Jose Francis Llenado (RPsy, MA.Org Psy, BS Psy)
Muriel Umurungi (BA Psych, cand JCU)
Carolyn Burr (M.Lead, Grad Dip Couns, B.A.)
AI readiness can be viewed at three levels: the organisation, the leader, and the individual. Across all three levels, a consistent theme emerges. Readiness appears to support appropriate reliance on AI, and the quality of that reliance is reflected less in how often AI is used and more in how it is used. In other words, AI readiness is not the same as AI intensity.
Drawing on this evidence, this article also proposes two ideas for further exploration. The first is human dispassion: the ability to evaluate AI outputs with balanced judgement rather than being driven by enthusiasm, convenience, trust, fear, or avoidance. The second is AI-optimised leadership: the deliberate allocation of leadership activities between people and AI, with each contributing where it adds the greatest value.
These concepts are not presented as established findings. Rather, they are interpretations that emerge from a synthesis of the literature reviewed. Together, they offer a way of thinking about how organisations, leaders, and individuals might work more effectively with AI, while also providing directions for future research and testing.
At the organisational level, AI Singapore's AI Readiness Index (AIRI) is a self-assessment tool that measures how prepared an organisation is to adopt AI. AIRI assesses key dimensions of organisational readiness, including organisational capability, business value, data preparedness, and infrastructure. Organisations are then placed on a maturity continuum ranging from "AI Unaware’ to ‘AI Competent’.
Psyrras and Kafa (2026) also look at AI readiness in schools. They ran a scoping review of 37 studies following PRISMA-ScR guidelines and identified seven dimensions of AI-enabled school leadership, grouped into three levels:
Individual- AI literacy, ethical leadership, and human–AI symbiosis.
Organisational- decision-making and administration, continuous professional development, and organisational readiness and culture.
Systemic- policy and governance.
Their organisational readiness dimension is somewhat complementary to AIRI’s pillars with coverage from infrastructure, strategy, and a culture open to innovation. They also draw on Holmström’s (2022) AI readiness framework from the business literature. On reliance, the review finds that leaders mostly use AI for administrative tasks rather than teaching-related ones. The literature they review raises concerns about over-reliance, loss of professional autonomy, and leaders lacking the skills to interpret AI-generated data. The authors conclude that AI should be a decision-support tool, not a decision-maker, and that leadership becomes a hybrid process in which humans keep final judgment. They also stress that individual readiness depends on the levels above it. AI literacy grows through professional development at the school level, and that in turn depends on policy and governance at the system level.
Sánchez-Ruiz et al. (2025) provide some direct evidence on how to assess reliance, at the learner level. They surveyed 686 Spanish university students and used k-means clustering and topological data analysis (Mapper) to identify four profiles: Low-Engagement, Active–Cautious, Balanced–Confident, and High-Use–Vigilant. The profiles differ in how often students use AI, whether they revise outputs, how much they over-rely on AI, and how ethically aware they are.
One finding is especially important for this topic. The Low-Engagement group, who used AI the least, reported comparatively higher over-reliance and the least revision of outputs. The High-Use–Vigilant group used AI the most, yet reported low over-reliance and strong revision habits. Frequency of use is therefore a poor stand-in for problematic reliance, and AI readiness cannot simply be read off AI intensity. What seems to matter is how people regulate their use, such as revising, verifying, and reflecting. Two cautions apply here - these are self-reported behaviours from a single cross-sectional study, and the profiles differ on several dimensions at once. Even so, the pattern runs against a common assumption that more use signals more maturity or that less use signals more caution.
The study also found that 29.9% of students accept incorrect AI output as it is. Self-rated digital competence was the strongest predictor of being in a higher profile (OR = 1.44 for Profile 3 and 1.67 for Profile 4, compared with Low-Engagement), and studying a STEM subject also predicted higher profiles. In other words, prior readiness is associated with the quality of reliance. Mapper showed that the profiles sit on a continuum rather than in fixed categories. On that basis, the authors propose AI CAR (AI Competency Assessment and Ranking) as a formative tool for placing students on the continuum and designing targeted support.
Reading across the three sources, and across the additional studies discussed below, we see a further quality that the literature points toward but does not name or measure. We call it human dispassion and put it forward as a proposed construct, emerging from our interpretation of the evidence, rather than already established. We define it as the ability to hold AI output at a critical distance and judge it on its merits, without being carried along by enthusiasm, convenience, or anxiety. Dispassion in this sense does not mean coldness or a lack of feeling. It means keeping emotion and trust in proportion, so that judgment stays with the person rather than drifting to the machine. The paragraphs that follow set out the evidence that, in our reading, points toward this quality.
