Sim Chen Xing

Artificial intelligence is often discussed as a technological challenge. Organisations ask whether employees can use emerging tools effectively, whether AI can be deployed safely, and whether it can improve productivity, reduce repetitive work or accelerate decision-making. These are important questions, but they may not be the most difficult ones. A deeper challenge emerges when technological adoption begins to alter the capabilities required to perform work itself.

AI literacy, on its own, is unlikely to be sufficient. Effective AI-enabled work depends on a broader constellation of capabilities: domain literacy, process literacy, data literacy, analytical and methodological literacy, technological literacy, governance literacy, ethical literacy, institutional literacy, and human and stakeholder literacy. AI may make work faster, but speed does not remove the need to understand the work itself. In many cases, it may make such understanding more important.

As organisations accelerate their use of AI, some employees will develop alongside them. Others may initially struggle before adapting. Some may move into roles where their existing strengths remain valuable. Yet there will also be individuals whose capabilities develop more slowly than the requirements of the organisation around them. Over time, the difference between what the organisation requires and what the individual can demonstrate may begin to widen.

The workforce challenge is therefore not simply to determine whether an employee is capable today. It is also to ask whether that employee’s capabilities are developing quickly enough for where the organisation is going tomorrow.

This essay proposes the Capability Divergence Model as a conceptual framework for thinking about that problem. The model distinguishes between present capability and capability trajectory, considers the relative speed at which organisations and individuals are changing, and introduces the idea of a tolerance threshold at which persistent divergence may require organisational intervention. It also places an important qualification around these ideas: capability development is not solely an individual responsibility. Organisations must provide sufficient clarity, opportunity, support and time for adaptation to occur.

The central argument is that, in periods of rapid technological and organisational change, static performance may no longer tell us enough.

What may increasingly matter is not only whether an individual can perform today’s work, but whether their capability trajectory is converging with the work of tomorrow.

AI and the Exposure of Capability Gaps

AI does not necessarily create capability gaps. In many cases, it reveals weaknesses that were already present but previously obscured by organisational structures.

Established organisations contain many mechanisms that can compensate for uneven individual capability. Highly structured procedures can reduce the need for independent judgement. Experienced colleagues may quietly correct errors. Managers may absorb problems before they become visible. Legacy systems may restrict how much discretion employees are expected to exercise. Organisational routines can also allow individuals to perform adequately without developing a deep understanding of the wider system in which their work sits.

AI can disrupt these arrangements. As routine components of work become automated or easier to execute, employees may increasingly be expected to interpret, verify, decide, coordinate, redesign and exercise judgement more independently. This changes the baseline of what competent performance requires.

An employee who understands the purpose behind a process may use AI to improve or redesign it. An employee with strong data and methodological literacy may use AI to explore questions more quickly while retaining the ability to assess whether the underlying method is appropriate. Someone with strong governance literacy may recognise when an efficient-looking AI-enabled process introduces unacceptable risks. Someone with strong stakeholder literacy may recognise that a technically accurate output may still be inappropriate for the audience receiving it.

Conversely, a person who lacks these surrounding capabilities may become faster without necessarily becoming better. AI can produce polished writing, analysis, presentations and recommendations, but polished output can obscure weaknesses in reasoning, judgement and understanding.

AI-enabled output is not necessarily equivalent to AI-enabled capability.

This distinction matters because organisations may otherwise mistake visible productivity gains for deeper capability growth. Increased output, faster turnaround or greater tool usage may demonstrate that employees are using AI. They do not necessarily show that employees are becoming more capable of evaluating, contextualising and governing the work being produced.

The relevant challenge is therefore broader than an “AI skills gap”. It concerns the relationship between the capabilities demanded by the role and the capabilities demonstrated by the individual.

From Capability Gaps to Capability Divergence

Let the capability required for a role at a particular point in time be represented as:$$R_t$$
and an individual’s demonstrated capability at the same point as:$$C_t$$​
The capability gap can then be represented conceptually as:$$\boxed{{\displaystyle G_t=R_t-C_t}}$$
​​where $G_t$ represents the capability gap at time $t$, $R_t$ the capability required by the role, and $C_t$ the capability demonstrated by the individual.

Neither $R_t$ nor $C_t$ should be interpreted as a single skill. Both are better understood as composites of multiple capabilities whose importance varies by context. A more complete representation may therefore be expressed as:$$G_t=\sum _{i=1}^n{w}_i(R_{i,t}-C_{i,t})$$
where $i$ represents a capability dimension and $w_i$ represents the relative importance of that capability to the role.

