Adaptability is the capacity to build the capability a changing situation requires.

A regional operations manager has spent years refining a delivery process that works. Forecasting is reliable, supplier routines are familiar, escalation thresholds are understood and the team knows how to recover when ordinary disruptions occur. The situation changes materially when the company replaces its fulfilment model with a new operating system. Orders are now routed through a new platform, customer promises are calculated differently and several decisions that used to sit with local managers are automated. Within weeks, the manager's experience still matters, but some of the capability built around the previous system no longer produces the same result.

The obvious response would be to say that the manager needs to adapt. That description leaves the difficult part untouched because it says nothing about what the person has to reconstruct. The change has to be noticed accurately, its consequences have to be understood, and the manager has to work out which parts of existing capability still fit. Some knowledge may transfer directly because the underlying structure of the work has barely changed. A familiar process may need only a modest adjustment to remain effective under the altered conditions. Several capabilities may need to be combined differently because no single existing response now fits the task. A new capability may have to be learned while the work continues. An old routine may need to be withheld because the conditions that once made it effective have disappeared.

We define Adaptability as the capacity to build the capability a changing situation requires. The definition uses build in a systemic sense that includes the construction of an effective capability configuration. Building capability does not always mean acquiring something new. It can involve mobilising capability already possessed, transferring it into a new setting, recombining existing elements, refining them, developing what is missing, or retiring a response that no longer fits. The academic literature does not use this exact definition as a settled construct, yet research across adaptive performance, adaptive expertise, transfer, self-regulated learning and workplace learning strongly supports the underlying problem: changed or unfamiliar demands can disturb the fit between what a person can currently do and what effective performance now requires (Jundt and Shoss, 2023; Pelgrim et al., 2022; Frie et al., 2024).

That formulation also places a hard boundary around the concept. Adaptability cannot be inferred from enthusiasm for change, a broad skill inventory or one successful response to novelty. It depends on whether a person can recognise a changed requirement and construct a capability configuration that works under the new conditions. Domain knowledge, practical expertise, feedback and opportunity all shape whether that construction succeeds.

Adaptability is a capacity, adaptation is a process

Academic research has used adaptability, adaptation and adaptive performance inconsistently enough that the terms can easily collapse into one another. The distinction becomes much cleaner when each is assigned a different level of analysis.

Adaptability refers to the capacity that makes successful adjustment possible across changing or unfamiliar demands. Adaptation refers to the process through which a person responds after those demands change. Adaptive performance concerns the behaviour or performance that becomes observable under the altered conditions. This separation is consistent with major organisational traditions even though no universal terminology is enforced across the literature (Baard, Rench and Kozlowski, 2014; Jundt, Shoss and Huang, 2015; Jundt and Shoss, 2023).

The distinction matters because the same visible result can arise from very different underlying situations. An employee may perform well after a software change because they recognised the new requirement quickly and reconstructed their workflow. Another employee may perform equally well because they had already used the software extensively elsewhere and needed almost no adjustment. The performance result is similar, while the evidence about adaptability is different.

Pulakos et al. (2000) helped establish adaptive performance as a multidimensional workplace criterion, identifying demands ranging from learning new tasks and technologies to interpersonal, cultural and emergency adaptation. Later research strengthened the view that adaptive behaviour varies with the specific change being faced and does not form one uniform type of performance (Jundt, Shoss and Huang, 2015). A person who handles an interpersonal disruption well has not thereby demonstrated the same capacity under a major technical or domain-specific change.

The expression of the capacity also varies within the same person across different types of change. Jundt and Shoss (2023) argue for studying adaptation as an in-situ process because the same individual can respond differently across successive changes. What looks like stable adaptability in a self-report may therefore interact strongly with task structure, prior experience, available resources and the kind of capability gap created by the situation.

Several neighbouring constructs capture parts of the same territory. Cognitive flexibility concerns appropriate adjustment of cognition and behaviour as contingencies change, but changing strategy is useful only when the change improves fit (Uddin, 2021). Resilience focuses on maintaining or recovering functioning under adversity, while adaptation can be required by a benign change that creates no meaningful adversity. Learning agility emphasises learning from experience and applying that learning in new situations, while some adaptation episodes can be solved through capability already possessed (Milani, Setti and Argentero, 2021; Milani et al., 2024). Career adaptability operates at a broader vocational horizon through resources such as concern, control, curiosity and confidence; it does not describe the complete process through which a person restores task capability after a specific change (Savickas and Porfeli, 2012; Stead, LaVeck and Hurtado Rúa, 2022).

