Capable managers often delegate least. They can do many tasks faster themselves, so taking work back feels sensible.
Over time it costs a lot. The team stops developing, decisions queue behind the manager and the manager's workload grows. Gallup's 2026 report found manager engagement fell from 27% to 22% in a year.
AI adds a new question. What should go to a person, what to an AI tool, and what stays with you?
Why managers under-delegate
The expertise trap. Doing it yourself is faster today. Teaching someone pays back every time after.
Perfectionism. If the outcome meets the standard, a different method is fine. Taking work back teaches people to stop trying.
Worry about accountability. You remain accountable, so delegate with support and check-ins. Build trust through progressively larger tasks.
No clear view of the workload. Without an audit, managers default to doing everything.
A five-step framework
1. Choose the task. Keep work that needs your authority, is politically sensitive or is too risky for current capability. Most other tasks are candidates.
2. Choose the person. Pick a task that stretches someone just beyond their comfort zone, and offer support.
3. Describe the outcome. "Write up the meeting notes" is a task. "Make sure the team has a clear record of decisions, owners and deadlines" is an outcome.
The outcome leaves room for judgement. Hackman and Oldham's research found that autonomy, ownership of a whole piece of work and feedback all strengthen motivation.
4. Agree check-ins. Match frequency to the stakes and the person's experience. Less experience means more check-ins.
5. Debrief. Ask what went well, what was hard and what they would change. This is where the development happens.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
Delegating in AI-changed work
AI changes what can be delegated and to whom. Microsoft's 2026 Work Trend Index found active agents in Microsoft 365 grew fifteen times in a year.
In Ju and Aral's 2026 experiment, people delegated 17% more work to AI partners than to human ones. Delegation improved text quality but made outputs more alike.
Apply the same discipline to AI. Be clear about the outcome, set checking points and decide who reviews AI output before it reaches anyone else.
Watch the workload too. Ranganathan and Ye's eight-month study found AI brought a faster pace, broader scope and longer hours. Delegation includes deciding what stops.
Redesigning who does what, and how work is checked, is Adaptation.
Building the habit
Audit your workload. List what only you can do. Most managers find more can move than they expected.
Start small. Delegate a low-risk task with a real development benefit. Use the framework and debrief.
Accept a different method. Judge delegated work against the agreed outcome rather than your own method.
Try this
Pick one task you do that someone else could own. Describe the outcome, agree check-ins and book a debrief. Track your workload and their capability for four weeks.
Then take the free Delegation Effectiveness Audit.