Security awareness training in the age of AI phishing

By Greg Markowski / Jun 8, 2026 / Cybersecurity & Compliance

Your firewall did not fall for the email. Karen in accounts did. That is not a criticism of Karen, it is the whole problem with cyber security in 2026: the technology has become genuinely hard to break, so attackers have gone back to working on people. A good security awareness training program is now one of the highest-return controls a small business can put in place, and AI is the reason.

The old advice was to teach staff to spot typos, dodgy sender addresses, and too-good-to-be-true offers. That advice is now close to useless. AI writes flawless emails in your CEO’s tone, clones a voice from a few seconds of audio, and builds a convincing fake invoice in the time it takes to read this sentence. The tells are gone. What is left is judgement, and judgement can be trained.

9 in 10

Breaches that involve a human element somewhere in the chain

Seconds

Time AI needs to clone a voice convincing enough to fool staff

Ongoing

Awareness is a habit, not a once-a-year video

Why annual training does not work

Most businesses tick the awareness box with a one-off induction video and a slideshow once a year. People click through it, pass the quiz, and forget it by lunch. Meanwhile the threats change monthly. A program built around a single annual event trains people to treat security as paperwork, not as part of the job.

What works is little and often. Short, regular touchpoints keep the topic alive. Real phishing simulations, sent to staff without warning, turn an abstract risk into a memorable moment when someone realises they nearly clicked. The goal is not to catch people out. It is to build the reflex of pausing before acting on an unexpected request.

What AI changed about the threat

Business email compromise used to rely on volume and luck. Now it relies on precision. An attacker can scrape your website and LinkedIn, learn your finance manager’s name and your suppliers, and generate a payment-redirection request that references a real project. Deepfake audio means a phone call from “the director” approving a transfer is no longer far-fetched.

The defence is not a better spam filter, though you need one of those too. It is staff who know the playbook: verify unusual payment requests through a second channel, never act on urgency alone, and treat a sense of pressure as a warning sign rather than a reason to hurry. That is what a modern program teaches, and it pairs directly with your technical cyber security controls.

What a program that works looks like

We build awareness the way you would build any habit: small, consistent, and measured.

  1. Baseline. Run an initial phishing simulation so you know where you actually stand, not where you hope you stand.
  2. Train little and often. Short monthly modules on current tactics beat a single annual marathon every time.
  3. Simulate realistically. Send safe but convincing test phishing, including AI-style lures, so the lesson lands without the damage.
  4. Coach, do not punish. When someone clicks, the moment becomes a quick, supportive lesson. Blame drives reporting underground, which is the opposite of what you want.
  5. Measure and report. Track click rates and reporting rates over time. Improvement is the proof the program is working, and it is exactly the evidence insurers and auditors want.

The return on a few dollars per head

Awareness training is among the cheapest controls you can buy and among the most effective. The maths is simple. A single successful invoice-redirection scam can cost tens of thousands of dollars. A program that turns your staff into people who pause and verify costs a fraction of that and reduces the chance of the scam landing at all.

This is the human side of the same story we keep coming back to: the controls that defend against AI-powered attacks are the established disciplines, run properly. Train the people, then the awareness program and your technical defences reinforce each other. We covered the foundation of access and data control in our piece on why your AI risk is really a permissions problem.

Frequently Asked Questions

What is a security awareness training program?
It is an ongoing process that teaches staff to recognise and respond to cyber threats like phishing and scams, usually through short regular training modules and simulated phishing tests. The aim is to build lasting habits rather than tick a compliance box once a year.
How has AI changed phishing and staff training?
AI removes the old warning signs. Scam emails are now grammatically perfect and personalised, and attackers can clone voices and create deepfakes. Training has shifted from spotting typos to building judgement: verifying unusual requests through a second channel and treating urgency as a red flag.
How often should staff do security awareness training?
Little and often works far better than once a year. Short monthly modules combined with periodic phishing simulations keep the topic current and build the reflex to pause before acting. Threats change monthly, so training should too.
Is security awareness training worth the cost for a small business?
Yes. It is one of the lowest-cost, highest-return controls available. A single successful scam can cost tens of thousands of dollars, while a training program costs a fraction of that and measurably reduces the chance of an attack succeeding.

Turn your team into your strongest defence

We run practical, ongoing awareness training built for the AI era, with real simulations and clear reporting. Talk to our Perth team on 1300 EPIC IT.

Book a Free Consultation

About the Author
Written by Greg Markowski, Founding Director of Epic IT, a CRN Fast50-recognised Microsoft Solutions Partner managing IT and cybersecurity for Perth businesses since 2003. Greg holds a Degree in Computer Science and a Diploma in Computer Systems Engineering from Edith Cowan University, and is ITIL certified.

Further Reading

Previous

Data classification: the control that actually makes Copilot safe

Return to News
Back to News
Next

Penetration testing vs vulnerability scanning: what is the difference?