These constructs remain useful because they illuminate resources and processes that can support adaptation. None provides a complete substitute for the capability-building definition because successful adaptation still depends on establishing what the changed situation now demands and bringing capability into sufficient alignment with it.

The first difficulty is recognising what changed

The operations manager may initially interpret the new fulfilment platform as a straightforward technical upgrade. The interface is unfamiliar, so training on the interface appears to be the obvious capability requirement. Performance problems continue after the training because the larger change concerns how decisions are distributed across the workflow. Local managers now receive different information, intervene at different moments and need new criteria for knowing when automated recommendations deserve override.

The example exposes a basic feature of adaptation: capability cannot be rebuilt intelligently until the person has recognised what changed and diagnosed why the previous response no longer fits. Jundt and Shoss (2023) place detection and diagnosis near the beginning of the adaptation process for exactly this reason. A person has to recognise a meaningful change in the task environment, interpret its cause and determine what strategy, knowledge or skill the altered situation requires.

Detection is less reliable than the existence of feedback might suggest. Performance can deteriorate gradually enough to be attributed to noise, temporary disruption or other people. Familiar routines can continue producing acceptable results for long enough that the underlying change remains hidden. Metacognitive research adds another problem because people are imperfect judges of what they understand and what they do not. Yang et al. (2023), in a large meta-analysis of metacomprehension, show persistent gaps between actual comprehension and judgments about comprehension, alongside evidence that calibration can improve under some interventions.

Expertise changes this monitoring problem in two directions because it strengthens recognition while also stabilising familiar interpretations. Deep knowledge allows experts to recognise structural features that novices cannot see, which makes diagnosis faster and more precise in familiar domains. The same success history can create stable schemas and expectations that reduce attention to disconfirming information. Zhang, Harrington and Sherf (2022) review conditions under which expertise can inhibit effective advice and feedback seeking, while the broader cognitive-entrenchment tradition explains why efficient expert representations can become restrictive when the environment changes substantially (Dane, 2010).

Diagnosis therefore matters independently of learning speed because the wrong learning target can deepen the original mismatch. A person can learn quickly after identifying the wrong capability gap and become highly efficient at solving a problem that the situation never required them to solve. A sales team may interpret falling conversion as a persuasion problem and invest in pitch training while the real shift lies in customer qualification. A manager may treat a conflict as a communication deficit while the underlying issue is an incentive structure that rewards incompatible behaviours. The adaptive problem begins with a model of what has changed, and that model remains provisional while evidence accumulates.

This gives the word requires in our definition real weight. The changing situation does not announce its capability requirement in a clean specification. The person has to infer it from performance, feedback, contextual knowledge and the structure of the task. Adaptability therefore includes the capacity to identify the capability problem before attempting to solve it.

The size and nature of that capability problem can vary considerably. Sometimes the gap is local: the person already possesses the necessary knowledge and procedure but needs to select a different response. In other cases, structural overlap with previous work is partial enough that transfer requires substantial reinterpretation. The most demanding episodes can expose missing knowledge, incompatible routines or a new configuration of responsibilities that cannot be restored through local adjustment. Research does not provide a validated general metric for measuring this distance, so a simple continuum from small adjustment to major reconstruction should be treated as a conceptual aid whose empirical status remains unvalidated.

Different forms of mismatch can also look similar in the final performance failure. A person may lack a capability entirely, possess it but fail to activate it, transfer an old response because the new case looks deceptively familiar, or understand the task while missing contextual knowledge that changes what competent action requires. These possibilities matter because the same observed error can demand very different developmental responses. Additional training is appropriate for a genuine knowledge deficit and largely irrelevant when the person already possesses the capability but continues selecting the wrong routine. Adaptability therefore concerns the relationship between diagnosis and reconstruction, not the amount of development activity that follows.

