A competency requirement is an employer’s answer to a question about work that does not yet exist in its final form. What employers describe for the end of this decade is less a list of tools than a description of judgment: deciding well once the routine part of the task is handled elsewhere.
Quick Overview
The usual way of writing about future competencies is to publish a list and let the reader assume the list is the point. It is not. A list of named skills ages badly, because the naming convention changes faster than the underlying requirement does. What is durable is the direction of travel, and the direction is legible without any forecast at all.
The direction is this. Work that can be specified precisely enough to be described in a procedure is steadily being separated from work that cannot. The first kind moves toward automation, toward shared services, toward software. The second kind stays with people and gets denser. That is not a prediction about a particular year; it is a description of what happens to any task once its rules become explicit enough to be written down.
The consequence for competencies follows directly. If the specifiable part of a role leaves, what remains is the part that requires reading an ambiguous situation, choosing between options that are all defensible, and being accountable for the result. Employers are describing that residue when they say they want critical thinking, adaptability and collaboration. Those words sound soft. The requirement behind them is not.
This article is about that residue across the whole labour market, not inside information technology. The IT-specific version of the question — which technology shifts reshape a developer’s or an administrator’s competence profile — is a different subject with a different answer, and mixing the two produces advice that fits nobody.
Why the Ground Moved: Skill Half-Life Against the Training Cycle
The pressure on competencies does not come from technology alone. It comes from a mismatch between the rate at which a skill loses value and the rate at which an organisation can design a response to that loss.
An organisation’s training cycle has a floor set by its own procedures. Someone has to notice a gap, describe it, get it funded, find a provider, schedule participants, run the programme and wait for the effect to show up in work. That sequence takes what it takes. It cannot be compressed below the speed of the slowest approval in it, and in most organisations that approval is not a training decision at all — it is a budget decision made on an annual rhythm.
When a skill stayed useful for the length of a career, that cycle was fine. It only had to run a few times per working life. The current problem is arithmetic rather than dramatic: when the useful life of a specific technique falls below the length of the cycle that replaces it, the organisation is permanently responding to a gap that has already changed shape.
There are only a small number of ways out of that arithmetic, and adding training days is not among them. Either the cycle gets shorter, or the thing being trained gets more durable, or the learning stops being an event and becomes part of the work. Serious employers are attempting all three at once, which is why their competency language keeps drifting away from named tools and toward capacities that survive a tool being replaced.
This is also why the phrase continuous learning has stopped being a slogan in job descriptions and started being a testable requirement. An employer asking about it is not asking whether you enjoy courses. It is asking whether you can acquire a working understanding of something unfamiliar without being taught, because there will be no time to teach you.
The Shift Employers Are Making: From Operating a Tool to Owning a Decision
Here is the axis that everything else in this article hangs on. The competency shift being described for the end of this decade is not a shift from old tools to new tools. It is a shift in which part of the task the person is responsible for.
Consider what a role looks like when it is defined by tool operation. The instrument is specified, the inputs are specified, the correct output is specified, and competence means producing that output reliably. Quality is measured as conformance. Someone who does this well is valuable in direct proportion to how much of it they can do. That description covers a very large share of white-collar work as it was organised until recently, and it covers it outside technology entirely — in finance operations, in claims handling, in procurement administration, in scheduling, in reporting, in large parts of legal and HR support work.
Now consider what happens when the specifiable part of that role is absorbed by a system. The instrument still exists; the person is still in the loop. But the question they are answering has changed. It is no longer did I produce the specified output but is this output right, and if it is not, what do I do about it. That question cannot be answered by conformance. It requires knowing what the output is for, what a wrong answer would cost, and which of the available responses is least bad.
That is the whole shift, and it is why the competencies employers name look like a psychology syllabus rather than a technology one. Critical thinking is the capacity to notice that a plausible answer is wrong. Adaptability is the capacity to keep working when the procedure no longer fits the case. Complex problem-solving is what you need when there is no procedure at all. Communication matters because a decision that cannot be explained cannot be accepted by the people it affects.
