You can see it going wrong. Someone on your team is three days into an approach you would not have taken, the deadline is close enough to feel, and every instinct is telling you to open the document and fix it yourself. Then the other voice arrives: if I jump in now, I've taken it back, and I'll be taking everything back forever.
The takeaway up front: stepping in is never free, and the cost is not your time — it is that ownership moves back to you and stays there. So the default is to wait. Three signals genuinely mean act now; three more feel identical and are only your discomfort wearing a suit. Telling them apart is most of what people mean by judgment in a manager.
Why stepping in feels free, and what it costs
When you intervene, the benefit is immediate: the work gets better, the risk goes away. The cost lands later and somewhere else, so it is never attributed to the decision that caused it.
- Ownership transfers. The moment you take the pen, the work is yours. If it goes badly afterwards that is your fault, not theirs — and they know it. People stop pushing hard on work they no longer own.
- The next escalation comes sooner. You have taught the team where your threshold is. Jump in at the first sign of difficulty and that is the point at which they will bring things to you from now on.
- The learning does not happen. Most of what makes someone senior is having been stuck and found a way out. Intervene at every stuck moment and the team stays one level below where it should be.
None of that means never intervene. It means intervention should buy something worth those three costs — usually a genuinely irreversible risk, and not much else.
The three signals that mean step in now
1. The mistake is about to become irreversible
The only truly urgent category: something is about to be sent, published, deleted, promised, or signed, and cannot be pulled back afterwards.
Reversibility is the cleanest test available, because it is a fact about the situation rather than a judgment about the person. A bad draft is fully reversible. A bad draft in the client's inbox is not. Money spent, data deleted, a commitment made in a meeting you were not in — those are one-way doors, and one-way doors are worth taking the pen for. The corollary matters as much: if the mistake is reversible, you have time, and time is what lets people work it out themselves.
2. They are stuck, not struggling
These look identical from outside and are completely different.
Struggling is effortful progress. The person is trying things, discarding them, and moving — slower than you would and in a direction you might not have picked, but moving. Struggling is how people learn, and interrupting it is how managers accidentally stall someone's development.
Stuck is when the next move does not exist for them. They have run out of options, or they are blocked on something outside their control: access they do not have, a decision only you can make, a person who will not reply. Stuck is not productive discomfort, it is just cost.
Ask rather than guess: "What's your next step?" If they have one, they are struggling — leave them to it. If the answer is vague, circular, or a version of "I'm not sure," they are stuck, and stepping in is right.
3. Something changed that they cannot see
You have information they do not: the client's priority moved, the budget is about to be pulled, another team is halfway through the same work, your own boss has changed what "done" means.
This is not really intervening — it is doing your job. Nobody can correct for a constraint that was never visible to them. Deliver the information plainly and early, then let them re-plan. What you must not do is watch the work go the wrong way because you did not want to interrupt.
The three signals that only look urgent
These produce exactly the same feeling in your chest as the real ones. That is the problem.
"They are slower than I would be." Almost always true, almost never a reason — you are comparing their first time to your hundredth. If the pace genuinely threatens a deadline, the honest conversation is about the deadline, not about taking the work back.
"They are doing it a different way." Different is not wrong. Before you act, be able to say what the other approach costs in rework, risk, or money. If you cannot name the cost, you have an aesthetic objection — and overriding a method on aesthetics teaches people not to bother thinking.
"It's visible, and it's making me anxious." Your boss asked about it; someone mentioned it in a meeting. Visibility raises your stakes, not the work's, and it is the most common trigger for taking back work that was going fine. Get an update you can relay, rather than seizing the work so you feel covered.
A blunt test for all three: if you took the work back, what specifically would be different about the outcome? If the honest answer is "it would feel better to me," that is not a signal.
The move that is not stepping in: set a checkpoint
Between "leave them alone" and "take it over" sits a third option, and it is the one to reach for most of the time: a named moment to look at the work together, agreed in advance.
"Keep going the way you're going. Send me what you've got Thursday morning and we'll look together — if it needs a change of direction, Thursday is early enough to make one."
That sentence leaves ownership where it is, puts a ceiling on how far the work can go wrong, converts your anxiety into a diary entry, and tells the person they are trusted until Thursday — the part that makes them work harder rather than quieter.
Set it at the last point a course correction is still cheap. If you cannot find such a point — if the next visible moment is also the irreversible one — the work needed more structure at handover. Naming the level of authority when you delegate prevents most of this; delegating at the right level covers how.
How to step in without taking the work back for good
When one of the three real signals fires, intervene in a way that returns the work afterwards.
- Ask before you tell. "Walk me through where you've got to" costs two minutes and often reveals a reason you had not considered. You then redirect from inside their thinking rather than over it.
- Say what you are doing and why. "I'm making the call on this one, because we can't unsend it once it goes to the client." The reason keeps it a decision about the situation, not a verdict on them.
- Never take it over silently. Rewriting it at night and handing it back without comment teaches people your feedback is not what it appears to be.
- Hand it back out loud. "That's the bit I needed to fix — the rest is yours, keep going." Without that sentence people assume the whole thing has been repossessed.
- Debrief in one question. What would have let you catch that yourself? That is what stops you needing to do it again.
When waiting has actually gone too far
Patience has a failure mode too, and it looks like a manager who never intervenes and calls it empowerment.
You have waited too long when the same problem has happened twice and you have said nothing; when someone has stopped asking for help because asking felt like admitting failure; or when a deadline has slipped and nobody has named it out loud. That last one is the giveaway — silence around a missed date is avoidance, not patience, and the team can tell the difference even when you cannot.
One distinction to hold on to: a piece of work going wrong is not the same as a person consistently falling short. A wobble on one project is handled with a checkpoint. A repeated pattern across different work is a performance conversation, covered in managing an underperforming employee.
Different people, different thresholds
The line is not the same for everyone. A new hire has no map: long struggle is not developmental for them, it is disorienting, and their checkpoints should sit closer together — daily at first, then weekly as their judgment shows itself. An experienced person on familiar work should go a long way without hearing from you, and shortening their leash for no reason reads as a loss of confidence even if you never say so.
The trade-off is worth naming: closer checkpoints buy a lower chance of a bad surprise and cost some of the ownership that comes from carrying something alone. Neither end is free. Choose deliberately, say which setting you are using, and move it as the person's track record changes — that is the everyday work of building trust, covered in the leading people guide.
FAQ
How long should I let someone struggle before I say something? Long enough to try two approaches, and no longer than the point at which a correction stops being cheap. Put a checkpoint just before that last safe moment rather than watching the clock.
Is it micromanaging to ask how something is going? Asking is not. Asking daily and unannounced about work you already set a checkpoint for is. Regular and expected reads as support; random reads as surveillance.
What if I step in and turn out to be wrong? Say so quickly. "You were right, I jumped in too early" costs nothing and tells the team decisions here can be revisited.
My boss wants me to intervene but I think the person should keep going. What now? Ask what outcome your boss is worried about, then offer the checkpoint as the answer: a named date when you will both see the work. That usually satisfies the concern without repossessing the task.
How do I get people to come to me sooner when they are genuinely stuck? Make being stuck cheap to admit. Ask "what's your next step?" routinely rather than only when things look bad, and meet "I don't know" with help rather than concern.
Waiting is the default, and it is the harder discipline. Take the pen only when the mistake cannot be undone, when someone is out of moves, or when you know something they do not — and set a checkpoint every other time. For more practical guides for new managers, see YouManageIt.