Building can begin with capability already possessed

The strongest qualification to a literal reading of build is that successful adaptation does not always require new learning. Sometimes the necessary capability already exists and the adaptive work lies in recognising that it should now be used.

Consider a procurement manager whose usual supplier-review process relies heavily on historical delivery reliability. A geopolitical disruption suddenly makes lead-time volatility far more informative than the previous averages. The manager may already know how to model volatility from an earlier role. The changed situation requires that capability to be mobilised at the right moment, not learned from the beginning.

Adaptive-performance process research explicitly allows people to leverage existing strategies, knowledge and skills after diagnosing a change (Jundt and Shoss, 2023). Cognitive-flexibility research similarly treats appropriate selection and switching as responses to altered contingencies (Uddin, 2021). Mobilisation therefore belongs inside capability building when building refers to constructing the configuration required by the situation, a meaning broad enough to include existing content as well as new development.

Transfer is a second route because capability developed in one setting can sometimes remain useful when surface conditions change. Its reliability depends heavily on structural similarity and domain knowledge. Gobet and Sala (2023) review a long history of cognitive-training claims and conclude that broad far-transfer effects have repeatedly proved much weaker than expected. Evidence from interleaving and contextual-interference research shows that varied practice can improve transfer under some conditions, while effects become less dependable as tasks move farther from the learning context and as applied settings introduce additional complexity (Firth, Rivers and Boyle, 2021; Czyż, Wójcik and Solarská, 2024).

Transfer therefore creates a disciplined question: what part of the previous capability still maps onto the new situation? Surface resemblance can encourage the wrong response when the causal or structural conditions have changed. A manager who handled one organisational restructuring effectively may recognise useful patterns in another restructuring, yet the stakeholder incentives, regulatory environment or technical architecture may make the old solution inappropriate. Adaptability requires both carrying capability forward and recognising where the transfer stops.

Recombination becomes relevant when no single prior capability is sufficient. An unfamiliar task can require elements that have previously been used separately to be assembled into a new configuration. Adaptive-expertise research provides convergent support for this kind of flexible construction because experts facing non-routine cases draw on domain knowledge without remaining confined to routine application (Kua, Teo and Lim, 2022; Gamborg et al., 2024). Recombination is less cleanly isolated as a distinct empirical mechanism than transfer or learning, so we use it as one useful form of reconfiguration while avoiding any claim that it is an established stage of adaptation.

Refinement sits closer to the existing response because the underlying capability remains broadly valid. The underlying capability remains valid while some part of its execution, timing, threshold or structure has to change. Jundt and Shoss (2023) explicitly describe strategy development and refinement after diagnosis. A new reporting requirement may leave an analyst's core modelling capability intact while changing validation rules and review checkpoints. The adaptive work is meaningful even though the underlying skill has not been replaced.

Refinement has its own failure mode because repeated optimisation can preserve a model whose basic assumptions are already invalid. Small adjustments are useful when the structure still fits. They become persistence when evidence indicates that the structure itself needs reconstruction.

Knowledge determines what can be adapted intelligently

Adaptability is sometimes presented as a general quality that becomes most valuable when prior knowledge runs out. The research supports a different relationship in which prior capability supplies the material from which adaptation is constructed. Prior knowledge and expertise provide much of the material from which adaptation is constructed.

Adaptive expertise is especially informative because its central problem is the flexible use of substantial domain expertise when familiar routines become insufficient. Pelgrim et al. (2022), reviewing reviews of adaptive expertise and adaptive performance, show a field organised around the interaction of knowledge, flexibility, innovation and performance. Kua, Teo and Lim's (2022) study of adaptive experts highlights recognition of knowledge limits, reflection, mentoring and supportive learning environments. Gamborg et al. (2024) similarly show that uncertainty can become an opportunity for adaptive learning without guaranteeing that the opportunity will be recognised or used.

Within the Unified Skills Map, Knowledge Domains and Practical Skills therefore form the starting capability from which adaptation proceeds. Knowledge helps a person recognise what is unusual, identify which evidence matters and understand which constraints are genuine. Practical Skills provide an available repertoire of actions that can be retained, transferred, modified or combined. A broader repertoire increases the number of possible responses while leaving diagnosis and selection unresolved.