These weights are necessarily contextual. A position involving sensitive information may place greater emphasis on governance and data literacy. A policy role may depend more heavily on analytical, institutional and stakeholder literacy. A research role may place substantial weight on methodological and domain competence, while a frontline role may depend more strongly on process understanding, communication and human judgement.

There is therefore no universal AI-era competency profile. The more meaningful question is whether a person possesses, or can develop, the specific combination of capabilities required for the work they are expected to perform.

Yet a capability gap alone provides only a snapshot. It tells us where someone stands at a particular moment, but not where they are heading.

Consider two employees. The first has a noticeable capability gap but actively learns, experiments, asks questions, incorporates feedback and demonstrates sustained improvement. The second appears more capable at present but shows little development as the role continues to evolve. A conventional performance assessment may favour the second employee. A longitudinal assessment may reach a different conclusion.

This is why capability trajectory matters.

Let:$$V_O=\frac{\mathrm{\Delta }R}{\mathrm{\Delta }t}$$​

represent Organisational Capability Velocity: the rate at which capability requirements are changing.

Let:$$V_I=\frac{\mathrm{\Delta }C}{\mathrm{\Delta }t}$$​

represent Individual Adaptation Velocity: the rate at which an individual’s demonstrated capability is developing.

The resulting rate of change in the capability gap may then be represented conceptually as:$$\boxed{{\displaystyle \frac{\mathrm{\Delta }G}{\mathrm{\Delta }t}=V_O-V_I}}$$​​
This relationship produces three broad situations.

If:$$V_I>V_O$$​

the individual is catching up with organisational requirements and the gap is narrowing.

If:$$V_I=V_O$$​

the individual is developing at approximately the same rate as organisational expectations, meaning that an existing gap may remain relatively stable.

If:$$V_I<V_O$$​

the individual is falling progressively further behind.

This produces an important insight.

A person can be improving and still be falling behind.

An employee may be learning, producing better work and becoming more capable than they were six months earlier. Yet if the capabilities demanded by the organisation are evolving even more rapidly, the distance between the individual and the role may continue to widen.

The relevant comparison is therefore not simply between the individual today and the individual yesterday. It is between the trajectory of individual development and the trajectory of organisational change.

Performance may tell us whether someone can perform today’s job. Capability velocity may offer insight into whether they are likely to remain able to perform tomorrow’s.

Organisational Velocity and Individual Adaptation

Organisational capability velocity is not determined by AI adoption alone. An organisation does not necessarily transform simply because it purchases an AI platform or gives employees access to new tools. What matters is whether technology changes how work is actually performed.

Organisational Capability Velocity may therefore be represented conceptually as:$$V_O=f(A,S,I,P)$$

where $A$ represents the pace of AI and technology adoption, $S$ the organisation’s strategic ambition and senior-management direction, $I$ the intensity of implementation, and $P$ the pace of process and operating-model change.

Two organisations could adopt the same technology while experiencing very different capability consequences. One might introduce an AI tool while leaving existing processes, responsibilities and decision structures largely intact. Another might use the same technological capability to redesign workflows, automate administrative work, decentralise decisions, remove intermediary layers and increase expectations of independent judgement.

The capability pressure on employees would be substantially greater in the second organisation.

Senior-management direction therefore matters because strategy determines how deeply technology is expected to reshape work. Yet strategic direction alone is insufficient. Organisational capability requirements rise through the interaction of technological adoption, leadership ambition, implementation intensity and changes to the underlying operating model.

The same principle applies on the individual side.

Individual adaptation should not be measured simply by the number of courses completed, workshops attended, certifications obtained or AI tools experimented with. These are learning inputs rather than capability outcomes.

Individual Adaptation Velocity can instead be conceptualised as:$$V_I=f(L,A_p,O,F)$$

where $L$ represents learning achieved, $A_p$​ practical application, $O$ demonstrated outcomes, and $F$ responsiveness to feedback.

One employee may attend extensive formal training yet demonstrate little change in practice. Another may receive relatively little structured training but learn quickly through experimentation, application, feedback and reflection. In capability terms, the latter may be adapting more rapidly.

This distinction matters because organisations often use proxies for development precisely because they are easier to measure. Training hours, course completion and certification can indicate that developmental opportunities were offered. They do not necessarily demonstrate that capability has increased.

