The Soft Skills Paradox
Why the skills that matter most, are the hardest to see.
Download the report (PDF)18-page PDF · about 25 minutes to read online
Contents
A person’s strongest and weakest soft skill sit 34 percentage points apart on average. No one in our data scored in the top tier on every skill at once.
Education, career history and tenure together explain about 8% of the variation in soft-skill scores. 92% is not on the page.
Average scores move less than two points between people in their twenties and people in their forties.
Soft skills barely correlate with each other (the strongest pair is 0.18), so an average can make two opposite candidates look identical.
Generalists average 72.2%, against 67.8% for specialists.
Decision-making is the least developed skill for more than one in five people.
It has the lowest median score (63%) and is the strongest skill for only about 3% of people.
Four assumptions hiring is built on, and where our data disagrees.
Almost every hiring decision made today rests on four beliefs so familiar that they are rarely examined. They feel less like assumptions than like common sense.
- The best candidate is well-rounded. The ideal soft-skill profile is strong across the board, with no obvious weakness.
- A good CV signals a good hire. Education, experience and history tell you who will perform.
- Soft skills grow automatically with experience. The more senior the person, the more developed their judgement, resilience and adaptability.
- A single score can rank one person above another. Reduce a candidate to one number and compare.
Drawn from more than two million skill observations, one of the largest datasets of its kind, the evidence contradicts every one of them. No one is well-rounded. The CV, the document hiring runs on, says little about soft skills. Experience can build these skills, yet in our data older candidates score no higher than younger ones. And the single number used to rank people can make two opposite candidates look identical.
This is the soft skills paradox. The capabilities that most determine whether someone will succeed at work are precisely the ones our hiring systems are worst at seeing. This report sets out the evidence for each of the four, explains why it matters for the people who hire and develop talent, and closes with a different question worth asking: if the old assumptions fail, what should replace them.
The purpose here is not to say hiring is broken. It is to show that hiring has been optimising the wrong signals, and that better ones are available to anyone willing to measure what the CV cannot.
The well-rounded candidate does not exist
Picture the candidate a hiring panel dreams of. Capable everywhere, weak nowhere, the safe pair of hands who ticks every box. This is the profile that survives a committee, because no one can object to it. It is also, on the evidence, a person who does not exist.
When we look at how people actually score across the full set of soft skills, a striking regularity appears. Everyone is uneven. Every capable person is strong in some areas and still developing in others, and the distance between their best and their least-developed skill is far larger than most hiring processes assume. On average, a person’s strongest and weakest skill sit 34 percentage points apart. The even, gap-free profile that many recruiters look for is not rare in our data. It is essentially absent.
Notice what the chart shows and does not show. At the population level the skills span a wide band, from teamwork near the bottom to the most common strengths at the top. But this is the aggregate. No single individual sits at the top of all of them. The population looks capable across the board precisely because different people are strong in different places, and that collective competence dissolves the moment it is applied to any one person.
So what does the ideal profile actually look like?
If the well-rounded candidate is a myth, the honest question is what replaces them. The old ideal was defined by subtraction: the best candidate was the one with the fewest weaknesses, the smoothest profile, the least to object to. It was an ideal of safety, and it quietly rewarded the people who revealed the least about themselves.
The ideal the evidence points to is defined by addition instead. The strongest profile is not the one without gaps, because no such profile exists. It is the one whose genuine strengths line up with what the role actually demands, and whose development areas fall where they matter least. A candidate who is exceptional at judgement and still growing in, say, project administration is not a compromise for a role built on decision-making. They are the ideal hire for it. The old model would have marked them down for the gap; the new one recognises that the gap is the ordinary cost of the strength, and irrelevant to the job at hand.
This is a quieter but far more useful definition of a great candidate. Not the person who is acceptable at everything, but the person who is excellent at the things this role cannot do without, honest about the rest, and self-aware enough to know the difference. Fit, not flawlessness, becomes the standard, and fit can only be judged once you can see the shape.