Research on job and task rotation illustrates this distinction. Mlekus and Maier's (2021) meta-analysis of 56 studies and over 280,000 participants found positive relationships with several learning, development and flexibility outcomes, while many popular expectations surrounding rotation were less strongly supported than assumed. Exposure can broaden experience and repertoire without creating a general capacity that automatically selects the right response later.

Domain knowledge also sets a limit on far transfer. An adaptable financial analyst cannot enter a specialised clinical role and produce expert judgments through adaptability alone. Strong adaptive processes may help the person recognise what they do not know, seek relevant support and learn more effectively. They cannot remove the time and practice required to build specialist knowledge and procedure. This is why an employability model that celebrates adaptability while treating expertise as disposable becomes internally incoherent.

Expertise also remains incapable of guaranteeing adaptation when established knowledge is applied outside the conditions that made it reliable. The person may possess deep knowledge while applying it under conditions where its assumptions no longer hold. Adaptability and expertise therefore interact reciprocally: expertise gives adaptation something substantive to work with, while adaptability helps keep expertise usable when context changes.

Sometimes capability has to be developed, and sometimes an old response has to be retired

Some changes create a genuine capability deficit that cannot be solved through mobilisation, transfer or local refinement. The person does not possess the necessary knowledge, skill or procedure in a form that can be mobilised or transferred, so adaptation requires development while performance is already under pressure.

Jundt and Shoss (2023) include learning additional declarative knowledge or procedural skill when changed demands make existing competency insufficient. Research on self-regulated learning provides stronger evidence for mechanisms that support this development. Guo's (2022) meta-analysis found that metacognitive prompts improved self-regulated learning activity and learning outcomes in computer-based environments, with intervention design and feedback influencing the effects. Theobald (2021) similarly found positive effects of extended self-regulated learning programmes on regulatory strategies, motivation and academic performance.

Those findings support the development of constituent learning processes while leaving a domain-general adaptability intervention unproven. They also show why learning ability cannot define adaptability on its own. The adaptive episode includes deciding that learning is necessary, locating the missing capability and selecting a developmental route that fits the performance problem.

Workplace learning introduces additional constraints because the person may have to learn while continuing to deliver. Time is limited, feedback can be ambiguous and the organisation may provide uneven access to expertise. Frie et al. (2024) describe career-long expertise renewal as a dynamic process shaped by new expertise needs, feedback and social interaction. Groenier et al. (2025), in a realist review of adaptive-expertise development through work-based learning, identify integrative challenging contexts, reflective practice, interaction with others, guidance and learner characteristics as recurring elements. Their review is particularly useful because challenging work becomes developmental through mechanisms and conditions, with difficulty alone providing an inadequate explanation.

Adaptation can also require withdrawing capability that remains psychologically available. A procedure may have been reinforced by years of successful use and still become unsafe under a new system. Research on workplace unlearning is heterogeneous, making literal claims that people erase knowledge difficult to defend (Kim and Park, 2022). Habit research provides a firmer language because established responses can persist through learned cues even when intentions change (Verplanken and Orbell, 2022).

The academically safer interpretation is that a person may need to inhibit, revise or replace a response whose conditions of validity have disappeared. Old knowledge can remain accessible while the person learns that it should no longer govern action in this context. Auqui-Caceres's (2023) systematic review of double-loop learning reaches adjacent territory by examining revision of underlying assumptions alongside local behavioural correction.

The operations manager learning the new fulfilment model may therefore face several forms of work at once. Some expertise remains valid and transfers, while other routines require refinement before they can function under the new operating conditions. One decision process must be learned from the beginning. Another familiar escalation rule needs to be retired because the new system already resolves the condition automatically. The capability required by the changed situation is constructed from this mixture, with no clean replacement of the old by the new.

Action keeps changing the problem

Adaptation rarely unfolds from perfect diagnosis into a finished solution because changed situations often remain uncertain while people are already required to act. The first response therefore becomes part of the evidence available for understanding what the situation actually demands.