The relevant question is not merely whether learning activity occurred.

It is whether learning changed the individual’s ability to perform.

Developmental Runway and Organisational Responsibility

A discussion of capability divergence would be incomplete, and potentially unfair, if it placed the entire burden of adaptation on employees.

Individuals cannot reasonably be expected to adapt to changing requirements that were never communicated. They cannot be faulted for failing to demonstrate capabilities they were never given an opportunity to practise. Nor can organisations credibly emphasise transformation while providing insufficient time, resources or support for employees to transform alongside them.

This requires a further concept: developmental runway.

Conceptually:$$D_R=f(T,O,S,F)$$

where $T$ represents reasonable time since expectations were communicated, $O$ genuine opportunities to practise and demonstrate the required capabilities, $S$ the support, resources, training and guidance provided, and $F$ the clarity and frequency of feedback.

Developmental runway matters because the same capability gap can imply very different things under different organisational conditions. A substantial gap accompanied by little time, weak communication and minimal support primarily signals the need for organisational intervention. The employee may not yet have been given a reasonable opportunity to adapt.

The interpretation becomes different when expectations have been clear, opportunities to develop have been genuine, feedback has been repeated, support has been provided and sufficient time has passed, yet the individual’s capability trajectory continues to show little evidence of convergence.

This creates reciprocal responsibility. Organisations must create conditions in which adaptation is reasonably possible. Individuals must engage seriously with those opportunities and demonstrate that development is occurring.

Without the former, organisations risk blaming employees for transformations they have not adequately supported. Without the latter, workforce development risks becoming an indefinite process in which the provision of training is mistaken for the achievement of capability.

When Does Divergence Become Unsustainable?

Not every capability gap is equally consequential.

A weakness in an emerging capability that remains peripheral to a role may be tolerable for a considerable period. A similar gap may be much more serious if it affects safety, governance, financial controls, customer outcomes, regulatory obligations or critical decision-making.

Organisations therefore operate with some form of Capability Tolerance Threshold, even if this threshold is rarely made explicit.

Let that threshold be represented as:$$\theta$$

When:$$\boxed{{\displaystyle G_t>\theta }}$$​

the capability mismatch has become sufficiently material that intervention is required.

This should not be interpreted as:$$G_t>\theta \Rightarrow \text{Dismissal}$$

Such a conclusion would be both simplistic and poor management. Crossing the threshold should instead be understood as a decision point at which the organisation must determine which intervention is most appropriate.

A useful sequence is:

Develop → Redesign → Redeploy → Separate

The first option is development. Can the capability gap reasonably be closed through targeted learning, coaching, practice, feedback or support?

The second is redesign. Is the problem partly a consequence of how the role or workflow has been structured? Could the work be redesigned in a way that better combines human and technological capabilities or makes more effective use of the individual’s strengths?

The third is redeployment. The employee’s capabilities may remain valuable to the organisation even if they no longer align well with the current role. A different position may provide a better match.

Only after these possibilities have been considered does separation become the final option: the possibility that continued employment in the existing arrangement is no longer sustainable.

The threshold itself cannot be universal. It depends on the criticality of the missing capability, the consequences of errors, the individual’s trajectory, the level of managerial support required, the additional workload transferred to colleagues, stakeholder impact, institutional knowledge, the availability of alternative roles, the difficulty of replacement and the pace at which the organisation itself needs to move.

This is why the model should be understood as a judgement framework rather than an arithmetic rule.

Its purpose is not to automate employment decisions, but to make the assumptions underlying those decisions more explicit.

The harder question follows naturally: when does continued retraining cease to make sense?

Organisations should invest in their people, particularly during periods of technological change. Yet that investment is not costless. Training requires resources. Supervision requires managerial attention. Persistent capability gaps may generate rework, errors, delays and additional burdens for colleagues. There is also an opportunity cost: time and resources devoted to repeatedly compensating for one capability gap cannot simultaneously be used elsewhere.

At some point, responsible management may require comparison between alternative courses of action.

