Leadership interview questions to assess a candidate

These leadership interview questions test what a candidate does when something needs an owner and nobody reports to them: a new engagement, a team rule, an incident. Each is quoted from a Hirezen interview plan, with three signals of a strong answer and an example of a Yes, a Not sure and a No. None asks about managing a team.

Updated . Written by Hirezen.

A signal is what a strong answer shows; mark each one as you listen. The examples are written for this page, and each Not sure ends with the follow-up that would settle it. Grouping these five under leadership is Hirezen’s call; each plan lists its own skills.

1. Taking charge of something new

Asked about one customer engagement the candidate chooses at the start and talks about for the whole hour.

The question

“Take me through your first week. Who did you talk to, in what order, and what was the first thing you looked at with your own eyes instead of hearing about it?”

Taking charge of something new: each signal, and what Yes, Not sure and No can sound like
SignalYesNot sureNo
Names the people met in the first week by role, in an order with a reason behind it“The operations lead on day one, because she owned the queue. Then two of her team, before I met IT.”“Ops first, then IT, then the sponsor.”Ask: Why in that order?“I spent the first week in meetings with whoever was available.”
Watched the work happen, read real records or ran a query on real data inside the first week“On day three I sat with a claims handler for a morning and counted what she looked up.”“I got into their data early on.”Ask: Which records, and on which day?“The sponsor walked us through their process, and we worked from the requirements document.”
Sent the customer something written by the end of the week that the customer could correct“On Friday I sent a one-page note of what I thought the problem was. They corrected two lines of it.”“We shared our findings with them at the end of the week.”Ask: In writing, and could they correct it?“Our first document was the proposal, in week three.”

2. Setting a rule the team will follow

The question

“Now the standard. Your team will use AI to draft — that is already happening and I'm not going to stop it. Write me the rule. What is it allowed to do, what is it not, who checks, and what happens the first time something wrong gets published?”

Setting a rule the team will follow: each signal, and what Yes, Not sure and No can sound like
SignalYesNot sureNo
Puts the accountability on the publisher rather than on the tool“Whoever publishes it owns every claim in it, however it was drafted.”“We’d all be responsible for what goes out.”Ask: Who signs off on a page?“If the model invents a statistic, that’s a tooling problem, and we’d switch tools.”
Specifies who reviews and whether they have authority to block publication“The section editor reads every page before it goes out, and can stop it without asking anyone.”“The editor would look over AI drafts.”Ask: Can the editor stop a page going out?“Writers check their own drafts before publishing.”
Gives an incident path: how a wrong published claim is found, corrected, and logged“A reader reports it, we correct the page that day with a dated note, and it goes in a log we read every month.”“We’d fix it straight away.”Ask: How would you find out, and where is the fix recorded?“The rule covers what AI may draft, and stops there.”

3. Changing how a team works, without blaming it

The candidate has a week of run data from a build pipeline whose typical run takes 38 minutes, with a script that retries failed jobs.

The question

“Three pull requests were merged past a failing required check this week. Before you touch the retry script, tell me why people did that, what the retries have been doing to the team, and what you would change so nobody needs to.”

Changing how a team works, without blaming it: each signal, and what Yes, Not sure and No can sound like
SignalYesNot sureNo
Starts from what a red run costs its author — another 38 minutes, maybe 51 — and reads the bypass as a rational response to that“A red run costs them another 38 minutes. From where they sat, merging past it was reasonable.”“Thirty-eight minutes is a long time to wait.”Ask: Is that why they merged?“Merging past a red check is a discipline problem. I’d raise it at the team meeting.”
Treats the administrator bypass as evidence about the pipeline before treating it as a permissions problem“Three bypasses in a week tells me nobody trusts the check. I’d fix that before touching who can merge.”“I’d look at both the pipeline and the permissions.”Ask: Which first, and why?“First, remove admin merge rights so it can’t happen again.”
Explains the change to the team where they merge, including what to do when a quarantined test fails“A line in the pull request template: this test is quarantined, here’s its owner, here’s what to do if it fails.”“I’d let the team know about the quarantine.”Ask: Where would they see it, and what do they do when the test fails?“Once the quarantine is in place, the pipeline handles it.”

