Why Good Leaders Sometimes Let Their Own People Fail

“You need to let him fail.”
“What??”
“That’s right.”
“I can’t. I’m responsible for the team’s success!”
“By covering his mistakes, you’re failing as a leader. You need to let him fail.”

When my coach first said this to me, it made no sense. I thought it was bad advice. After all, what good leader wants a direct report to fail?

In my case, my analyst owned our monthly cloud spend report. I reviewed it each cycle. Every time, I found mistakes—small ones, but they mattered. Finance and engineering used the report to make calls on hiring, OKRs, and team investments. I felt responsible for sending out accurate numbers.

And yet, my coach told me to stop fixing the errors and let my analyst ship it—then face the consequences, learn, and improve.

Why this feels wrong

Leaders get rewarded for preventing problems. So the reflex is to sweep in, correct, and protect. That instinct is natural, but it blocks learning. People judge their progress by how easy a task feels, even though harder conditions can produce stronger learning and transfer. That mismatch—comfort versus long-term mastery—misleads managers into removing difficulty and, with it, growth. (Gwern, bjorklab.psych.ucla.edu)

Another reason it feels wrong: we want our people to feel confident. Yet confidence grows when autonomy and competence increase together, not when managers repeatedly step in. Environments that support autonomy and competence raise intrinsic motivation; controlling or overly protective environments suppress it. (Self Determination Theory, ScienceDirect)

What happens when you keep rescuing

Rescuing creates dependency. Your report learns that quality only emerges after you edit, so they wait for you. Their judgment stalls. Your time disappears. Team throughput slows.

There’s also a perception cost. Constant, visible help can dilute others’ view of the performer’s ability, which undermines credibility over time. (Harvard Scholar)

Finally, the team misses the performance gains that come from engaging with errors directly. Allowing people to try, err, and correct builds durable skill and better transfer to new problems compared with error-avoidant approaches. (PubMed)

“Let them fail” doesn’t mean “be careless”

The goal is not to manufacture failure. The goal is to create conditions where people do real work, encounter mistakes safely, and learn fast.

Well-run teams make room for errors and talk about them openly. When people believe they can surface issues without punishment, they share more, learn more, and improve performance. (Massachusetts Institute of Technology)

A simple pattern you can use

1) Set the bar upfront.
Define the outcome, deadlines, acceptable error range, and escalation triggers. Ask: What’s the minimum standard for this to be shippable?

2) Right-size the blast radius.
Choose scopes where a mistake is recoverable—a report to a friendly internal audience before the exec review; a canary deployment before full rollout.

3) Coach in the open, not in the shadows.
Ask questions that force judgment: What changed versus last month? Where are the anomalies? What would make finance challenge this? Then let them decide and own the call.

4) Let the consequence land.
If the report goes out with an error, don’t quietly fix it at midnight. Support your report as they correct the record and explain the fix. Learning sticks when the person confronts the outcome.

5) Run a short after-action.
What happened, why it happened, and what changes next cycle. Keep it blame-free and specific. (Massachusetts Institute of Technology)

Back to the cloud spend report

Here’s what I changed:

• We wrote a one-page checklist: data sources, variance thresholds, and sign-off rules.
  
  
  

• We agreed on two checkpoints: mid-cycle spot check and final QA—but no line-item editing from me.
  
  
  

• If variance exceeded a set threshold, my analyst had to escalate within an hour.
  
  
  

• If an error slipped through, they owned the correction note to finance within the same day.
  
  
  

The next report still had a miss. Finance asked a pointed question. My analyst answered, corrected the sheet, and documented the fix. Two cycles later, error rate dropped and cycle time improved. That pattern repeats across domains because people learn more from generating answers, encountering mistakes, and correcting them than from watching someone else do the work. (Gwern, PubMed)

Why this works (and what the research says)

• Harder conditions lead to stronger learning. Training that introduces “desirable difficulties” looks slower in the moment but improves long-term retention and transfer. (Gwern)
  
  
  

• Allowing errors during training improves transfer. Meta-analytic evidence shows error management training outperforms error-avoidant methods for applying skills to new tasks; recent studies in complex domains replicate these benefits. (PubMed, PMC)
  
  
  

• People grow when autonomy and competence rise together. Support for self-direction and mastery increases intrinsic motivation, engagement, and persistence. (Self Determination Theory, ScienceDirect)
  
  
  

• Teams learn more when it’s safe to speak up. Psychological safety predicts learning behaviors like admitting mistakes, asking for help, and sharing ideas. (Massachusetts Institute of Technology)
  
  
  

Guardrails for “safe failure”

Ask yourself:

• What can we constrain so the cost of an error stays low—scope, audience, time window, or rollback plan?
  
  
  

• Where will a pre-brief and clear “done” criteria remove ambiguity?
  
  
  

• What data will tell us the learning is happening—error trend, time to independent delivery, rework rate?
  
  
  

If a task is compliance-bound, irreversible, or high-risk, scale down the scope first. Use simulations, sandboxes, or pilot runs to create the same learning loop with less downside. (PubMed)

Scripts you can borrow

• “You own the result. I’m here for questions, not edits. If X happens, page me immediately.”
  
  
  

• “Ship when your checklist is complete and variance is under Y%. If you miss the window, send a note and propose a new time.”
  
  
  

• “After delivery, we’ll do a 15-minute review: what worked, what didn’t, what changes next cycle.”
  
  
  

A wholesome takeaway

You help people grow by giving them room to try, miss, and recover—on purpose and within guardrails.

As a leader, you need to grow your direct reports for the long term, even if that means allowing short-term mistakes. A coach can help you build those skills. You can also try AI Coaching for continuous feedback. We’ve built an AI coaching assistant that does this job well. Give it a try, and tell us what you think.