Empathy

The most common complaint in exit interviews is about feeling unheard, ahead of compensation or workload. Employees leave managers, not companies, and the managers they leave are disproportionately those who fail to understand what their people are experiencing.

The pattern repeats in customer relationships. People will tolerate imperfect products and occasional mistakes if they feel the organisation genuinely cares. They will abandon superior offerings if they feel like a transaction rather than a person. These are not new observations. What has changed is the context. As organisations adopt AI systems that can answer questions, process requests, and even simulate concern, the question of what distinguishes genuine human connection from its appearance has become operational.

The nature of the capacity#

Empathy is the ability to understand and share another person's emotional state while maintaining the distinction between one's own experience and theirs. It requires precision because it is routinely confused with adjacent concepts. It differs from sympathy, which involves feeling for someone without necessarily understanding their perspective. It differs from compassion, which adds a motivation to help. It is a component of emotional intelligence but not equivalent to it.

Research distinguishes between cognitive empathy, understanding another's perspective or mental state, and affective empathy, actually experiencing some echo of the other's emotion. Both contribute. A person with only cognitive empathy might accurately read others but remain unmoved; a person with only affective empathy might share feelings intensely but struggle to understand the thinking behind them. The skill lies in the underlying process of perceiving and responding to another's inner experience, not in any particular external form, which varies across cultures.

What the manager studies found#

A large international study of nearly 7,000 managers across 38 countries found that those rated as more empathetic by subordinates also received higher performance ratings from their superiors. Teams led by highly empathic leaders show markedly lower turnover, higher engagement, and better objective performance.

In healthcare the evidence is unusually consistent. A systematic review pooled 28 randomised trials covering 6,017 patients. Seven of those trials tested empathic communication specifically, and across them the effect on pain, anxiety and satisfaction was small but real. The remaining trials tested positive framing, which is a related but different intervention, and the distinction is usually lost when this review is cited. Medical students who scored high on empathy subsequently had patients with better disease control and fewer complications.

The evidence includes qualifications. A cross-temporal meta-analysis of 72 samples of American college students found that empathic concern fell by 48 per cent and perspective taking by 34 per cent between 1979 and 2009, most of it after 2000. Empathy is not fixed and can erode. Behavioural economics also shows that vivid empathy for a single identifiable individual can lead to resource allocation that neglects larger but less salient needs. Empathy requires calibration with broader judgement.

How the mechanisms operate#

At the cognitive level, empathy provides more accurate models of what others intend or need, reducing friction and error. At the relational level, it builds trust: when people sense their inner experience is understood, they feel safe to surface problems, admit mistakes, and propose ideas. This creates a reinforcing dynamic, empathy generates openness, which provides more information, which enables more accurate empathy, compounding into relationship capital. At the motivational level, feeling another's distress creates motivation to relieve it. And the act of empathising requires stepping outside one's own perspective, which can check impulsive or self-serving decisions.

The AI relationship#

AI can simulate empathic expression with increasing sophistication. Studies comparing AI-generated and human responses to patient questions have sometimes rated AI responses as more empathetic, particularly when the human is overloaded. Yet research reveals an AI empathy paradox: users rate AI empathetic responses as effective, but when given the choice, they overwhelmingly prefer to receive empathy from humans. The words may be right, but knowing there is no genuine concern behind them limits trust.

The limitation is not merely psychological. AI lacks the experiential understanding that lets humans generalise from their own emotional life to novel situations, and it lacks moral agency: empathy in humans implies an ethical commitment not to cause more suffering, a commitment AI does not make. The risk for humans is that heavy reliance on AI for empathic functions could erode the capacity, which develops through the repeated practice of attending to others. If AI handles initial interactions and humans engage only with escalations, the everyday practice that builds empathic skill is reduced, a particular concern for younger workers. The more useful relationship positions AI as support, identifying who needs human attention, while humans remain responsible for the relationships that matter.

When empathy fails#

The capacity has failure modes. Empathic distress occurs when feeling another's pain becomes overwhelming rather than informative, leading to burnout; the remedy is better boundaries, not less empathy. Bias means empathy is more easily felt for those who are similar or individually salient, which can produce inequitable treatment; awareness allows correction. Misplaced action follows when empathy is not coupled with judgement about the appropriate response. Performative empathy, simulating concern for instrumental purposes, erodes trust when detected. And empathy can fail through cultural mismatch, which calls for cultural humility rather than abandoning the capacity.

Organisational implications#

In hiring and promotion, behavioural indicators of empathy have become explicit criteria for roles involving leadership, collaboration, or customer relationship, because technical skills can often be taught or supplemented by tools, but the capacity to connect is harder to develop and more consequential for outcomes that depend on trust. In leadership development, empathy does not transfer well through classroom instruction; it develops through practice in contexts that require it. In culture, the presence or absence of empathy at senior levels sets a tone that propagates through the organisation. The stakes are higher during technological change, when transitions generate emotional responses that affect how smoothly change proceeds.

The compounding structure#

Empathy strengthens over time rather than depreciating. Each successful empathic interaction builds relationship capital; experience broadens range; the capacity transfers across contexts, whether the other is a colleague, customer, patient, or family member. The conditions that threaten it, extreme stress, cultures that punish vulnerability, isolation, are environmental rather than inherent. As AI handles more of what can be automated, what remains for humans is increasingly defined by what automation cannot replicate: connection, care, and the trust that flows from genuine understanding. The relative value of empathy does not decline as technology advances. It concentrates.

Related SuperSkills research#

See also outsourced recognition and what stays human.

Key research and primary sources

• Sadri, G., Weber, T. J. and Gentry, W. A. (2011). Empathic emotion and leadership performance: An empirical analysis across 38 countries. The Leadership Quarterly, 22(5).
• Howick, J., Moscrop, A., Mebius, A. et al. (2018). Effects of empathic and positive communication in healthcare consultations. Journal of the Royal Society of Medicine, 111(7).
• Konrath, S. H., O'Brien, E. H. and Hsing, C. (2011). Changes in Dispositional Empathy in American College Students Over Time. Personality and Social Psychology Review, 15(2).

Every figure on this page has been checked against the source that reports it. Where a number could not be confirmed, it was removed rather than left standing, and what was removed is recorded in the build register.

About this research#

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.

How this research works · Reviewed quarterly · Found an error? Tell me and it is corrected on the page.