Most managers were trained for a job that no longer exists.
The role they were prepared for involved a team of six or seven, work that arrived through them, and a reasonable expectation that the person who did the job well would eventually do the managing. That version of the role has quietly disappeared. Teams are bigger. Layers have been removed. A significant share of the coordination work that used to justify the position is now handled by software. And managers are expected to lead their teams through an AI transition they were never briefed on and are often more anxious about than the people they manage.
The result is measurable. Gallup found manager engagement fell from 31% in 2022 to 22% in 2025, while individual contributor engagement stayed roughly flat. DDI’s Global Leadership Forecast reports 71% of leaders experiencing increased stress, with 40% of those considering leaving leadership altogether, and employee trust in immediate managers falling from 46% to 29%.
The good news is that the skills that close the gap are learnable, and there are only about ten of them. Here’s each one: what it actually means, and where managers most often get it wrong.
1. Delegating to AI and knowing what not to
What it means. Deciding which work leaves human hands entirely, which comes back for judgment, and which never goes near a model. This is a managerial decision, not a technical one.
Why now. Microsoft’s 2026 Work Trend Index found 49% of Copilot conversations now support cognitive work such as analysis, problem-solving, strategic thinking rather than simple drafting. The line between “AI helps” and “AI decides” moved, and most teams have no agreed position on where it sits.
Where managers get it wrong. Treating it as a tooling question and delegating the decision downwards. Every person on the team then draws their own line, and quality becomes unpredictable. Set the line explicitly, in writing, for the two or three types of work your team does most.
2. Modelling AI use rather than mandating it
What it means. Using the tools visibly yourself, including in front of your team, including when it doesn’t work.
Why now. This is the most underrated finding in the current research. Microsoft found that when managers actively modelled AI use, employees reported a 17-point increase in the value they got from AI and a 30-point rise in trust in AI agents. Yet only one in four AI users said their leaders were clearly aligned on AI at all.
Where managers get it wrong. Announcing an adoption target while never using the tools in public. Teams read the gap instantly. Showing your team a prompt that produced something useless is worth more than a mandate.
3. Judging quality when output is cheap
What it means. Holding a standard when producing plausible-looking work costs almost nothing.
Why now. Gartner has a name for the failure mode, workslop, describing fast but poor-quality output produced because people are pushed to use AI for as many tasks as possible without the time or authority to assess whether the result is fit for purpose. Gartner’s counsel is to point AI at the most arduous, friction-filled work rather than the easy wins, and to measure effort saved rather than time saved.
Where managers get it wrong. Reviewing for presence rather than quality. If the document exists and reads fluently, it passes. Ask instead what decision this work supports, and whether it’s good enough to support it.
4. Coaching when you have too many people to coach
What it means. Structured, prepared development conversations that survive a large team, rather than ad-hoc chats when there happens to be time.
Why now. DDI found employees whose managers are effective coaches are nine times more likely to trust them. But average spans of control have widened from around eight direct reports in 2013 to roughly twelve today, and McKinsey’s work suggests a manager doing genuine coaching realistically tops out around six or seven.
Where managers get it wrong. Assuming coaching means availability. With twelve reports, availability isn’t a strategy, preparation is. Know before each conversation how that person takes feedback and what will actually land.
5. Redesigning the work, not just distributing it
What it means. Looking at a process end to end and asking whether it should exist in its current shape, rather than allocating its existing steps more efficiently.
Why now. Gartner is direct that this beats technical AI fluency: the organisations getting real value are prioritising people who can redesign whole processes over those who optimise individual tasks, and recommend recruiting for judgment and critical thinking ahead of technical skill.
Where managers get it wrong. Automating a bad process and calling it transformation. Faster nonsense is still nonsense.
6. Translating uncertainty into something a team can act on
What it means. Turning ambiguous strategy into this week’s actual priorities, and being honest about what isn’t decided yet.
Why now. In 2025, 71% of companies said it had become more crucial for leaders to operate amid constant change, up from 58% the year before.
Where managers get it wrong. Waiting for clarity before communicating. The team fills the silence with worse information than you’d have given them. Say what’s known, what isn’t, and when you’ll know more.
7. Managing your own state
What it means. Having enough capacity left to respond well in the moments that decide how people feel about working for you.
Why now. The limiting factor for most managers isn’t knowledge. They can explain a feedback model perfectly well, then default to reactivity under pressure. Gartner now frames this as an organisational responsibility, arguing managers must be equipped to recognise the psychological effects of pervasive AI at work including in themselves.
Where managers get it wrong. Treating this as a personal wellness issue rather than a performance one. Your regulation is the ceiling on your team’s psychological safety.
8. Building psychological safety deliberately
What it means. Making it demonstrably safe to raise a problem, admit an error, or say an idea isn’t working, through what you do when it happens, not what you say in advance.
Why now. Microsoft found employees were 1.4 times more likely to be high-frequency users of agentic AI when their manager created safety around experimentation. Safety isn’t a wellbeing initiative; it’s the precondition for every other change you’re trying to make. SHRM’s State of the Workplace 2026 found 46% of CHROs naming leadership and manager development their top priority for the second year running.
Where managers get it wrong. Declaring an open-door policy and then reacting badly the first time someone walks through it. One visible response to bad news sets the norm for a year.
9. Handling conflict between people who genuinely disagree
What it means. Surfacing and working through disagreement rooted in real differences of values, style or background not smoothing it over.
Why now. Workplace conflict is at record levels; UK conciliation service Acas found 44% of workers experienced conflict in the past year, with nearly a third involving line managers. Edelman’s 2026 Trust Barometer found 70% of people unwilling or hesitant to trust someone with different values or background, and 34% saying they’d put less effort into a project led by someone with different political beliefs.
Where managers get it wrong. Managing the temperature instead of the substance. Conflict that gets suppressed reappears as attrition.
10. Developing people when the middle rungs are gone
What it means. Building deliberate paths into leadership now that the roles people used to learn it in are disappearing.
Why now. Gartner flags that the decline of entry-level roles is fracturing the pathway into mid-level jobs. Those middle rungs were where future managers learned to manage. Meanwhile 41% of employees globally say their organisation has eliminated management layers.
Where managers get it wrong. Assuming succession takes care of itself because it always used to. It used to because the structure did the work. The structure is gone.
Turning the list into capability
Reading a list like this changes nothing, and manager development tends to fail the same way every time: it hands people a model and never gives them anywhere to use it.
The distinction worth holding onto is that these ten skills don’t all learn the same way. Process redesign, AI delegation and quality judgement are largely cognitive, they improve with good examples, clear thinking time and someone senior to argue with. Coaching, conflict, psychological safety and self-regulation are behavioural, and behaviour is stubborn. Nobody ever got better at difficult conversations by reading about difficult conversations. That half of the list only shifts through repetition in situations that carry some pressure, followed by an honest look at what you actually did.
Which is where most organisations get stuck. The behavioural half of this list needs somewhere to be practised, and there is rarely an obvious place for that in the working week.
This is the gap Jambar exists to fill. Whichever of the ten skills your managers need: coaching, conflict, psychological safety, adapting to AI, or something specific to how your teams work, we design a programme around that outcome rather than fitting you into a fixed course. We start from the result you’re after, build the experience backwards from it, and make sure the debrief converts what happened into something your managers can apply the following week.
Tell us what your managers are struggling with and we’ll design the programme around it. Get in touch.