4. Running the room when something breaks

Asked on the night of a sale, about the checkout system the candidate designed earlier in the same interview.

The question

“It is 20:10. Payment authorization's median is 6 seconds and climbing, 9% of checkouts are failing, and basket reads are fine. Take me through the next fifteen minutes in order, and ask me for anything you want to see.”

Running the room when something breaks: each signal, and what Yes, Not sure and No can sound like
SignalYesNot sureNo
Declares an incident within the first minutes and says who commands, who works the problem and who talks to the business“I declare it now. I’ll command, one engineer works the provider, and the support lead keeps the business updated.”“I’d declare an incident and get people on it.”Ask: Who commands, and who talks to the business?Opens the provider’s dashboards and starts diagnosing, with no incident declared.
Tells support and marketing what customers will see before customers start asking“Support and marketing hear now that some payments will show as pending, and what to tell customers who ask.”“We’ll put out a status update.”Ask: When, and do support and marketing hear first?“Once we know the cause, we’ll tell customers.”
Sets a time to review the decision and says what would make them switch the mode back off“Pending mode goes on now, and we look again at 20:40. It comes off once latency has been normal for a while, the breaker is closed and the pending queue has drained.”“Pending mode goes on, and we keep watching.”Ask: When do you look again, and what turns it off?Turns pending mode on and sets no time to look at it again.

5. The review after a failure

The draft postmortem names an engineer as the root cause: they typed 3 as a rate limit, a reviewer approved it, and the change reached every region in 90 seconds.

The question

“Rewrite the draft's root cause, impact and action items. Which of the four items survive, what replaces the others, and what does the review say about the engineer who typed 3?”

The review after a failure: each signal, and what Yes, Not sure and No can sound like
SignalYesNot sureNo
Keeps the engineer out of the root cause, and asks what made 3 look right to them and to their reviewer“The root cause isn’t her. What did the screen show that made 3 look like a normal value?”“The review should be blameless.”Ask: What made 3 look right to her and to her reviewer?“Root cause: human error during a configuration change.”
Drops "be more careful" and "add more monitoring" as unverifiable, and replaces them with specific changes, each with an owner and a date“‘Be more careful’ becomes a check that refuses a limit cut by more than half. ‘Add more monitoring’ becomes an alert on rejected requests per customer. The gateway team owns both, by the 30th.”“Those action items are too vague.”Ask: What does each become, who owns it, and by when?Keeps “be more careful” and “add more monitoring” as written.
Turns "retrain the team" into a change to the tool or its output, and says what the training was standing in for“The training was standing in for a check. The tool should show the per-minute value beside the number typed.”“I’m not sure training fixes this.”Ask: What would change in the tool instead?“Everyone who pushes config does the training this quarter.”

What these questions leave out

Managing people. Motivating a team, delegating and handling someone who is underperforming are not covered here, and a manager’s interview needs its own questions for them.

Questions

What are good leadership interview questions?
Questions that hand the candidate something to own, in the room or in their past, and ask what they did first. “Take me through your first week” asks who they met, in what order, and what they looked at themselves. “What is your leadership style?” asks for a self-description, which gives the interviewer nothing to mark against a signal.
Can you assess leadership in someone who has never managed people?
Yes. These five come from plans for a forward deployed engineer, a content strategist, a DevOps engineer and a site reliability engineer, and none of them asks about direct reports. Each one reads how the candidate leads without authority over anyone: setting a rule the team follows, running an incident, changing a habit the team has formed.
Can I use these questions for other roles?
Yes, once the setup fits. The first-week and rule-setting questions work for most roles as written. The pipeline, incident and postmortem questions need materials like the ones in their plans: run data, a live scenario, a draft postmortem. Without those, ask about a time the candidate faced the same situation, and keep the signals.

Start from a plan

Each question links to the plan it comes from, with that plan’s other questions and their signals. In Hirezen you can use a plan for one of your roles, change its questions, and mark its signals during the interview. The free plan covers one open role and up to ten people.