The practical test for a manager is simple and uncomfortable. Take any role in your organisation and ask what remains of it if the routine output is produced by something else. If the honest answer is nothing remains, the role is a procedure with a person attached, and the competency question for that person is a redeployment question. If the answer is the judgment remains, then the training question is how to make that judgment better, faster and more explicit — which is a completely different investment from teaching another tool.
Digital Competence Stopped Meaning Tool Operation
The most common error in workforce planning is to treat digital competence as a proficiency level with an instrument. That framing was never quite right and is now actively misleading, because it produces training that expires with the software it was written for.
The European reference model for this is worth reading precisely because it refuses that framing. The Digital Competence Framework (DigComp) organises digital competence around what a person is trying to accomplish — finding and evaluating information, communicating and collaborating, creating content, staying safe, solving problems — rather than around which application they are holding. Information evaluation, in particular, is treated as a competence in its own right, which is exactly the capacity that a role needs once the system in front of it can produce fluent output that is sometimes wrong.
The consequence for hiring is that a candidate’s tool list is weak evidence. It tells you what they have been exposed to, not what they can do when the exposure does not match the problem. A stronger signal is whether they can describe a case where the tool gave them an answer they refused to use, and why.
The consequence for job design is larger. Digital competence spread outward from the IT function years ago, and it did so through infrastructure rather than through office software. Connected sensors, industrial networks and low-latency links now sit inside facilities management, logistics, quality assurance, maintenance and field service — functions whose job descriptions were written before any of that existed. The people in them are not being asked to become network engineers. They are being asked to understand what the connected layer can and cannot tell them, which readings deserve trust, and where an operational decision now depends on data whose provenance they need to be able to question. That boundary — the operational meaning of connected infrastructure for people who do not build it — is what the training programme 5G and IoT: The Synergy of Future Technologies is organised around, and the audience for it is not the technology department.
Working Alongside a Model Is a Competence, Not a Licence
The competency most frequently misdescribed in current job advertisements is the ability to work with generative systems. It is usually written as a tool skill, which is the same category error again, and it produces training that teaches people to phrase requests rather than to evaluate answers.
What the work actually requires has a different shape. A person working alongside a model has to hold two things at once: what the system is good at, and what it will produce confidently when it is wrong. The second is harder and matters more. A system that fails loudly is easy to work with. A system that fails fluently, in the register of a correct answer, transfers the entire burden of verification onto the person using it — and that verification is a domain competence, not a prompting competence. You cannot check an answer in a field you do not understand, no matter how skilfully you asked the question.
This is also where competence acquires a legal edge. The European regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) does not treat familiarity with these systems as an optional professional nicety; it establishes obligations around the competence of the people deploying and using them, alongside its risk-tiered requirements on the systems themselves. Whatever an organisation’s own view of the training question, the obligation to be able to demonstrate that its people understand what they are operating is not a matter of internal preference. It sits in the same category as any other compliance capability the organisation has to be able to evidence.
The practical requirement, then, is a combination that no tool course delivers on its own: enough understanding of how these systems generate output to know which failure modes to expect, plus enough command of the subject matter to catch a wrong answer, plus the professional confidence to reject an output that looks authoritative. Employers describing this as AI literacy are not being vague. They are naming a competence whose absence is invisible until the day it is expensive.
The Sustainability Transition Reaches Roles That Never Had an Environmental Brief
The second force reshaping competency requirements is the sustainability transition, and it has a property that surprises people: its competency effects are largest outside the sustainability function.
When environmental performance was a specialist concern, it could be handled by a specialist team producing a report. Once it becomes a reporting obligation with defined boundaries and audited figures, it stops being a report and becomes an input to ordinary decisions. Someone in procurement now needs to know what a supplier declaration is worth. Someone in product design needs to know what a material choice does downstream. Someone in facilities needs to know what an efficiency claim rests on. Someone in finance needs to know why a number that looks like an estimate is being disclosed with the same seriousness as a financial one.