Jundt and Shoss's (2023) process model is explicitly recursive. Detection, diagnosis, strategy, learning and performance are loosely ordered, and later experience can send the person back towards an earlier part of the process. The operations manager may believe that a revised exception rule will restore performance, test it for a week and discover that delays are now concentrated in a different part of the workflow. The action has produced evidence that changes the diagnosis.

This recursive quality matters because adaptation under uncertainty may require provisional action before complete understanding becomes possible. A person can form a working explanation, choose an action that is sufficiently safe and informative, monitor what happens and revise the capability configuration as evidence improves. Adaptive-expertise research supports the importance of working with uncertainty while preserving high standards of performance, especially when routine scripts no longer provide a complete answer (Gamborg et al., 2024).

Feedback becomes developmental only when it can be interpreted. Katz, Rauvola and Rudolph's (2021) meta-analysis links supportive feedback environments with performance, feedback orientation and other work outcomes, while the evidence remains largely correlational. Guo's (2022) experimental synthesis shows more directly that feedback can moderate the effectiveness of metacognitive support. Feedback that arrives late, lacks credibility or is filtered through organisational politics can do little to improve the person's representation of what went wrong.

Reflection faces the same condition because its value depends on the quality of the evidence available for reconstructing what happened. Guo's (2022) separate meta-analysis of reflection interventions supports structured reflection under appropriate designs. Reflection can help reconstruct causal understanding and expose a weak strategy. It can also become retrospective storytelling when the person lacks credible evidence about what produced the outcome.

Adaptation is therefore better represented as continuing fit work than as a one-time switch from an old answer to a new one. Diagnosis shapes the response, action changes the available evidence, and the next interpretation can alter what capability needs to be mobilised or developed.

Change becomes developmental under specific conditions

A person who has adapted successfully once may finish the episode with additional knowledge, a broader repertoire, better awareness of their own knowledge limits or a stronger strategy for monitoring future change. That creates a plausible route through which one adaptation can change the starting conditions of the next. The evidence does not support an automatic compounding rule.

Georganta et al. (2023) provide unusually useful counterevidence because two laboratory experiments on team adaptation found no simple advantage from prior adaptation experience. Experience with an earlier adaptive episode did not reliably translate into superior later adaptation. The result challenges the familiar intuition that repeated exposure to change naturally makes people more adaptable.

What matters is what the experience allows the person to learn. Variation can be useful because repeated performance under identical conditions risks overfitting capability to one pattern. Contextual-interference research shows that varied practice can improve later transfer in some domains, while Czyż, Wójcik and Solarská's (2024) meta-analysis found weaker effects in applied settings than in laboratories. Schmidt, Kemena and Jaitner (2021) provide additional counterevidence against the assumption that increasingly large amounts of task variation produce increasingly large benefits.

Development therefore depends on variation that the learner can interpret well enough to distinguish changing features from stable structure. The learner needs enough prior structure to understand what changed, enough feedback to distinguish useful from weak responses and enough opportunity to try again. Challenge without these conditions can create confusion or defensive persistence and fail to produce adaptive learning.

Simulation is valuable for the same reason because meaningful requirements can be perturbed deliberately. Fernandez et al. (2022) integrate adaptive-performance and instructional-design research into a simulation-based training model that emphasises exposing learners to changing conditions. The model provides a strong design logic without proving that simulation automatically produces general adaptability. Transfer still has to be demonstrated outside the trained scenario.

The workplace can provide equally rich developmental conditions when challenging assignments are accompanied by guidance, interaction and reflection. Groenier et al. (2025) find these mechanisms repeatedly in their realist review of work-based adaptive-expertise development. Mentoring and coaching can contribute feedback, models and alternative perspectives, while their value depends on the quality of the guidance and whether the learner can apply it to real performance problems (Wang et al., 2022).

Taken together, this evidence supports a qualified developmental claim about what can realistically be strengthened through deliberate practice. Adaptability can be developed through constituent processes and through practice under changing domain-relevant conditions. Current research does not justify a generic course that claims to create durable, far-transfer adaptability across unrelated professions. Development becomes credible when it improves how people recognise, diagnose and respond to variation in settings where the relevant knowledge can also grow.

