The most useful finding in The State of Organizations 2026, McKinsey's March 2026 survey of 10,000 senior leaders across 15 countries, is not the one being shared on LinkedIn. The one being shared is the framing about AI rewriting work. The one worth a second look is the operational one: organisations that adopted what McKinsey calls a "human-centric" leadership approach reported a 56% lift in employee satisfaction and retention, a 56% strengthening of trust, a 42% improvement in decision-making and a 40% improvement in organisational adaptability.
That is not a soft-skills argument. That is a retention argument, a decision-quality argument and an adaptability argument, in one survey, from the same set of leaders.
What McKinsey Actually Reported
The headline framing of the report is the "inside-out" leadership shift. McKinsey's own phrasing: "leading others also means leading oneself." The next frontier of leadership, in their data, is about cultivating inner motivation and the willingness to continuously relearn.
That sounds abstract until the numbers underneath it are read in sequence. The four uplifts (56%, 56%, 42%, 40%) are not from a wellbeing intervention or a values campaign. They are from leadership behaviour. The behaviours specified in the report cluster around four observable moves: how the leader sets a direction the team can describe; how the leader signals that interpersonal risk is safe; how the leader makes their reasoning visible when a decision is hard; how the leader keeps developing themselves in public, not in private.
The McKinsey framing connects directly to two findings the wider literature has held for years. Edmondson (1999), in Administrative Science Quarterly, defined psychological safety as "a shared belief that the team is safe for interpersonal risk-taking" and the construct has held up across two decades of replication. Google's Project Aristotle, the multi-year internal study of team effectiveness, ranked psychological safety first among the dynamics that distinguished higher-performing teams. The McKinsey 2026 numbers are, in effect, the same story in fresh data: when the leader's behaviour creates the conditions, retention, trust and adaptability move together.
Why The "Soft" Reading Misses
The default response to "human-centric leadership" in most boardrooms is a values document, a coaching budget and a 360. None of those move the McKinsey numbers on their own. The reason is mechanical: the behaviours are skills, the skills decay without rehearsal, and the rehearsal does not happen on the average leadership programme.
Two findings explain the gap. The first is Roediger and Karpicke (2006), in Psychological Science, on the testing effect: being asked to retrieve and use material increases long-term retention by around 50% compared with re-reading the same content. Most leadership development is the second kind: a recorded keynote, an LMS module, a 90-minute workshop with a feedback form. Eight weeks later, the leader who is in the difficult conversation is operating on instinct.
The second is Kirkpatrick and Kirkpatrick (2016), the standard four-level model of training evaluation. Most programmes measure Level 1 (reaction) and stop. The McKinsey uplifts sit at Level 4 (results). The gap between the two is where most leadership budgets quietly disappear.
What Most Organisations Do (And Why It Plateaus)
The familiar pattern, repeated across the last decade. A flagship leadership programme is procured. Senior leaders attend. The cohort photographs are taken. Twelve months later, the engagement survey is unchanged, the manager-of-managers layer has not moved, and the Chief People Officer is asked, in the next strategy session, why the investment has not landed. The honest answer is rarely supplied: the programme covered what the leaders should know, not what they should do, and the doing was never rehearsed.
The McKinsey report names this gap obliquely. In its words, leaders need new qualities to navigate "an increasingly complex, high-stakes world" in which AI handles more execution. New qualities are not new content. They are new behaviours, and behaviours are taught by being practised, observed, corrected, practised again.
What Works
The organisations whose leadership numbers are moving in the McKinsey direction tend to share three habits.
They distinguish between awareness and behaviour. Awareness is the keynote. Behaviour is what the leader does in the next difficult one-to-one. They name the second one, then ask their leaders to practise it.
They put the difficult conversation in front of the leader in a low-stakes setting before the real one arrives. Building on academic behaviour-change work from UCL, Cambridge and Bocconi, our own research found that immersive role-play with professional actors was around 20% more effective than passive modalities such as slide-show or video e-learning at moving observed skill. The same study found that participants overestimated their skill level before they were asked to perform: they believed they would handle the conversation well, until they had to.
They re-measure. Not attendance, not satisfaction. The real questions, in McKinsey's idiom, are whether retention has held, whether trust has lifted, whether the next decision was made better and faster than the last one.
This is the work behind our leadership development programmes and our immersive simulations. We rehearse the behaviour the leader will eventually need, with the help of professional actors trained in the dynamics, before the real version of it arrives unannounced.
The honest test for any organisation reading the McKinsey numbers this quarter is not whether their leadership framework mentions empathy, judgement, or purpose. It almost certainly does. The harder test is whether the leader who is currently sitting opposite a struggling direct report would, this week, hold the conversation in a way the McKinsey 56% would recognise. If the answer is uncertain, the inside-out shift is still on the slide deck. It is not yet in the meeting.
Book a free 30-minute consultation → or read about our research-backed approach.