None of those people are environmental specialists, and none of them will become one. What they need is narrower and harder to teach: the ability to trace a claim to whatever produced it, and to recognise when a figure is a measurement, a model output or a convention. That is a data-literacy competence wearing an environmental jacket, and it is the same competence the previous section described in a different setting.
There is a second competence that the transition creates and that organisations consistently underestimate: the ability to communicate an environmental claim to an audience that will test it. Internally that audience is a board asking what a commitment will cost and when it becomes binding. Externally it is customers, investors, auditors and regulators, none of whom accept enthusiasm as evidence. The person doing this has to be able to say what is measured, what is estimated, what is assumed and what is not yet known — in plain language, without either overstating the position or hedging it into meaninglessness. That is a demanding communication competence resting on a technical understanding of the underlying figures, and it is now being asked of people whose job title says nothing about sustainability at all.
The pattern is worth naming explicitly, because it recurs. Every time a specialist domain becomes an obligation that touches ordinary decisions, the competence requirement it generates is not learn the specialism. It is be able to interrogate the specialism’s outputs well enough to use them responsibly. Regulatory literacy, environmental literacy and data literacy converge on the same underlying skill, which is why an organisation that trains them as separate curricula usually ends up paying for the same capacity several times over.
What Occupational Classifications Show, and What They Cannot
There is a reasonable objection to everything above: if the shift is real, it should be visible in the way occupations are defined and counted. It is, partly — and the way it is only partly visible is itself informative.
Occupational classification systems exist to make labour markets comparable. The International Standard Classification of Occupations 2008 (ISCO-08) groups jobs by the kind of work performed and the skill level it requires, which is what makes statistics from different countries mean the same thing. Skill and competence taxonomies such as ESCO go further and describe the underlying capabilities, linking occupations to the skills they draw on rather than to the titles they carry. Cedefop’s Skills intelligence work exists precisely to read those signals across European labour markets.
What these instruments do well is describe structure. What they cannot do is move quickly. A classification earns its value by being stable — that is the entire point of a shared standard — and stability means that a genuinely new configuration of work will exist in the labour market for some time before it exists in the classification. Job titles arrive first, then postings, then the taxonomy.
The sequence is observable if you look for it. A capability first appears as an awkward paragraph in a job advertisement, written by a hiring manager who knows what they need and has no established vocabulary for it. Then it acquires an unstable name, usually borrowed from whichever vendor or consultancy described it first, and that name is used inconsistently across employers for a while. Only after the usage settles does it become a discrete entry in a taxonomy, at which point everyone can finally agree on what they were all separately hiring for. The competence existed throughout; only its label was missing. An organisation reading the awkward paragraphs is reading the same signal as an organisation reading the taxonomy, several years earlier and with more noise.
For anyone doing workforce planning, that lag has a straightforward implication. If you wait for a competence to be formally named before you build it, you have chosen to be late by exactly the length of the standardisation process. The organisations that are not late are the ones reading the demand signal directly — in what their own roles are struggling with, in what candidates are being asked for, in what work keeps being escalated because nobody in the current structure owns the decision it requires.
Social and Emotional Competence Becomes Load-Bearing
The counterintuitive part of the competency shift is that the more of the specifiable work leaves, the more the interpersonal part of the job carries structural weight. This is not a sentimental claim, and it does not rest on any survey.
The mechanism is about where the difficulty relocates. When a role is defined by producing a specified output, most of its interpersonal content is coordination: who needs what, by when, in what format. When the role is defined by judgment, its interpersonal content becomes something else entirely — establishing what a request actually means, negotiating between people who want incompatible things, explaining a decision to someone who will be worse off because of it, and holding a position under pressure without turning it into a conflict.