Adaptability belongs to the person, while adaptation happens in a system

The capability to adapt belongs partly to the individual, but observable adaptation depends on what the environment allows the person to do. This distinction prevents a serious attribution error in organisations: failure to adapt is not automatically evidence of low adaptability.

A manager can recognise a workflow failure, diagnose the required redesign and possess the knowledge to implement it while lacking authority to change the system. An employee can identify a capability gap and have no access to training, time or expert support. Incentives can reward preservation of the old routine even after performance evidence favours change.

Organisational research supports this contextual dependence by showing that adaptive behaviour changes with leadership, resources and work design. Bonini et al. (2024), in a systematic review and meta-analysis of leadership and adaptive performance, found meaningful relationships between leadership and employee adaptive performance across a large evidence base. Demerouti and colleagues' quasi-experimental work during organisational change shows that job-crafting interventions can influence change-related behaviour and outcomes (Demerouti et al., 2021). Dhondt, Kraan and Bal's (2022) longitudinal research similarly shows that organisational arrangements shape how technological change translates into skill use.

Recent work continues to connect autonomy, proactive job redesign and demands-abilities fit with adaptive performance. Lau and Aguirre Reid (2026) examine how work autonomy can support adaptive performance through job crafting and fit during crisis conditions. The study does not establish adaptability as a pure environmental product; it reinforces the need to separate individual capacity from the conditions that allow that capacity to become behaviour.

There are also limits to how much change a person can absorb. Beaulieu, Seneviratne and Nowell's (2023) integrative review of change fatigue identifies stress, exhaustion and burnout across environments characterised by repeated organisational change. de Vries and de Vries (2023) connect repetitive reorganisations with uncertainty, workload and change fatigue. Continuous disruption therefore has costs that cannot be reframed indefinitely as opportunities to demonstrate adaptability.

A serious account of adaptability must preserve joint responsibility. Individuals can develop stronger capacities for recognising and responding to change. Organisations determine whether changes are intelligible, resourced and actionable, and whether employees have enough information, authority and developmental opportunity to construct the capability now required.

Assessment has to introduce a real change

Adaptability is difficult to assess because many existing measures capture self-perception or broad adaptive behaviour without observing the process that gives the construct its meaning. Someone can report that they adjust easily to change, while the statement tells us little about whether they can diagnose a domain-specific capability gap or rebuild performance after an unexpected requirement shift.

Hissink et al. (2025) reviewed measurement instruments for adaptive expertise and adaptive performance in healthcare and found 19 instruments across 17 articles. Self-evaluation and job-requirement instruments dominated, while design scenarios, mixed-method approaches and collegial verbalisation were less common. The review also found substantial variation in conceptualisation, operationalisation and evidence for instrument quality. Although the healthcare scope limits generalisation, the measurement problem is directly relevant: adaptability-related constructs are frequently assessed without observing adaptation as it unfolds.

A credible assessment should therefore distinguish the capacity available before the episode, the adaptation process after change, the resulting adaptive performance and any capability that persists afterwards. These four levels answer different questions about what existed, what changed, what was achieved and what endured.

The strongest design starts with competent baseline performance and then introduces a meaningful perturbation. The change should invalidate some part of the previous response without making the entire task unrelated. Assessment can then observe whether the person notices the change, how accurately they diagnose its consequences, whether they mobilise or reconstruct appropriate capability, how they use feedback and whether performance recovers.

Repeated perturbations are especially important if the claim concerns capacity. One successful episode may depend on fortunate prior experience. Performance across several qualitatively different but domain-relevant changes provides better evidence that the person can repeatedly work with changing requirements.

Process traces can also separate mechanisms that a final score would hide. Detection latency, diagnostic quality, strategy revision, help seeking, learning rate and the recurrence of previously corrected errors can reveal different features of adaptation. A later transfer task can then test whether useful capability from one episode remains available when the next change shares structure without sharing all of the surface features.

This assessment architecture is a synthesis from adaptive-performance process research, adaptive-expertise measurement and transfer science, and it has not been validated as a complete battery. Its value lies in matching the construct more closely: adaptability should be evidenced by what happens when a previously sufficient capability configuration becomes insufficient and the person has to restore fit.