A hiring process built to find the well-rounded candidate is built to reward the person who is merely inoffensive, the one whose weaknesses are well hidden rather than absent. It systematically overlooks the candidate with a real, role-critical strength and a visible gap somewhere unrelated to the job, because that gap looks like a flaw rather than the ordinary price of being human. For a recruiter, this is the difference between screening people out for weaknesses that do not matter and selecting them in for strengths that do. The more useful question is not how good a candidate is in general, but what they are genuinely strong at, where they are still developing, and whether that particular shape is the one this particular role needs.
Across every complete profile in our data, the number of people who scored in the top tier on every skill at once was zero.
The CV is the least informative document in the room
The CV is where hiring begins. It is the first filter, frequently the only one before a human is involved, and an entire industry of screening software now exists to read it faster. But speed was never the problem. The uncomfortable truth is that we were never able to read the CV properly in the first place, because most of what matters was never in it. And if we could not extract the signal ourselves, it is worth asking how we expect an algorithm to, when we are the ones teaching it on the very same document.
The numbers are stark. Everything a CV reliably contains, education, career history and tenure combined, explains under a tenth of the variation in how people score on the behavioural skills that determine whether they will thrive. More than ninety percent of that signal is nowhere on the page.
The degree tells you almost nothing about soft skills
Education level explains effectively none of the variation in soft-skill scores. A qualification is a real signal of discipline and domain knowledge, but whether a person adapts when a plan collapses, keeps their head under pressure, or genuinely lifts the people around them is simply not contained in it. Every degree filter applied at the top of a funnel is a bet that the qualification predicts the behaviour, and our data says that bet does not pay.
What the research says predicts performance
Decades of research measure how well different hiring methods predict later job performance. The most recent large review (Sackett et al., 2022) puts structured interviews, job knowledge tests and work samples near the top, and situational judgement tests, the method Owiwi builds on, in the middle. At the bottom sit the two things a CV mostly contains: years of education and years of experience. They do predict performance, but only weakly.
And the cover letter does not rescue it
If the CV is thin, the natural instinct is to reach for the cover letter, the place a candidate supposedly reveals their motivation and character. Even that signal is fading. A carefully tailored letter used to say something about the effort behind it. Now AI can write a polished one in seconds, and recent research from a large freelancing platform finds that well-written applications no longer predict who actually delivers the work. When anyone can produce a persuasive letter, the letter stops telling you much about the person who sent it.
The clue hiding on LinkedIn
There is a revealing irony here. The one place online where a professional reputation now lives, the social profile, hints at what the CV lacks, and then mostly fails to deliver it. In principle, how a person behaves on a platform like LinkedIn is a behavioural signal rather than a self-reported one: whether they engage at all, how they treat people in a comment thread, how they respond to disagreement, whether they lift others or only broadcast themselves. That is closer to observing conduct than to reading a claim, and it is precisely the kind of evidence a CV can never contain. The trouble is that most profiles are simply a CV with a photograph, the same list of destinations, the same polished self-description. The potential is behavioural; the reality is usually another self-report.
But the instinct behind it is the right one, and worth naming. Even the internet has quietly worked out that watching how people actually behave tells you more than reading how they describe themselves. Hiring has not caught up. It still leads with the single most self-reported artefact in the entire process, and treats direct evidence of behaviour as an afterthought, if it gathers it at all.
The two documents at the heart of hiring, the CV and the cover letter, are among its weakest predictors. This is not an argument for reading them more cleverly, or for building AI to read them faster. It is an argument for measuring the thing itself. The signal that predicts performance was never on the page, which means it has to be gathered somewhere else.
Experience does not build soft skills the way we think
Of the four assumptions, this is the one that feels most like wisdom. Surely people grow. Surely two decades in the workforce make someone more adaptable, more resilient, better with people. It is the logic beneath every hire that favours the seasoned candidate over the promising one, and beneath the quiet belief that soft skills are something you accumulate rather than something you have.
“The older and more experienced you are, the stronger your soft skills.”
Intuitive, widely held, and quietly shaping who gets hired and promoted. It is also not what the data shows.