Yan et al. (2025) systematically reviewed 34 studies on AI ethics in education published between 2020 and 2024. Data privacy, algorithmic bias, and transparency were the most frequently raised concerns, and the studies they reviewed show that AI systems can carry bias from training data and institutional records into grades and predictions. If the machine is not neutral, we would argue that any dispassion has to come from the people using it. Yan et al. make this point directly for learners, arguing that ethics education should help students question the perceived neutrality of AI recommendations and evaluate AI-generated content with informed scepticism. They also suggest that adaptive systems can nudge learners, especially younger ones, into over-reliance on AI-generated pathways, which weakens self-directed learning. Notably, students who use generative AI chatbots often show little concern about privacy or trustworthiness (Wamba-Taguimdje et al., 2024, as cited in Yan et al., 2025). In readiness terms, frequent use without critical distance is a gap, not a sign of maturity.
Dispassion also cuts the other way. Yan et al. (2025) note that teachers who understand AI tools show less algorithmic avoidance and anxiety, and develop more calibrated trust and self-efficacy. On our definition, a dispassionate user would neither hand judgment over to AI nor reject it out of fear. In our reading, both of those responses appear to be driven more by feeling than by evaluation.
The emotional data in Sánchez-Ruiz et al. (2025) are the closest the reviewed evidence comes to this idea, and they are consistent with it, although they do not measure it directly. The High-Use-Vigilant profile reported the highest curiosity (M = 4.51) and motivation (M = 4.41) but the lowest stress or anxiety (M = 2.24), alongside strong revision habits and low over-reliance. These students were engaged but composed. The Balanced–Confident profile also used AI heavily, but reported the highest stress (M = 3.33), guilt, and ethical doubt. The Low-Engagement profile showed the lowest curiosity and motivation and relatively low ethical concern, yet the highest over-reliance. Taken together, these patterns are consistent with the possibility that appropriate reliance depends less on how strongly people feel about AI than on whether those feelings are regulated. Enthusiasm without scrutiny, anxiety without resolution, and indifference all sit apart from the vigilant, even-tempered stance that accompanies the strongest regulation behaviours. The 29.9% of students who accept incorrect output as it is can be read as a lapse in exactly this kind of critical distance. These are cross-sectional associations, so they show that composure and good regulation occur together, not that one causes the other.
At the leadership level, Psyrras and Kafa (2026) describe a similar theme through their Human-AI symbiosis dimension. They argue that while AI can support decision-making, qualities such as ethical judgment, empathy, relational understanding, and contextual awareness remain distinctly human responsibilities. Leaders therefore need to engage critically with AI systems while retaining accountability for final decisions. The literature they review highlights risks at both extremes: too little trust in AI can limit its value, while excessive reliance can undermine professional judgment and autonomy. From this perspective, human dispassion can be understood as the capacity to navigate between these extremes, treating AI output as evidence to be considered rather than a verdict to be either accepted unquestioningly or dismissed outright.
Hutson (2024) adds an important qualification. Dispassion is not the same as detachment from human values. Hutson argues that empathy, ethical judgment, and nuanced understanding are precisely the skills AI cannot replace, and that they should shape how AI is designed and used in education. The paper also extends critical distance from the individual to the group. It notes that AI systems trained on homogenised data, and the teams that build and use them, are prone to groupthink, where the wish for agreement overrides careful evaluation. Hutson proposes that neurodivergent perspectives, with strengths such as attention to detail and pattern recognition, can help detect bias and challenge prevailing assumptions. For organisations, this suggests that dispassion might be built into readiness structurally, by involving people who think differently in reviewing AI outputs and decisions, rather than relying on each individual to stay sceptical alone.
Leadership and AI - The Case for AI-Optimised Leadership
The discussion so far has focused on AI readiness and reliance across organisations, leaders, and learners. The leadership literature takes this a step further by asking a practical question: what aspects of leadership should be supported by AI, and what should remain firmly human?
In response to this question, we propose the concept of AI-optimised leadership. This is not the idea of AI replacing leaders, nor is it the ad hoc use of AI tools by individual managers. Rather, it is the deliberate allocation of leadership activities between people and AI, with each contributing where it adds the greatest value. In this context, optimised does not mean using more AI. It means making thoughtful decisions about which tasks are best supported by AI and which depend on distinctly human capabilities. An organisation that uses AI extensively but without clear intent may be less AI-optimised than one that uses it selectively and purposefully. As with human dispassion, this concept is our interpretation of the literature rather than a framework proposed by the authors we reviewed.