Those are different capacities from coordination, and they are the ones that determine whether a judgment-heavy role functions. A person who reaches a good decision and cannot get it accepted has not done the job. A person who cannot tell the difference between a stakeholder being obstructive and a stakeholder having information nobody has surfaced will keep making confident decisions on incomplete inputs.
Distributed and cross-border working sharpens all of this, because the informal channels that used to carry it are thinner. Reading hesitation is harder through a video call. Establishing enough trust for someone to say I think this is wrong takes deliberate effort when the relationship has no corridor time in it. That effort is a competence, it is developable, and organisations that treat it as a personality trait simply forgo the development.
Emotional self-management belongs in the same category for an unglamorous reason: judgment degrades under sustained pressure before anything else does. A person who cannot regulate their own state under uncertainty will not exercise good judgment in exactly the conditions where judgment is what they were hired for.
Leadership Turns Into the Design of Learning Conditions
If the competency requirement is shifting from output to judgment, then the leadership requirement shifts with it, and in a direction that many management populations are not selected for.
Managing conformance work is a matter of specification and monitoring: define the standard, check adherence, correct deviation. Managing judgment work cannot be done that way, because the manager frequently does not know the right answer either. What the manager can do is shape the conditions under which the person deciding is likely to decide well — which means making it safe to surface an uncertainty early, making it normal to be wrong in a recoverable way, and making the reasoning behind decisions visible enough that other people can learn from it.
That is a design task, not a supervisory one. It shows up in unglamorous specifics: whether raising a concern costs the person raising it anything, whether a post-mortem produces a cause or a culprit, whether the calendar has any space in it that is not committed to delivery, whether an experiment that failed is treated as information or as a performance issue.
Leading a group with a wide spread of adaptation is a distinct sub-problem and a common one. During any transition, a team contains people who have already reorganised how they work and people who have not, and the gap between them is usually not about capability. It is about how much of a person’s professional identity is invested in the way the work used to be done. Treating that as resistance guarantees resistance. Treating it as a legitimate cost — someone is being asked to become a beginner again, in public — makes it negotiable.
The leadership competence, then, is not inspiration. It is the ability to run a group in which learning is happening continuously and visibly, without that learning being punished by the same performance system that rewards flawless delivery.
Distributed Work Made Self-Management a Job Requirement
Flexible and hybrid arrangements changed the competency profile of ordinary roles in a way that is still under-acknowledged, because the discussion has been dominated by questions of policy rather than capability.
When work happens in a shared physical space, a large amount of coordination is free. Priorities get corrected in passing. Confusion surfaces because someone looks confused. Focus is partly enforced by the environment. Remove the shared space and none of that disappears — it becomes an individual responsibility that was previously carried by the setting.
The capacities this promotes into job requirements are prosaic and rarely taught: deciding what matters today without being told, protecting attention from an interruption stream that has no natural boundary, communicating proactively enough that nobody has to guess your status, and noticing your own drift before it becomes a delivery problem. These were once personal working habits. In a distributed role they are load-bearing, and their absence looks like poor performance rather than like a missing skill.
The office does not disappear in this arrangement; its purpose narrows. It stops being where routine work is executed and becomes where the things that need physical presence happen — the difficult conversation, the collaborative design session, and above all the informal transfer of judgment from experienced people to inexperienced ones. That last function is the one most at risk, because it was never scheduled and therefore never protected. Early-career professionals lose disproportionately when it stops, and they lose exactly the tacit judgment this whole article says the market is about to demand from them.
Who Actually Gets to Acquire These Competencies
A competency shift is also a distribution question, and ignoring that produces workforce plans that fail for reasons the plan never modelled.
The capacities described here are not evenly acquirable. Judgment develops through exposure to consequential decisions with feedback, which means it develops fastest for people who are already given consequential decisions. Adaptability develops through supported experience of change, which means it develops fastest for people whose failures are survivable. Cross-boundary literacy develops through contact with other functions, which means it develops fastest for people whose roles are not narrowly boxed. Each of these is easier to obtain from a position that already has some standing.