Overall soft-skill scores are remarkably flat across age groups. Between people in their twenties and people in their forties, the average moves less than two points. The line barely tilts.
This does not mean people cannot grow. They can, and research that follows people over many years shows some of these qualities maturing with time. What our data shows is that it does not happen on its own. Growth is not automatic, and it is not conferred by time served. The twenty-six-year-old and the forty-six-year-old are far closer on these skills than the seniority ladder implies. Assuming otherwise leads to two expensive mistakes at once: overvaluing experience as a proxy for capability, and undervaluing younger candidates who may already have the behavioural strengths the role demands.
If soft skills do not reliably arrive with tenure, they cannot be waited for. They have to be identified and developed deliberately, at every level, from the graduate hire to the senior leader. For a learning and development function, this reframes the entire task. Soft-skill development is not a remedial programme for junior staff who will grow out of it. It is a permanent responsibility, because the evidence gives no reason to believe the gaps close on their own. There is a more hopeful reading too: if capability is not locked to age, then a person’s stage of life tells you very little about their ceiling, and the strongest candidate for a role that demands adaptability or judgement may be exactly the one the seniority filter would have skipped.
The skills are independent, so one score cannot capture a person
The fourth assumption is the most mechanical, and the most consequential. It is the belief that a candidate can be reduced to a single overall number and ranked. Every scorecard that averages a person down to one figure depends on it, and the belief only holds if soft skills travel together, if being strong in one makes you likely to be strong in the rest. The intuition is seductive. The strong communicator feels like they should also be the strong collaborator. Good people feel broadly good. But the data does not cooperate.
| Acc | Adapt | Dec | Change | Integ | Learn | Resil | Team | |
|---|---|---|---|---|---|---|---|---|
| Acc | .09 | .09 | .07 | .10 | .04 | .08 | .06 | |
| Adapt | .09 | .02 | .10 | .18 | .17 | .16 | .10 | |
| Dec | .09 | .02 | .01 | .03 | .01 | .06 | .01 | |
| Change | .07 | .10 | .01 | .10 | .09 | .09 | .06 | |
| Integ | .10 | .18 | .03 | .10 | .15 | .15 | .12 | |
| Learn | .04 | .17 | .01 | .09 | .15 | .11 | .14 | |
| Resil | .08 | .16 | .06 | .09 | .15 | .11 | .10 | |
| Team | .06 | .10 | .01 | .06 | .12 | .14 | .10 |
The chart shows how every core skill relates to every other, and the pale field is the entire story. Across the board, the skills barely move together. The average relationship between any two is close to zero, and even the most closely linked pair correlates at just 0.18, weak by any standard. Knowing that someone is strong in one skill tells you almost nothing about the others.
This is not a statistical footnote. It is the reason a single score misleads. When skills are genuinely independent, averaging them does not summarise a person, it blurs them. Two candidates can arrive at exactly the same overall rating while being near-opposites underneath, one strong precisely where the other is still developing. Ranked by that one number, they are indistinguishable. In reality one of them fits the role and the other does not, and the score has hidden the only information that would have told you which.
The shift. A single score is not a summary. It is a loss of the very information the decision depends on. The profile, the shape of strengths and development areas, is what a hiring or development decision actually needs.
Any process that shortlists, ranks or rejects on one composite number is discarding signal it has already paid to collect. Reading the profile instead costs nothing extra and changes the quality of the decision, because it restores the distinction between two people the single score had made look the same.
Everyone is a shape, and the best shapes are broad
If no one is uniformly excellent and the skills do not move together, then the honest unit of a person is not a level but a shape: a distinct pattern of strengths and development areas, unique enough that two people with the same average can look nothing alike. This is not a softer way of describing candidates. It is a more accurate one, and it leads somewhere genuinely useful.
When we compare those shapes against overall performance, a clear and slightly counter-intuitive pattern emerges. The people who score highest overall are not the dramatic specialists with one towering strength. They are the ones with even, broad profiles. In soft skills, breadth beats brilliance-in-one-thing, which is the reverse of how technical hiring usually works.