The assessment design also needs to account for the opportunity available to express an adaptive response. A participant who diagnoses the change correctly but is denied access to necessary information or tools should not be scored in the same way as someone who never recognised the requirement. Likewise, rapid recovery can reflect unusually rich support as much as stronger person-level capacity. Recording the assistance available during the episode helps separate the quality of the adaptive process from the environment that enabled or constrained its expression.

A higher-order Adaptability score would eventually require evidence that performance covaries across several kinds of perturbation after controlling for domain knowledge, prior exposure and task opportunity. Current measurement research is not yet strong enough to establish that structure. Until such evidence exists, assessment is more defensible when it reports the quality of detection, diagnosis, reconfiguration and learning across repeated situations alongside the final performance result.

AI can change the capability requirement before it changes the job title

Generative AI provides a current example of why adaptability cannot be reduced to learning a new tool. A role can retain its title while the distribution of tasks, information and judgment inside it changes quickly.

Noy and Zhang's (2023) preregistered experiment with 453 professionals found that access to ChatGPT substantially reduced completion time and increased judged quality on a set of professional writing tasks. Brynjolfsson, Li and Raymond's (2025) field study of 5,172 customer-support agents found significant productivity gains from generative-AI assistance, with larger gains among less experienced and lower-performing workers. These findings establish task-specific augmentation and heterogeneous effects across expertise levels. They do not establish that the workers became more adaptable.

The adaptive problem appears in the capability configuration around the technology. Some existing knowledge becomes easier to access through AI, while verification, workflow design, delegation, confidentiality, provenance and judgment about when automation should be trusted can become more consequential. Hettrich, Krings and Kock (2025) similarly find that AI support interacts with expertise in project-planning tasks while leaving expertise differences visible.

Successful use of AI in one workflow therefore demonstrates a local performance result. Evidence that AI-mediated work develops Adaptability would require something harder: later changes in how the person detects, diagnoses and responds to different capability disruptions, preferably after the original scaffolding or exact tool has changed. Current productivity research generally does not test that outcome.

The same uncertainty applies to expertise development because automation can change both access to expertise and opportunities to practise it. AI may expose less-experienced workers to patterns drawn from stronger performers, while automation may also reduce some practice opportunities through which expertise was previously acquired. Current evidence supports both mechanisms as possibilities without justifying a general prediction that AI will inevitably accelerate or erode human capability development.

Capability has to keep finding a new fit

Adaptability becomes necessary when the relationship between a person and a situation changes enough that the previous capability configuration no longer produces sufficient performance. The change may be objective, such as a new technology or regulation, or it may be unfamiliar relative to the person, such as a role requirement they have never previously faced. Complexity and uncertainty increase the likelihood that familiar routines will fail, while neither one defines adaptation by itself.

The work begins with recognition and diagnosis because a capability gap that is misunderstood can send development in the wrong direction. Building the required capability can then take several forms. Existing capability may be mobilised or transferred, several elements may be recombined, a valid routine may be refined, genuinely missing capability may be developed, and an obsolete response may need to be inhibited or replaced. These operations do not form a mandatory sequence, and action continually produces new evidence that can reopen the diagnosis.

Existing Knowledge Domains and Practical Skills determine what the person has available to adapt with. Adaptability can help preserve their usefulness as conditions change, while it cannot substitute for the specialist knowledge and practice that competent performance requires. The environment also determines whether a good adaptive response can be enacted, making opportunity, authority, feedback and learning resources part of any serious interpretation of adaptive success or failure.

Development follows the same logic because exposure becomes useful only when the experience can alter later capability. Exposure to change creates a developmental opportunity without guaranteeing development. Variation, feedback, reflection, guidance and repeated practice can help a person extract capability from an adaptive episode, while unsupported instability can reinforce weak assumptions or simply exhaust the learner. One adaptation can change the starting point of the next when useful learning is retained, but current evidence does not support automatic career-long compounding.

We therefore use build to describe the active construction of capability fit under changing conditions. In some situations that construction adds genuinely new knowledge or skill to the person's repertoire. In other situations, it rearranges, sharpens or withholds what the person already possesses. Adaptability lies in being able to work out which of those responses the situation requires and continue rebuilding fit as new evidence changes the problem.