Public policy has a stated interest here — Europe’s Digital Decade sets out the European Union’s digital objectives for 2030, including targets on the digital capability of the population and of businesses, precisely because leaving that distribution to itself is not expected to reach everyone. But an organisation cannot outsource its own version of the problem to policy, because its version is internal and specific.
Internally, the question is concrete. Which roles in this organisation get exposure to real decisions, and which are structurally denied it? Who gets sent on development that changes what they can do, rather than development that certifies what they already did? When work is redesigned around a new system, who is given time to learn it properly and who is expected to absorb it while maintaining full output?
The answers tend to be uncomfortable and are almost always correlated with things the organisation would not defend if asked directly: seniority, function, location, working pattern, and who happened to have a manager with slack in their budget. Naming the correlation is the first useful step, because none of it is a competence problem. It is an access problem that will later be diagnosed as a competence problem.
What an Organisation Does With This
The practical response has a small number of parts, and their order matters more than their content.
The first is to stop planning competence at the level of tools. A skills inventory organised around named instruments will be obsolete on the schedule of those instruments, and it will systematically under-record the capacities that actually differentiate performance. Organising the same inventory around what people must be able to decide produces a shorter list that ages far more slowly and maps directly onto the judgment residue described earlier.
The second is to distinguish two questions that are usually merged. Which competencies does this organisation need is a strategy question and belongs with people who know where the business is going. Which competencies does this person have is a diagnostic question and needs an instrument designed for it — structured observation, simulation exercises, evidence from real work — rather than an opinion collected once a year. Confusing the two produces development plans that are precise about the wrong thing.
The third is to accept that a portion of the required learning cannot be delivered as a course. Judgment is built by doing consequential work with feedback from someone more experienced, which makes mentoring and structured review part of the competence infrastructure rather than a benefit. Formal training remains necessary, and how its delivery is changing — personalisation, simulation, shorter formats embedded in the working day — is a substantial subject in its own right, examined in The Future of Training: How AI, VR, and Personalization Are Changing Professional Development.
The fourth is to think about breadth deliberately rather than accidentally. A workforce of narrow specialists is fragile when the boundaries between specialisms move, and a workforce of generalists cannot do anything hard. The deliberate middle — deep capability in something specific, plus enough command of adjacent domains to work across the seam — is a design choice with a literature behind it, discussed for technology roles in T-shaped Learning in IT: Why the Future of Specialists Requires a Broader Perspective, and applicable well beyond them.
The last part is retention, and it is not a separate topic. In a market where capability decays and has to be rebuilt continuously, the organisation that builds capability and cannot keep it is funding its competitors’ workforce development. The condition that keeps people is rarely the training budget itself. It is whether the work they are given afterwards uses what they just learned.
Frequently Asked Questions
Which competencies will matter most to employers by the end of this decade?
The durable answer is not a list of tools. It is the capacity to make and defend decisions in situations that no procedure covers: evaluating information whose reliability is uncertain, working across a boundary between domains, communicating a decision to people it affects, and continuing to learn without being taught. Specific technical competences still matter, but they are the part of the requirement that changes name most often.
Is this only relevant to technology roles?
No, and treating it as a technology question is the most common planning error. The shift is largest in roles that were built around producing a specified output — operations, administration, reporting, first-line professional support — because those are the roles where the specifiable part is easiest to separate from the judgment part. Technology roles have their own version of the question, with a different answer.
How is a competency requirement different from a training requirement?
A training requirement names something to be taught. A competency requirement names something the person must be able to do, and is neutral about how they get there. That distinction matters because a substantial share of what employers now describe cannot be delivered by teaching alone: judgment develops through consequential work with feedback, which is an assignment decision, not a curriculum decision.
What should an organisation do first?
Rewrite the competence inventory around decisions rather than tools, then check honestly which roles are given the exposure that develops judgment and which are not. That second check usually explains more about an organisation’s capability gap than any assessment of individual skill, and it costs nothing but candour to perform.