The balanced and generalist profiles outscore the spiky specialist. Range, not a single peak, is the mark of the strongest performers. But breadth alone is not the whole answer, because the most valuable profile combines the two. It is the shape often described as T-shaped: broad capability across every skill, with one or two genuine strengths rising above the rest. Versatile and pointed at the same time.
Even, no peaks
Broad and dependable. Adapts to most roles because little falls outside their reach, but without a signature strength to build around.
Broad, with depth
Generalist breadth plus one or two real strengths. Flexible enough for a changing role, sharp enough to lead on something.
Spiky, narrow
A pronounced strength with matching development areas. Formidable in the right seat, exposed in the wrong one.
This turns hiring from a search for the highest number into a question of fit between shape and role. Neither shape is a better or worse hire in itself. A specialist is a higher-variance one, brilliant where their peak lines up with the job and struggling where it does not. Knowing a candidate’s shape tells you not just whether they are good, but where they will be good, and what a team will need to put around them. It reframes team-building entirely, because a team of identical strong shapes shares identical blind spots, while a team of complementary shapes covers them.
Every profession has a signature, and a blind spot
If individuals are shapes, so are the groups they form. When we look at soft-skill profiles by professional function, each one carries a recognisable signature: a skill it leans on, and one it consistently has the most room to grow. What is striking is how rarely the signature matches the stereotype.
| Function | Development area | Distinctive strength |
|---|---|---|
| HR | Willingness to change −5 | Teamwork +3 |
| Engineering | Willingness to change −3 | Teamwork +3 |
| Finance | Decision-making −4 | Adaptability +1 |
| Customer service | Decision-making −6 | Accountability +5 |
| IT | Integrity −2 | Teamwork +2 |
| Sales | Willingness to change −3 | Adaptability +1 |
| Marketing | Willingness to change −3 | Accountability +1 |
| Commercial | Decision-making −4 | Accountability +3 |
| Retail | Teamwork −9 | Adaptability +5 |
Two patterns stand out. The first is how often the real strength contradicts the stereotype. Engineering distinguishes itself not on raw logic but on teamwork. Finance leads not on rigour but on adaptability, and its clearest development area is decision-making, the very skill its reputation takes for granted. Customer service, so often assumed to run on warmth and people skills, in fact stands out most on accountability while carrying the widest decision-making gap of any function. IT, the supposedly solitary profession, over-indexes on teamwork. Retail leans hard on adaptability, unsurprising for a world of constant change, yet shows the largest teamwork gap we see anywhere. Again and again, the trait a function is known for is not the trait that actually sets it apart, and the gap it needs to close is rarely the one its reputation would predict.
The second pattern is more universal, and more useful. One development area recurs across almost every function regardless of field: willingness to change. From people-facing roles to deeply technical ones, the appetite to adapt, to let go of a familiar way of working, is in short supply nearly everywhere. It is not that any profession lacks intelligence or drive. It is that the readiness to change is the workforce’s most widely shared blind spot, precisely at the moment when change has become the defining condition of work.
Hiring on the stereotype of a profession means solving a problem that may not be there while missing the one that is. A team’s real development need is rarely the one its function is teased for, and naming it accurately is the first step to building the capability the group genuinely lacks. These are group-level tendencies, not verdicts on any individual, but they show how much sharper the picture becomes the moment you measure rather than assume.
We ask for the skills we are least able to find
If soft skills are independent and unevenly distributed, some are simply rarer than others. The question that follows is whether the market has noticed which, and the answer appears to be no. In two important cases, the skills employers demand most loudly are among the ones people are least likely to have.
Judgement is the common gap, and the one machines cannot fill
Decision-making under ambiguity is the single most frequent development area in our data. For more than one in five people, it is their least developed skill, more often than any other. This is not a shortage of knowledge, which training supplies readily. It is a shortage of judgement, the capacity to choose well when the answer is not written down and the information is incomplete. The timing could hardly be more pointed. As automation absorbs the routine and rule-bound parts of work, the decisions left for humans are precisely the ambiguous ones, the calls that cannot be reduced to a procedure. Judgement is the human capability least exposed to automation, and it is the one our data shows to be in shortest supply. An organisation that does not screen for it is leaving its scarcest and most future-proof skill entirely to chance.