Voigt and Timinger (2025) provide a useful starting point for thinking about this allocation. Their review of 72 studies suggests that AI is best suited to leadership activities that are structured, repeatable, and data-driven. Tasks associated with transactional leadership, such as monitoring performance, tracking goals, and providing routine feedback, appear particularly compatible with AI support. By contrast, leadership approaches that depend on empathy, vision, trust, coaching, and individual understanding remain much harder for AI to perform authentically. The authors also highlight an important risk: when decision-making becomes overly dependent on autonomous systems, leadership can drift into a situation where responsibility is unclear and accountability becomes weakened. Their conclusion is not that AI should replace leaders, but that the strongest outcomes are likely to come from hybrid approaches that combine AI capability with human judgement, transparency, and ethical oversight.
Santiago-Torner et al. (2026) help explain why such deliberate choices matter. Their review suggests that AI is no longer simply a tool that sits alongside leadership. It is increasingly becoming part of how decisions are informed, how information is interpreted, and even how authority is distributed within organisations. AI can extend analytical capability, influence how evidence is considered, and reshape the way leaders coordinate people and technology. If AI is already influencing leadership in these ways, the question is no longer whether organisations should involve AI in leadership, but how intentionally they do so. From this perspective, AI-optimised leadership is an attempt to consciously design that relationship rather than allowing it to emerge by accident. In our view, this is where the greatest opportunity lies: not in replacing human leadership, but in creating a thoughtful partnership between human judgement and AI capability.
The Gains of AI-Optimised Leadership
The gains cluster around the analytical and transactional side of leadership. Voigt and Timinger (2025) report that AI can increase efficiency, reduce some human biases and errors in decision-making, support more data-informed and individualised employee development, make evaluations more consistent, and improve the timeliness of feedback. Santiago-Torner et al. (2026) add greater predictive capacity and strategic precision in complex environments. These findings fit with the school leadership evidence in Psyrras and Kafa (2026), where AI is used mainly to reduce administrative workload and support data-informed planning, freeing leaders to concentrate on teaching and people.
Joshi (2025) offers the most optimistic account. In a broad review of AI in leadership and management, Joshi argues that hybrid models combining AI tools with human judgment improve both the speed and the accuracy of decisions, and reports one study in which a combined game-theory and AI approach to management training produced decisions that were 22% more accurate and 37% faster. Joshi also proposes a hybrid decision model in which the weight given to the AI recommendation, as against human judgment, changes with the AI’s confidence, the level of risk, and cultural factors. This is, in effect, a formal version of calibrated reliance: the AI’s voice should count for more in low-risk, well-understood decisions and for less where stakes and uncertainty are high. These figures need caution, however. The paper is a preprint that has not been peer reviewed, the cited gains come from a simulated training environment, and they are reported second-hand rather than from the original study.
Two further qualifications apply to all of these gains. First, the promise that AI makes leadership fairer is conditional. As Yan et al. (2025) show, AI systems can carry bias from their training data into evaluations, so a gain in consistency is not automatically a gain in fairness. Second, Santiago-Torner et al. (2026) note that the literature develops opportunities much less systematically than it develops risks. The gains are plausible and partly evidenced, but they are not yet well measured.
The Costs of AI-Optimised Leadership
While AI offers significant opportunities for leaders, the literature also highlights a number of challenges that organisations need to manage thoughtfully.
The first relates to relationships and trust. Voigt and Timinger (2025) suggest that an excessive reliance on AI-supported leadership could weaken some of the human connections that effective leadership depends upon. Santiago-Torner et al. (2026) explain this through established leadership theories, which emphasise trust, reciprocity, communication, and shared understanding. When decisions are influenced by systems that employees do not fully understand, leaders may find it harder to explain the reasoning behind those decisions, and trust can be more difficult to maintain. The research suggests that acceptance of AI-supported leadership depends heavily on whether systems are perceived as transparent, explainable, competent, and fair. Interestingly, people do not always trust algorithmic recommendations, even when those recommendations may be objectively better, a phenomenon known as algorithm aversion.
