By Dr Dorottya Sallai
Few phrases in the AI debate are used as liberally, or as loosely, as “human-centred”. Many organisations proudly claim to take a human-centred approach to AI, yet few define how they commit to the principle in practice. Dr Dorottya Sallai, Associate Professor in the Department of Management at LSE and Chair of its AI Working Group, examines what a human-centred approach to AI demands from leaders, and why AI transformation is a leadership challenge.
As the leadership of corporate AI transformation is moving from CIOs to CEOs1, leaders increasingly find themselves facing a managerial dilemma: how to increase productivity, drive AI adoption and achieve return on investment (ROI) on AI investment while also placing people at the centre.
While two years ago, organisations mostly focused on trying to decide whether to invest in Copilot, ChatGPT or Claude; now, the majority have invested in off-the-shelf or bespoke technological solutions, worked out a clear AI strategy, hired some of the necessary technical talent, and turned their attention to adoption. Yet challenges remain. Today, leaders are struggling with how to define their roles in the AI race, how to align people across the organisation around a shared direction, and how to support their teams through this unprecedented technological change.
Although the narrative has shifted from the technology to the people, and many organisations started to use ‘human-centred AI’ as a slogan for their approach, the term is often used as a normative phrase without much clarity on what it means.
How is human-centred AI defined?
Understanding how AI affects the workforce should take priority before the enthusiasm for deploying it.
In management literature, human-centred AI is defined as AI that augments human capabilities while maintaining “human control, and considering the necessity, context, and ethical and legal conditions to promote individual and societal well-being”2. Consulting firms claim that organisations create the greatest value with AI when they invest in upskilling and redesigning roles3 so that people and AI agents complement each other4. According to McKinsey, a human-centred approach to leadership requires self-exploration, curiosity, data-informed decision-making and strong relationships5 – practices that strengthen leaders’ ability to navigate increasing complexity.
While these definitions are helpful, they offer leaders little practical guidance when they are facing critical decisions. HR is a good example, where research shows that firms have adopted AI faster than they have understood its effects on the workforce, which is likely to cause more problems than it solves besides undermining the human-centred agenda itself6. Understanding how AI affects the workforce should take priority before the enthusiasm for deploying it.
The mismatch between what is measured and what AI changes
Leaders tend to approach AI investment, especially the introduction of generative AI systems, as a technological challenge rather than as an organisational transformation and they measure it accordingly. In traditional digital transformation, the organisation implements a new IT system, trains people to use it, and the technological change is complete. AI is different. These are continuously adaptive systems that learn from data, can act autonomously, and are often opaque in their working methods and unpredictable in their outcomes.
While AI tools, such as chatbots, can generate content, summarise and evaluate information or provide advice, agentic AI systems complete a sequence of tasks and are designed to achieve business outcomes with very limited or no human involvement7. When AI agents take decisions and run processes end to end, the user’s activity turns from task execution to setting the intent of the action, judging whether the AI has produced the desired outcome and intervening if something goes wrong8. This means that integrating AI into workflows – particularly systems that act autonomously and produce outputs that cannot be fully predicted or audited – significantly changes decision-making, control, and agency, along with people’s role in producing the desired business outcomes.
As autonomous AI systems are integrated into an increasing number of organisational processes – including as members of teams in the form of AI agents – people experience more and more algorithmic management, where AI-powered systems track their data and other information to organise, assign, monitor, supervise and evaluate their work9. The increasing delegation of managerial functions to algorithms – in customer service, transport, logistics, banking and health care among others – create a work environment in which employees interact with an AI system rather than a human manager10. Hence while the technology potentially leads to more efficiency, it also has a significant impact on how people communicate, behave and feel in control in the workplace.
Yet, research shows that when organisations measure the success and the financial return on AI investment, they leave the effect on employee experience unexplored11. I see the same pattern in my work with executive leaders and leadership teams. Despite the human-centred narrative, most organisations measure the return on AI investment through adoption rates, time saved, output volume or cost reduction. Such measures serve a purpose, but the nature of the technology demands a more nuanced exploration what AI does to employee experience.