Teamwork is demanded everywhere and held by almost no one
Teamwork appears in nearly every job description ever written. It is the default requirement, the thing everyone claims to want. Yet it is the single lowest-scoring skill in our data, at a median of 63%, and it is the strongest skill for only around 3% of people. The most-requested capability in hiring is one of the least commonly held. Part of the explanation may be that we have been measuring the wrong thing. “Team player” has quietly come to mean agreeable, easy, pleasant in a room. But genuine collaboration is a harder and rarer set of behaviours: coordinating with others, disagreeing constructively without damage, and following through on what you promised the group. If we hire for congeniality and call it teamwork, we should not be surprised that the real thing stays scarce.
The pattern. The market is systematically asking for the skills it is least able to find, and in at least one case, using a definition that does not capture what it actually needs.
This is more than an awkward mismatch. It is a structural failure with a compounding cost. Every organisation is fishing in the same shallow pool for the same rare skills, bidding against one another for the small share of people who visibly have them, while the assessment tools most companies use cannot reliably tell those people apart from the confident self-presenters who do not. The result is a market that overpays for the appearance of judgement and collaboration and underinvests in developing the real thing. The scarcity is genuine, but it is made worse by our inability to measure our way out of it. You cannot buy, build or retain a capability you have not learned to see, and the skills that matter most are precisely the ones the standard process is blindest to.
What this means in the age of AI
Every finding in this report becomes more urgent, not less, as artificial intelligence reshapes work. It is worth spelling out why, because the implications run deeper than the usual reassurance that human skills will matter more.
For most of the modern era, hiring could afford to lean on proxies. A degree, a job history, a familiar-looking CV were rough but serviceable stand-ins for capability, because a large share of the work itself was routine and learnable. As machines absorb that routine and technical layer, the residue left for people is exactly the part that resists automation: judgement in ambiguity, adaptability when the ground shifts, resilience under sustained pressure, and the human work of getting a group to function. These are the skills this report has shown to be scarce, unevenly distributed, and invisible to the CV. In other words, the labour market is being reorganised around precisely the capabilities our hiring systems are least equipped to see.
The temptation will be to let AI deepen the old mistakes at speed. An algorithm trained on CVs will scale CV-based bias, not correct it. A model that ranks candidates on a single fit score will industrialise the very reduction this report warns against. The risk is not that machines replace human judgement in hiring. It is that they automate our worst assumptions before we have questioned them.
The organisations that will hire well in this environment are the ones that use technology to measure what the CV cannot, rather than to process the CV faster. That means assessing behaviour directly and early, reading candidates as shapes rather than scores, and treating soft-skill data as a first-class signal in decisions that AI increasingly informs. Used this way, technology becomes the instrument that finally makes the invisible visible.
There is a broader stake here that reaches past hiring. As the human contribution to work narrows to its least automatable core, how we identify, develop and deploy human skill stops being an HR concern and becomes a question of how organisations remain distinctively valuable at all. The companies that understand their people as shapes, that develop judgement and collaboration deliberately, and that stop mistaking credentials for capability, are the ones building an advantage that compounds precisely as the routine work disappears.
Which skills should we prioritise now?
If the old signals fail and the human skills are what remain, a hiring or development leader is left with a concrete question: of the many soft skills one could screen and build for, which deserve the weight? The evidence, combined with where work is heading, points to a clear order of priority. These are not the only skills that matter, but they are the ones where the combination of scarcity, impact and future relevance is highest.
Teamwork, because it is both scarcest and most exposed
Teamwork earns the top place for a simple reason: it is the skill we demand most and possess least, and its value is rising just as its supply looks thinnest. As roles fragment across hybrid arrangements, cross-functional projects and increasingly AI-assisted workflows, the ability to coordinate, disagree constructively and carry a group forward becomes the connective tissue that holds work together. Machines can draft, calculate and summarise; they cannot make a team cohere. With genuine teamwork the strongest skill for only a small share of people, prioritising it is less a preference than a necessity.