A second challenge concerns governance and accountability. As AI becomes more embedded in organisational decision-making, questions naturally arise about responsibility, transparency, and legitimacy. Santiago-Torner et al. (2026) identify several recurring risks, including opaque decision-making processes, uncertainty about accountability when things go wrong, and the potential for biased data or flawed models to influence outcomes. The authors argue that these issues cannot be solved through individual capability alone because they are characteristics of the wider organisational system. Effective governance, clear accountability, and appropriate oversight therefore become important parts of AI readiness.
A third challenge involves people, capability, and investment. Joshi (2025) highlights the risks of over-automation, diminishing human judgement, and capability gaps among both leaders and employees. Similar concerns about professional autonomy are noted by Psyrras and Kafa (2026). There are also practical investments that organisations must make if they want AI adoption to succeed, including leader education, AI literacy, explainability mechanisms, governance processes, and ongoing oversight. At the same time, Santiago-Torner et al. caution that organisations can create too much control. Governance is important, but overly restrictive approaches may reduce experimentation and slow innovation.
When viewed together, the benefits and costs of AI-optimised leadership have very different characteristics. The benefits often appear quickly and are relatively easy to measure, such as improved efficiency, faster feedback, or better analytical support. The costs, by contrast, tend to emerge more gradually and are often harder to quantify because they relate to trust, relationships, fairness, legitimacy, and employee experience. This distinction is our interpretation of the literature rather than a direct finding from any single study. If it proves accurate, organisations that focus only on productivity gains may underestimate the broader impacts of AI on leadership. A more balanced assessment would consider trust, perceived fairness, professional autonomy, employee wellbeing, and organisational culture alongside measures of efficiency and performance.
Conditions Under Which the Gains Outweigh the Costs
Santiago-Torner et al. (2026) suggest that successful AI-enabled leadership depends on three factors working together: organisational integration of AI, leadership capability, and effective governance. Their AI-Leadership Configurational Framework (ALCF) proposes that AI is most likely to enhance leadership when leaders have the skills to use it well and organisations provide appropriate safeguards, transparency, and accountability. Where these elements are misaligned, organisations face predictable risks, including over-reliance on technology, weak oversight, or stalled innovation.
This aligns closely with the multi-level readiness model discussed earlier. Organisational readiness creates the conditions for AI adoption, individual capability supports effective use, and governance provides the safeguards needed to maintain trust and legitimacy. On this reading, AI-optimised leadership is not separate from AI readiness. Rather, it may be viewed as one outcome of readiness when applied to leadership practice.
Human dispassion may also play an important role. Leaders need the ability to interpret, question, and, when necessary, override AI outputs rather than accepting or rejecting them automatically. In practice, the strongest case for AI-optimised leadership appears to be where AI supports analytical and transactional activities, while people retain responsibility for relational, ethical, and context-dependent leadership decisions.
Together, the literature suggests a layered approach to AI readiness and reliance. Organisational capability, leadership support, individual AI literacy, and appropriate governance create the conditions for success. Appropriate reliance then becomes visible through behaviours such as reviewing, verifying, challenging, and contextualising AI outputs. Across the frameworks reviewed, readiness is treated as a developmental process rather than a single score, highlighting the importance of ongoing capability building as AI continues to evolve.
The Human Factor in AI Readiness
When developing and implementing AI policies, organisations have an opportunity to place the human experience at the centre of AI adoption. This means thinking beyond AI as simply an efficiency tool and considering how people learn to use it, how they navigate ethical questions, and how confident they feel working alongside it. Concerns about AI replacing human roles are real and can influence patterns of AI use, as highlighted by Sánchez et al. (2025). Creating opportunities for learning, reflection and experimentation can therefore be an important part of building confidence with AI.
Hutson (2024) also draws attention to distinctly human qualities, such as empathy, that AI cannot replicate. This reinforces the importance of treating AI adaptability as both a technological and human learning process. Inductive approaches, where people learn and adapt through their lived experiences, provide a valuable way for organisations to support this transition.
AI competency can therefore be incorporated into an inductive learning cycle, where employees are encouraged not only to use AI, but also to reflect on their everyday experiences with it: what worked, what did not, where it added value, and where human judgement remained important.
This approach is reflected in our Seven Embers module, where everyday experiences become opportunities for deeper reflection and learning. Connecting these experiences with the observations of Sánchez et al. (2025) provides a way to consider not only how often AI is being used, but how and why it is being used.
Final Thoughts - Readiness Is Not Intensity
One of the messages emerging from the literature is that AI readiness should not be confused with AI intensity. The evidence reviewed suggests that using AI more often does not necessarily mean using it well. In Sánchez-Ruiz et al. (2025), some lower-use individuals showed greater over-reliance than those who used AI frequently, while high-use users often demonstrated stronger habits of checking, revising, and questioning AI outputs. If this pattern holds more broadly, the goal of readiness is not simply more AI adoption, but more thoughtful, responsible, and effective engagement with AI.
