Most organisations already accept that people are part of the security problem. The harder question is what to do about it. Another round of awareness training is the usual answer, and on its own it rarely changes the outcome.
What the evidence says, and what it does not
Verizon’s 2025 Data Breach Investigations Report puts the figure plainly: “Human involvement in cybersecurity breaches remained about the same as the previous year — 60%.” The report analysed 12,195 confirmed data breaches. Verizon 2025 DBIR key findings
The same report notes “a significant overlap between social engineering and credential abuse”. In practice, a stolen or tricked-out password is often the point where a human mistake turns into a breach. Verizon 2025 DBIR announcement
Read the 60% carefully. It measures involvement across the breaches Verizon analysed. It is not a finding that staff are careless, and it does not tell you the rate inside your own organisation.
Why awareness training alone falls short
Training changes what people notice. It does not change what a single click, approval or password reset is able to do. If one person acting alone can move money, grant access or open a production system, the control depends on nobody ever having a bad day.
That makes the more useful question for leaders a structural one: where does one person’s mistake lead straight to harm, and what stands in the way?
Design processes that survive one mistake
- Authentication that phishing cannot capture. CISA describes phishing-resistant multi-factor authentication as “the most secure form of MFA” and names FIDO/WebAuthn as the only widely available form of it. Start with administrators and finance approvers.
- Independent verification for payment and account changes: a callback to a number already on file, never one supplied in the request itself.
- Least privilege: each role gets the access it needs, reviewed whenever someone joins, moves or leaves.
- A reporting route that is quicker and safer than hiding a mistake: one address or button, and no blame for reporting early.
None of these asks people to be perfect. Each one means that a single error is caught by a second factor, a second person or a second check before it becomes an incident. CISA: Implementing phishing-resistant MFA
Train for decisions, not just recognition
Generic modules test whether someone can spot a suspicious email in a quiz. What matters more is what people do under pressure in their own role: the finance approver asked to rush a payment, the helpdesk agent resetting a password, the engineer with production access.
Scenario exercises built around those roles show where judgement holds and where the process needs another check. Measure observed behaviour, such as whether a request was verified before acting and how quickly it was reported, rather than course completion rates.
Where to start this quarter
- List the handful of actions in your organisation where one person can cause the most harm alone.
- For each, confirm whether a second factor, second person or independent check already exists.
- Move privileged and finance users to phishing-resistant MFA first.
- Run one role-based exercise and record what changed in the process afterwards.
What leaders should ask for
Boards and executive teams rarely need more training statistics. They need a short, repeatable view of where people-related risk sits and whether it is shrinking. Four questions usually get there:
- Which high-impact actions can still be completed by one person without a second check?
- What share of privileged and finance users are on phishing-resistant MFA today, and by when will it be all of them?
- In the last exercise, how many people verified before acting, and how long did reporting take?
- Which process changes came out of the last incident or exercise, and have they been tested since?
Answers to these give leaders something they can govern. A training completion rate does not.
The aim is not a workforce that never makes mistakes. It is an organisation in which one mistake is not enough.