Judgement, because it is what is left when the routine is automated
Decision-making under ambiguity is the most common development area in our data and the human capability least exposed to automation. As machines take over the calls that have clear answers, the decisions left for people are the ambiguous ones, and the premium on getting them right only grows. An organisation that does not actively select and develop judgement is leaving its most future-proof skill to chance.
Adaptability and willingness to change, because change is now the constant
Willingness to change is the single most widespread development area across every function we examined. That would be a manageable weakness in a stable world. It is a serious one in a world where the half-life of a skill keeps shrinking and the shape of most jobs is being rewritten within a few years. The capacity to let go of a familiar way of working, and to learn the next one quickly, is fast becoming the meta-skill that determines whether all the others stay relevant.
Integrity and resilience, because trust and endurance underwrite the rest
Beneath the visible skills sit two that make the others durable. Integrity is what allows an organisation to extend autonomy without constant oversight, which matters more, not less, as work becomes distributed and harder to supervise directly. Resilience is what determines whether capability survives pressure or collapses under it. Neither tops a job description, but both quietly decide whether the rest of a person’s strengths hold up when conditions get hard.
The through-line is unmistakable. The skills worth prioritising are the ones that are most human, most difficult to automate, and most badly served by the way we currently hire. They are, in other words, exactly the skills the paper on the desk was never able to show.
So how should we measure people instead?
It is easy to dismantle the old signals. The harder and more honest question is what to put in their place. If the CV is nearly silent, the cover letter barely better, the degree beside the point and the single score misleading, then what actually reveals how a person behaves at work?
The most truthful answer is also the least practical one. If you genuinely wanted to know how someone handles pressure, ambiguity, conflict and collaboration, you would spend a year working alongside them. You would watch them make a hard call with incomplete information, recover from a setback, disagree with a colleague and repair it, adapt when the plan fell apart. Behaviour observed over time, in real situations, is the gold standard, and everyone knows it. It is simply impossible to run a hiring process that way. Nobody can shadow a candidate for a year before deciding.
So the real task is not to find the perfect measure. It is to find the closest honest approximation to that year of observation that can be run in an afternoon. And the principle that gets you there is simple: stop asking people to describe themselves, and start looking at the choices they actually make.
This is where self-report quietly fails. Ask a candidate whether they are resilient and they will say yes, because who would say otherwise, and because most people genuinely believe it. The same problem haunts the interview, where confident storytelling stands in for evidence, and the cover letter, where the skill being measured is writing rather than working. Each of these asks a person to narrate their own behaviour, and narration is exactly where the signal breaks down.
The alternative is to put people in front of realistic situations and look at the choices they make. This is the logic of the work sample, the structured exercise, and the situational judgement assessment: rather than asking whether someone is a good decision-maker, you place them inside a realistic decision and see what they choose. Strictly speaking, a situational judgement test is still a form of self-report, since the candidate says what they would do rather than being observed doing it, but it asks for a decision in context rather than a description of themselves. It compresses that impossible year of observation into a controlled, comparable, repeatable experience. It is not a shadow of the real thing. It is a sample of it.
The goal was never to read people more cleverly on paper. It was to stop reading about them at all, and to start seeing how they choose, even if only for a scenario at a time.
Every organisation already knows, intuitively, that you learn more about someone from an hour of doing than from a page of describing. The evidence simply makes that intuition rigorous, and points to a practical way to act on it: measure behaviour directly, early, and in situations that resemble the work itself. That is the approximation of the year we cannot spend, and it is available now.
The talent everyone else is ignoring
Turn the report around and look at it from the candidate’s side, and the argument stops being about fairness and becomes about advantage. Every weakness in the old system is a mispriced asset for the organisation clear-sighted enough to notice.