Conceptually, the organisation may compare:$$E{V}_{develop}$$​

with:$$E{V}_{redesign},E{V}_{redeploy},E{V}_{separate}$$​

The expected value of continued development might be represented heuristically as:$$E{V}_{develop}=(P_{close}\times V_{future})-(C_{training}+C_{management}+C_{errors}+C_{opportunity})$$

where $P_{close}$​ represents the likelihood that the capability gap can be closed within the relevant timeframe, $V_{future}$ the expected future contribution if development succeeds, $C_{training}$​ the cost of additional development, $C_{management}$​ the managerial and supervisory cost, $C_{errors}$ the expected cost or risk while the capability gap persists, and $C_{opportunity}$​ the value foregone because organisational resources are being committed here rather than elsewhere.

The same logic applies to the alternative courses of action. Role redesign may preserve valuable capability while changing how work is structured. Redeployment may allow the organisation to retain an individual’s strengths in a role for which they remain better suited. Separation may become appropriate where the capability gap remains consequential, the probability of convergence is low, reasonable developmental runway has been provided, and neither redesign nor redeployment offers a viable alternative.

This should not be treated as a literal accounting formula. Many of these considerations cannot meaningfully be reduced to dollars. Institutional knowledge, loyalty, team cohesion, fairness, organisational culture and the human consequences of employment decisions matter as well.

The framework is therefore better understood as a decision heuristic. Its purpose is to make one principle explicit: continued development, redesign, redeployment and separation are all organisational choices involving trade-offs and consequences for multiple stakeholders.

Leaders therefore carry obligations in both directions. They have a responsibility to give employees a genuine opportunity to grow, but they also have responsibilities to colleagues, customers, stakeholders and the organisation as a whole.

There may eventually be circumstances in which continued investment produces insufficient improvement to justify its wider costs. Recognising that possibility does not require organisations to become less humane. Indeed, humane performance management requires clarity. Leaving an individual indefinitely in a role where expectations continue to rise while their ability to meet those expectations continues to fall may ultimately serve neither the organisation nor the individual.

The Individual Perspective

The Capability Divergence Model also has implications for how individuals think about their own careers.

Employees should not wait for formal performance conversations to tell them whether their capabilities are beginning to fall behind. By the time a capability gap becomes visible through formal processes, divergence may already have been occurring for a considerable period.

Individuals therefore need to monitor not only how well they currently perform but also the direction in which their profession, organisation and work are moving.

The relevant question is broader than whether one can use AI.

It includes whether the individual understands the work deeply enough to evaluate what AI produces; whether they can recognise an answer that is plausible but wrong; whether they understand how their organisation actually functions; whether they can work confidently with data; whether they understand the analytical and methodological choices underlying their work; whether they can redesign processes rather than merely automate existing steps; whether they understand the governance and ethical implications of technology; whether they can communicate across different stakeholder perspectives; and whether they can exercise judgement when the correct answer is ambiguous.

These capabilities matter because AI can compress the distance between intention and execution. Tasks that previously required substantial effort may become relatively easy to produce. Yet when the cost of producing an output falls, the ability to determine whether that output is useful, correct, appropriate and responsible becomes increasingly important.

This leads to a more demanding form of self-assessment.

Am I becoming more capable, or merely becoming faster?

There is also a temporal dimension to this problem. Capability development takes time. There is a delay between recognising that a skill or form of judgement matters, learning it, practising it, becoming competent and eventually being able to exercise it independently. This can be understood as adaptation lag.

By the time an organisation formally requires a new capability, someone beginning from zero may already be behind.

The implication is uncomfortable but important:

The best time to develop the capability required by tomorrow’s organisation may be while today’s organisation can still function without it.

Waiting until a capability becomes mandatory reduces the developmental runway available. AI may make this problem more pronounced because the pace at which some forms of work evolve may exceed the pace at which conventional reskilling occurs.

The greater long-term risk is therefore not simply lacking a particular capability today. It is remaining on a trajectory in which the capability gap widens faster than it can reasonably be closed.

Implications for Organisations and Human Resource Management

Taken together, the Capability Divergence Model can be reduced to three questions.

First:$$\boxed{{\displaystyle G_t=R_t-C_t}}$$​​

Where is the individual relative to what the organisation currently requires?

Second:$$\boxed{{\displaystyle \mathrm{\Delta }G=V_O-V_I}}$$​​

Is that gap closing or widening?

Third:$$\boxed{{\displaystyle G_t>\theta }}$$​

Has the divergence become sufficiently consequential that intervention is required?

These questions correspond to three different dimensions: position, trajectory and consequence.

This distinction has implications for how organisations think about workforce planning and performance management. Traditional competency frameworks are often static. They describe what an individual should know or be able to do at a particular level or within a particular role. This remains useful, but it may be insufficient where the requirements of the work itself are changing rapidly.