So, how is AI affecting employees?
In both 2025 and 2026, the most common personal uses of AI were therapy and companionship, followed by troubleshooting and ‘fun and nonsense’, in contrast to the administrative efficiency organisations have in mind. At work, BCG’s 2026 survey of nearly 12,000 employees found that 72% say AI has changed the skills expected of them, and 67% that it has absorbed their simpler tasks and left them with more complex work12. The bar for what counts as “good enough” is now higher for the majority of employees, requiring people to spend more time reviewing and correcting AI output or spending more time on making AI-supported decisions13. In my practice, I find that training and support are sporadic at most organisations, focused on AI literacy rather than on meaningful investment in helping people use the technology within their teams and business contexts. Hence, while most leadership teams focus on how AI can create more productive employees, people focus on addressing their loneliness, challenging tasks, their entertainment needs and try to manage the impact of AI on their work responsibilities and career prospects.
It is time to change how organisations approach AI. AI systems should be configured around the employees who use them and centred on improving their experience, so that people produce better outcomes in more rewarding roles. This also requires employees to be involved in the selection, design, testing and introduction of AI systems. Most organisations do the opposite. Invest in the technology and then ask staff to adjust their working practices to AI systems that take no account of their context, skills and needs.
By underinvesting in training, support, as well as the necessary bespoke adaptation of AI products to the organisation’s real objectives, many organisations fail to use AI as an innovation device and use it instead as a sticking plaster over existing problems. A human-centred approach adjusts the technology so that people excel in their jobs, treating AI as a means of creating value rather than as an instrument in the race to the bottom.
How organisations are “human-washing” AI rollouts
While the nature of roles, processes, responsibilities and human agency is changing, the focus remains limited to productivity, which often results in the surveillance of employees. There is a real risk that organisations end up using AI to measure and monitor people more intensively than before, while claiming to follow a human-centred approach.
Nevertheless, mislabelling is the smaller problem. Algorithmic management tends to erode transparency and predictability, along with the complexity and problem-solving that make a job worth doing14. Consequently, when the term is used without the meaning, human-centred AI becomes a form of ethics-washing, or “human-washing”, which companies can readily claim to endorse without any pressure to demonstrate how they implement it15.
What should a human-centred approach to AI look like in practice?
To lead successful AI transformation, leadership teams should shift their initial focus from productivity gains and cost cutting to the experience, training and support of employees as well as to the collection of meaningful metrics about the organisational and human impacts of AI in the workplace.
Rolling out a technology that drastically reshapes how people work without paying attention to how it does can only lead to failure and wasted investment.
Collecting qualitative and quantitative data on how AI systems affect people will not only allow leaders to measure ROI more accurately but also help to deliver transformation that will lead to competitive advantage. Putting people at the centre means prioritising employee experience and approaching technology to an end and not the end in itself.
A lot of AI investment is done for the wrong reasons: the intention is to catch up with market pressure rather than differentiation through innovation. Many organisations fail to answer the question of what objectives they want to reach, or pain points they want to solve with their AI investment.
Rolling out a technology that drastically reshapes how people work without paying attention to how it does can only lead to failure and wasted investment. To support people – including managers – through this shift, change management should be integrated in every AI transformation as a strategic priority.
A real human-centred approach starts with the leaders, who create culture, norms and values with their own role-modelling behaviours. If leaders just focus on using AI systems to get people to better perform in their jobs without paying attention to how technology changes the experience that people have in their workplace, then human-centred transformation remains a slogan rather than a genuine management principle.
Good intentions are not enough. Systems introduced in good faith can still cause real harm to employees, clients and customers. Organisations cannot put people at the centre by announcing that they have. Leaders must understand the technology well enough to know when its output can be trusted, how it affects people’s roles, experiences, communication and relationships, and which decisions should never be handed to it. Leading in a human-centred way is difficult and requires the courage to resist market pressure and do what is right. That is the job.