Consider what the evidence implies. If the CV hides most of what makes a person good at their work, then the labour market is full of capable people who are systematically undervalued, filtered out not for what they can do but for how they read on paper. Everyone else is competing for the same narrow set of candidates who present well, bidding their salaries up and their availability down. Meanwhile a large pool of people with the actual skills sits overlooked, because they interview modestly, lack the expected pedigree, or simply do not fit the template. The employer who can measure behaviour directly gains access to that pool. It is the closest thing hiring has to arbitrage: strong people, real skills, and almost no competition for them, precisely because no one else is even looking.
The practical edge. Measuring what the CV cannot show is not only fairer. It is a sourcing advantage. It lets you hire strong, overlooked people before the rest of the market learns to see them.
And it changes how we should talk about our own worth
There is a second, quieter shift here, and it reaches every professional, not only the ones doing the hiring. If behaviour matters more than credentials, then the way we narrate our own careers is due for a rewrite. The CV is a list of destinations: the titles held, the companies joined, the results delivered. It says almost nothing about the part that actually reveals capability, the how and the why. How did you make the call when the information was incomplete? Why did you change course when the plan stopped working? What did you do when the team fell apart, or the project failed, or the obvious answer turned out to be wrong?
These are the moments that show what a person is made of, and they are exactly what the destination-focused CV erases. The most useful thing a professional can learn to do, in a world that is slowly learning to measure behaviour, is to give worth to the journey rather than only the arrival. Not what you achieved, but how you achieved it, and why it mattered. That is a better account of a working life, and it happens to be the account the evidence says we should have been listening to all along.
The two shifts are the same shift, seen from two ends. Organisations that measure behaviour find the talent everyone else missed. People who learn to speak about the how and the why finally become visible for what they can actually do. The destination was never the point. The journey always was.
Four new assumptions to hire by
If the four beliefs this report began with no longer hold, the useful response is not simply to abandon them but to replace them. The evidence points to a different set of starting assumptions, ones that are less comfortable but far better supported.
Everyone is a shape, not a level
Stop looking for the well-rounded candidate. Look for the profile whose strengths match what the role genuinely needs, and whose development areas fall where they matter least.
Capability must be measured, not inferred
The CV, the cover letter and the credential are among the weakest predictors we have. What predicts performance has to be observed directly, in behaviour, not read off a page.
Soft skills are built, not banked
They do not accrue automatically with time served. They must be developed deliberately, at every level and every age, or the gaps simply persist.
The profile is the signal, not the score
Never let one number stand in for a person. The shape underneath it is where the decision actually lives.
None of this argues for less rigour in hiring. It argues for pointing rigour at the right target: the behavioural capabilities that actually predict performance, rather than the proxies that stood in for them because they were easy to measure. The future of work will not be staffed by people who are good at everything. It never has been. It will be built, as it always has, by people who are exceptional at something, deployed in the roles that need that something, and developed by organisations clear-eyed enough to see what each person actually offers. In an age where the human contribution is narrowing to its most valuable core, that clarity stops being a nicety. It becomes the difference between organisations that thrive on their people and those that merely employ them.
These findings draw on more than two million skill observations from Owiwi, a gamified soft-skills assessment platform. Rather than asking people to describe themselves, Owiwi places candidates inside realistic, interactive workplace scenarios and scores the choices they make against validated behavioural indicators, an approach well established in organisational psychology for capturing how people actually behave.
The analysis focuses on the core soft skills common across the assessed population. Three more recently introduced skills sit on lower assessment volume and are shown only where noted. All figures describe population-level associations and are presented as such, not as causal claims.
Try Owiwi for free.
5 assessments. 21 days. No credit card required.
- Galdin, A., & Silbert, J. (2025). Making talk cheap: Generative AI and labor market signaling. Working paper, arXiv:2511.08785.
- Roberts, B. W., Walton, K. E., & Viechtbauer, W. (2006). Patterns of mean-level change in personality traits across the life course: A meta-analysis of longitudinal studies. Psychological Bulletin, 132(1), 1–25.
- Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 107(11), 2040–2068.
- Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124(2), 262–274.
© Owiwi. All rights reserved. This report is intended as thought leadership and reflects population-level patterns in anonymised assessment data.
