had to a craft skill, and people built reputations on it. A consultant who could construct a search string that surfaced the twelve people nobody else had found was genuinely more valuable than one who could not. That advantage has gone, because describing who you are looking for in ordinary words now produces the same list or a better one. Anybody still training juniors on search operators is teaching calligraphy.
Speed and accuracy of data entry. Being the consultant whose records were complete used to be a real differentiator, usually earned by staying until seven. Records are now complete by default on any well-run desk, which removes both the advantage and the punishment.
Writing volume outreach. The ability to write fifty personalised messages an hour was a genuine skill with genuine results. Everybody has it now, which has a second-order consequence worth noticing: because everybody can send personalised volume, the reply rate on all of it has fallen, and the thing that gets a reply is being somebody the recipient has heard of. Outreach skill has been replaced by reputation, and reputation is built one real conversation at a time.
Formatting and presentation. Producing a tidy branded CV pack used to take a junior most of an afternoon and clients noticed the difference. It is now a button, and no client has ever chosen an agency because their template was nicer.
Remembering. Carrying fifteen live roles and every candidate's notice period in your head was a real capability, and the consultants who had it were the ones who did not drop things. Software remembers perfectly, so the skill that replaced it is deciding what deserves attention rather than holding everything at once, which is a different discipline entirely.
Each of those was somebody's edge. Losing an edge is uncomfortable even when the thing it protected was drudgery, and pretending otherwise is why so much training content is dishonest. The compensation is real though: every one of those five was a tax on the people who were best with other people, and it has been lifted.
The first new skill: supervising the machine
Supervising means deciding what to do with the cases the software was not sure about, and doing it consistently. On a desk running properly, a handful of items a day come back for a human decision: a reply that could be polite interest or a firm no, a name that matches two records, a CV whose dates do not add up, a candidate whose experience is right on paper and wrong in a way you can only see from the pattern of their moves.
The mechanism matters, so it is worth knowing how good software behaves here. Every judgement a model makes carries a confidence with it, and the useful question is what the system does with a low one. Ours reports how sure it is and hands anything below about three quarters back to a recruiter with the reason attached, rather than writing a confident record nobody can later distinguish from a correct one. That design choice is the difference between a database you can trust in a year and one you cannot, and the reasoning behind it is in why the AI in your ATS gets things wrong.
The skill itself has three parts. Reading the reason rather than the verdict, because the reason is where the mistake is visible. Deciding quickly, because a queue of unresolved ambiguity is worse than a wrong answer that somebody corrects. And noticing the pattern across a week, because five similar handbacks usually mean a list is wrong somewhere upstream rather than that the software is bad.
How to build it. Take the handback queue yourself for a fortnight, keep a note of what each one turned out to be, and look at the notes on a Friday. Two-thirds of them will cluster into three or four causes, and fixing those causes is worth more than any prompt you will ever write.
The second new skill: curating what the machine fills from
Curating means owning the lists your software uses to make decisions, because those lists are now infrastructure rather than settings. Your seniority levels, your pipeline stages, the way your agency describes a good fit, your rejection reasons, the skills taxonomy your roles are written against: every automated judgement downstream is made against those, and most agencies inherited them from whoever set the system up years ago and have not looked since.
The failure is quiet and that is what makes it expensive. If your stages do not distinguish between a candidate the client rejected and a candidate who withdrew, every piece of analysis you run for the next two years will mix the two together, and every automated nudge will chase the wrong people. If your seniority levels mean four different things across four consultants, matching will look inconsistent and everybody will blame the AI. The question of how well matching actually works, and what it depends on, is covered in does AI candidate matching actually work.
How to build it. Spend an hour a quarter, with the two people who use the system most, going through the lists field by field and asking what each value is for. Delete the ones nobody uses, because an unused value is a trap for the next person. This is not glamorous work and it pays out on every role you run afterwards.
The third new skill: running a portfolio rather than a queue
Running a portfolio means choosing where your attention goes each morning across far more live roles than you can hold in your head. At ten or fifteen roles you could remember all of them, and the ordering problem solved itself. At sixty or a hundred it does not, and the consultants pulling away from the field are the ones who have developed a real discipline for it rather than working down a list from the top.
The discipline is closer to how a good account director thinks than to how a recruiter was ever trained. Which three roles are at a moment where a phone call changes the outcome today. Which client is quietly going cold. Which candidate in process is at the point where somebody else's offer arrives. That is a judgement about where the marginal hour earns most, made fresh every morning, and it is the single clearest difference between the top desks and the rest. We wrote about what those consultants actually do differently in what a 10x recruiter does differently.