Organisations may increasingly need to distinguish between current performance and future adaptability. An employee can be performing satisfactorily while simultaneously becoming less prepared for what their role is becoming. Conversely, another employee may be weaker today but developing at a rate that makes them a stronger long-term proposition.

A person who is behind but learning rapidly may therefore be a better long-term investment than someone who appears adequate today but has stopped developing.

This also suggests that talent management should pay greater attention to learning velocity, responsiveness to feedback, transfer of learning into practice and demonstrated capability growth. These characteristics may become increasingly significant where organisational requirements cannot be assumed to remain stable.

The model also cautions against equating workforce transformation with training provision. Courses, workshops and development programmes are important, but they are inputs into adaptation rather than evidence of its completion. The more meaningful measure is whether individuals can translate those opportunities into improved capability.

At the same time, organisations should avoid allowing the language of adaptability to become a justification for transferring all responsibility onto employees. If capability requirements are changing because leadership has chosen to transform the operating model, leadership also carries responsibility for communicating those changes, providing developmental runway and ensuring that employees have a realistic opportunity to respond.

Effective workforce transformation therefore requires reciprocal accountability. Individuals must engage with development. Organisations must create conditions that make development possible.

Perhaps the greatest challenge for human resource management will be avoiding two opposite errors. The first is premature separation: treating any capability gap as evidence that a person no longer belongs in the organisation. The second is indefinite accommodation: allowing consequential capability divergence to continue without confronting its effects on other employees, stakeholders and organisational performance.

The Capability Divergence Model sits between these extremes. It does not attempt to prescribe when an individual should leave an organisation. Rather, it provides a way of asking whether divergence exists, whether it is widening, whether sufficient developmental runway has been provided and whether the gap has become consequential enough to require a different response.

Limitations and Conclusion

The Capability Divergence Model is conceptual rather than empirical. The equations proposed here should not be understood as validated predictive formulas or as variables that can necessarily be placed on a common numerical scale.

Capabilities such as judgement, ethical reasoning, stakeholder awareness and institutional understanding are difficult to reduce to singular numerical values. Capability requirements may also change unevenly rather than linearly, and the relative importance of different capabilities will vary considerably across organisations, professions and roles.

The Capability Tolerance Threshold is similarly normative. Different organisations will have different levels of risk tolerance, resource availability, employment practices and institutional responsibility. A capability gap that is manageable in one setting may be unacceptable in another.

The model should therefore not be used as an automated scoring mechanism for employment decisions. Its value is analytical rather than predictive. It provides a language through which organisations can distinguish between present capability and capability trajectory, between organisational velocity and individual adaptation, between development opportunity and demonstrated development, and between a temporary capability gap and persistent divergence.

Future empirical work could examine whether these constructs can be operationalised meaningfully, how capability velocity might be assessed in practice, and whether measures of developmental trajectory provide additional explanatory or predictive value beyond conventional performance assessments.

Much of the contemporary discussion surrounding AI and employment focuses on which jobs technology may eventually automate or eliminate. That remains an important question, but another transformation may be taking place more quietly inside organisations.

AI may increasingly differentiate not simply between tasks that can and cannot be automated, but between individuals whose capabilities continue to evolve with their work and those whose capabilities progressively diverge from it.

This should not become an excuse to abandon employees whenever technology changes. Nor should lifelong learning become a slogan that places the entire burden of adaptation onto the individual. Organisations need to communicate expectations, invest in development, create opportunities to learn and recognise that transformation requires time. Individuals, in turn, need to take responsibility for understanding where their work is heading and begin developing before change forces the issue.

Perhaps the central question for organisations is therefore no longer simply:

“Is this person capable enough today?”

It may increasingly become:

“Is this person’s capability trajectory converging with where the organisation needs to go?”

For individuals, the corresponding question may no longer simply be:

“Am I still performing well?”

but:

“Am I developing quickly enough for the world of work I am moving into?”

AI may accelerate organisations. The deeper challenge is whether people are given both the opportunity — and whether they develop the willingness and capacity — to accelerate with them.

---

The Capability Divergence Model proposed in this essay is a conceptual framework intended to support thinking and discussion about workforce development, organisational transformation and AI adoption. It has not been empirically validated and should not be treated as an assessment instrument, mathematical prediction model, or substitute for appropriate human-resource, performance-management, legal or employment processes.