This shifts how organisations might think about success. Usage statistics and adoption rates can indicate whether AI is being used, but they say little about the quality of that use. Readiness may therefore be better understood as a combination of organisational capability, leadership practices, individual competence, and the behaviours that help people use AI well. These behaviours include verifying information, revising outputs, exercising judgement, and recognising when AI should not be used at all.
Viewed through this lens, AI readiness is not only about technology, infrastructure, or skills. It is also about how people relate to AI in practice. This article proposes two ideas that may help explain that relationship. The first is human dispassion, or the ability to engage with AI outputs with balanced judgement rather than blind trust, fear, enthusiasm, or avoidance. The second is AI-optimised leadership, which involves intentionally allocating work between people and AI so that each contributes where it adds the greatest value. Both ideas remain conceptual and require further testing, but they offer a useful way of thinking about what appropriate AI reliance may look like in organisations.
Future research could examine whether human dispassion is a distinct capability that contributes to better regulation behaviours, and whether organisations that deliberately design leadership roles around complementary human and AI strengths achieve better outcomes than those that adopt AI more informally. More broadly, there is an opportunity to combine organisational readiness measures, such as AIRI, with behavioural indicators of AI reliance to create a more complete picture of what successful AI adoption looks like.
The literature reviewed also has important limitations. Several frameworks remain theoretical, many studies rely on self-report data, and empirical evidence on AI-supported leadership is still emerging. Human dispassion has not yet been directly measured, and AI-optimised leadership remains a proposed interpretation of the evidence rather than an established model. Nevertheless, the studies reviewed point in a consistent direction. Readiness appears to support appropriate reliance on AI, and that reliance is best understood by examining how people use AI rather than how much they use it. If that conclusion is correct, organisations should focus less on maximising AI usage and more on developing the judgement, capabilities, and governance needed to use AI well.
The reference listing, simplifying language and flow included in this article were constructed with the assistance of AI. AI organized APA formatting based on the meta tagging of the bulleted journal articles and references.
This Article contains cited materials from existing evidence-based sources. All referenced content is cited using APA format to ensure academic rigor and transparency. A comprehensive list of references is provided at the base of the article, as well as in text citations is the article sections
References
AI Singapore. (n.d.). AI Readiness Index (AIRI). Retrieved October 9, 2026, from https://aisingapore.org/innovation/airi/
Holmström, J. (2022). From AI to digital transformation: The AI readiness framework. Business Horizons, 65(3), 329–339. https://doi.org/10.1016/j.bushor.2021.03.006
Hutson, P. (2024). Embracing the irreplaceable: The role of neurodiversity in cultivating human-AI symbiosis in education. International Journal of Emerging and Disruptive Innovation in Education: VISIONARIUM. https://doi.org/10.62608/2831-3550.1020
Joshi, S. (2025). Artificial intelligence in leadership and management: Current trends and future directions [Preprint]. Preprints.org. https://doi.org/10.20944/preprints202504.1429.v2
Psyrras, N., & Kafa, A. (2026). Artificial intelligence in school leadership: A multi-level framework for leadership capacity building. International Journal of Educational Management, 40(8), 124–138. https://doi.org/10.1108/IJEM-04-2026-0573
Sánchez-Ruiz, L. M., Llobregat-Gómez, N., Vega-Fleitas, E., & Moll-López, S. (2025). AI competency assessment and ranking: A framework for higher education. Applied Sciences, 15(22), Article 12248. https://doi.org/10.3390/app152212248
Santiago-Torner, C., Corral-Marfil, J.-A., & Tarrats-Pons, E. (2026). Artificial intelligence and leadership in organizations: A PRISMA systematic review of challenges, risks, and governance dynamics. Sustainability, 18(8), Article 4085. https://doi.org/10.3390/su18084085
Voigt, H. I., & Timinger, H. (2025). Artificial intelligence as a leader: A systematic literature review of the status quo and future challenges. Journal of Advanced Research in Leadership, 4(2), 16–34. https://doi.org/10.33422/jarl.v4i2.1191
Yan, Y., Liu, H., & Chau, T. (2025). A systematic review of AI ethics in education: Challenges, policy gaps, and future directions. Journal of Global Information Management, 33(1), 1–51. https://doi.org/10.4018/JGIM.386381