You made a mistake at work. Here is what to say next

Deal with the live problem before writing its history

For a routine, low-risk error, that could mean warning the recipient not to use the earlier version and checking the corrected document before resending it. For a safety, security, personal-data or other serious incident, follow the relevant reporting process immediately. Do not improvise a cleanup that destroys evidence or makes the situation harder to investigate.

Stay within your authority. You may be able to correct a draft, but not recall a payment, alter an official record or contact affected customers without the right involvement.

If you are unsure whether an error is serious, tell the relevant responsible person what you know and ask how to proceed. Waiting for complete certainty can leave others acting on information you already suspect is wrong.

This article concerns everyday work communication. It is not a replacement for professional incident handling or jurisdiction-specific reporting obligations.

Give a short first account

A useful first message covers four things: what happened, what is known about the impact, what you have done within your authority and what help or decision you need.

For a fictional internal planning report: “I found an error in the summary I sent at 10:00. One region's figures were counted twice, so the total is overstated. Please do not use that version. I am checking the source figures with Maya and will update you by 11:30, even if the correction is not yet complete. Has the summary been forwarded or used in another document?”

The time is an example of an update commitment, not a universal response target. Commit to what you can control: reporting progress. Do not promise a completed correction before you understand the problem.

Add a direct apology where appropriate: “I am sorry I sent the incorrect version.” You do not need to repeat it in every sentence or ask the recipient to reassure you.

Separate facts from the story you are telling yourself

“I used the wrong input tab” may be established. “I did it because I was distracted” may be your first interpretation. “Nobody checks these reports” is a broader claim that needs evidence.

A factual account helps other people respond. A self-accusation such as “I am terrible with detail” does not tell them which document is affected or how to prevent another error.

Equally, do not use process problems to erase your own contribution. “The file names were similar, and I selected the wrong one” is clearer than “The system made me do it”. You can own an action while still examining the conditions around it.

Keep uncertainty visible. “I have confirmed the error in this summary; I am still checking whether the same input reached the monthly report” is more useful than either claiming everything is affected or assuring everyone it is isolated without checking.

Correct the work and close the communication loop

Once the correction has been checked, make the replacement easy to identify using the organisation's normal versioning process. Tell the relevant people which version to use and what changed. Ask the appropriate owner to help trace further use if the earlier information has travelled.

Do not silently overwrite a shared record when people need to understand the correction or when retention rules require an audit trail. Do not spread sensitive details to a wider group in the name of transparency. The right audience is the people responsible for handling the issue and those who need an update.

Close with the current state: corrected, still under review or awaiting a decision. If the impact remains uncertain, say who is checking it and when another update is expected.

Replace “I will be more careful” with a specific change

Google's site-reliability guidance^[1] describes postmortems as a way to document incidents, understand contributing causes and identify preventive actions without personal blame. That comes from operating technical systems, not a study proving a particular apology will protect your career.

The useful principle for ordinary work is to examine the failure point rather than stop at a character judgement.

In the fictional reporting example, the failure may involve several things: an old input tab remained in the workbook, two regions had similar labels, and the final total was not reconciled against the source. “Check more carefully” does not identify which of those to change.

A better follow-up could be: “For the next reporting cycle, I will reconcile the regional totals against the source export before review. Maya and I will check whether the template can make inactive tabs clearly distinguishable.” One action has an owner and a next use; the other is a proposal to the person who can help assess it.

Avoid adding five approval steps for one harmless typo. Match the safeguard to the consequence and likelihood of recurrence. Check later whether it actually helps rather than treating the existence of a checklist as proof that the risk is gone.

If you fear the reaction

Not every workplace handles mistakes fairly. A clear report cannot guarantee a reasonable response or remove genuine consequences. If you are worried about retaliation, scapegoating or a formal allegation, seek appropriate support from a trusted manager, representative, union or qualified adviser.

Do not conceal harm or change records to protect your image. You can ask for support while still using the correct reporting route. Keep your own account factual and handle company information according to its rules.

Your next move

If a mistake is active, write the short first account now: what happened, known impact, immediate action, help needed. Save the detailed explanation for when the situation is contained and the facts are clearer.

The most useful thing you can demonstrate next is not that you never make mistakes. It is that people can rely on you to surface them, help correct them and learn something specific.

If competing deadlines contributed to the problem, ask what should move when your manager adds more work. For AI-assisted work, check the parts that could hurt you before sharing.

About this guide

Google Site Reliability Engineering: Postmortem culture, learning from failure. Supports learning from contributing causes and preventive follow-up rather than individual blame. The reporting example and communication wording are original CareerFood illustrations, not a formal incident-response procedure or a guarantee about workplace consequences.