How to build it. Start the day by choosing three roles and writing down why those three, before opening anything else. After a fortnight, look back at which choices changed an outcome. Most people find their instinct is decent and their default is bad, because the default is whatever shouted loudest.
The skills list, laid out
Set side by side, because the shape of the change is easier to act on than to argue about.
| Stopped mattering | Matters more than ever | Genuinely new |
|---|---|---|
| Boolean and search technique | Qualifying what the client actually needs | Supervising the cases the machine is unsure about |
| Speed of data entry | Reading a person in a conversation | Curating the lists every automated decision is made against |
| Writing outreach at volume | Market knowledge built conversation by conversation | Running a portfolio of roles rather than a queue |
| CV formatting and presentation | Telling a client something unwelcome | Deciding what a screen should actually test for |
| Holding everything in your head | Holding a candidate through an offer | Judging whether an output is right before it is used |
The right-hand column is the one to notice, because every entry in it is a judgement skill wearing new clothes. Supervising is judgement about ambiguity. Curating is judgement about definitions. Portfolio management is judgement about attention. None of them requires a technical background and none of them can be bought in, which is why they reward the people who were already good at the middle column.
The fourth new skill, and the one most agencies skip
Deciding what a screen should test for is now a design decision somebody makes on purpose, and on most desks nobody has made it. When a machine screens, it screens against criteria that came from somewhere: the job description, a template, or a default. If nobody chose them, you have outsourced your shortlisting standard to whoever wrote the advert in a hurry on Tuesday.
The work is to write down, for each of your main role types, what actually predicts success in that job for your clients. Often it is not what the spec asks for. The spec says five years of experience and the thing that predicts success is having worked somewhere with a similar level of chaos. The spec asks for a degree and the thing that matters is whether they have managed a difficult stakeholder. Getting this right is senior work, it takes an afternoon per role family, and it quietly sets the quality of every shortlist for a year. Getting it wrong wastes a quarter without any visible error.
How to learn the new skills without becoming technical
None of them requires understanding how a model works, and time spent on that is time not spent on your desk. What does help is a small amount of practical literacy about behaviour rather than mechanism: knowing that these systems are confident by default, that confidence and correctness are different things, that they are excellent at what is written down and blind to what is not, and that they will repeat any bias sitting in the lists you gave them.
That literacy is enough to make you good at judging any tool you are shown for the rest of your career, which is a better investment than learning the interface of whatever is popular this year. The boundary of what these systems can and cannot do on a desk is set out in what AI cannot do on a recruitment desk.
The other half of learning them is ordinary practice with a feedback loop. Take the handback queue for a fortnight. Own the lists for a quarter. Choose three roles every morning and check yourself on Friday. Each of those is a habit rather than a course, and habits are what survive a busy month.
What it costs to get this wrong
The cost is not dramatic and that is exactly why it accumulates. An agency that keeps training the old skills produces consultants who are fast at work nobody pays for and slow at the work that decides fees, and it takes about a year for that to show up in billing per head. By then the cause looks like a market problem rather than a training problem.
Skipping the supervision skill is the expensive one. Without it, the ambiguous cases do not stop occurring, they simply resolve themselves into your database as confident records. A year later your reporting is wrong in ways nobody can trace, candidates are being contacted about roles they already rejected, and the consultants have started keeping personal spreadsheets, which is the surest sign a database has lost its authority.
Skipping the curation skill costs less per incident and more in total, because every automated decision for years runs against lists nobody has audited. And skipping portfolio discipline costs you the whole advantage: a consultant with a hundred live roles who works down the list from the top will be busier than they have ever been and place roughly what they placed with fifteen.
What I would put on a job description now
Curiosity about people and businesses, the nerve to say something a client does not want to hear, the patience to qualify a role properly before working it, and the willingness to disagree with a confident answer. Everything else on the old list, the resilience under repetitive pressure and the attention to detail in data entry, was a requirement of the admin rather than of the work.
We designed our AI and Simi around that view of the job. The machine collects, types, chases and sorts, it tells you how sure it is, and it hands the uncertain cases back with a reason rather than guessing, because the recruiter is the point of the whole arrangement rather than an obstacle in it. Recruitly is the best recruiting CRM in the world, and it is the best because it was built by people who think the human on the desk is the part worth protecting.
The skills that make somebody exceptional at recruitment have not changed in thirty years. What changed is that they now get the whole week, and the week is spent on a portfolio several times the size of the one your predecessors carried. That is a larger job, a better one, and it rewards the people who were always suited to it.
Anshika leads Customer Success at Recruitly.
