Phishing in the Age of AI: Teaching Employees What to Verify

May 5th, 2027

Red shield and padlock over a circuit board pattern

For years, security training taught employees to look for bad grammar, generic greetings, and awkward phrasing. Those cues are disappearing. AI can write a message that matches a company's tone, references a real project, and uses the recipient's name correctly. It can also clone a voice from a short recording.

The old advice to "spot the phishing email" is no longer enough. Employees need to verify the request, not just the writing. That shift changes what training should teach and what procedures the business should enforce.

What AI Changed

Perfect Language at Scale

AI removed the language errors that used to give phishing away. Messages arrive with clean grammar, plausible context, and a professional tone. The employee can no longer rely on "it looks wrong" as a warning sign.

This does not mean every AI-written message is convincing. It means the presence of good writing is no longer evidence of legitimacy. Training should stop treating grammar as a security control.

Better Research and Personalization

Attackers can use public information, social media, and previous email threads to make a message specific. A reference to a real conference, a real vendor, or a real project makes the request harder to dismiss.

The more specific the message, the more important the verification procedure. Familiarity is a vulnerability when it replaces confirmation.

Voice Cloning and Deepfakes

Voice cloning tools can produce a convincing imitation of a known person from a short sample. Attackers use it to leave voicemails, make phone calls, or combine audio with a video call. The target hears a familiar voice and lowers their guard.

The defense is not better listening. It is a policy that certain requests require confirmation through a separate channel, even when the voice sounds right.

Teach Verification, Not Detection

The Two-Channel Rule

Any request involving money, credentials, or sensitive data should be confirmed through a second channel that the employee initiates. If the request arrives by email, call the person at a number on file. If it arrives by phone, send a message through the known internal channel.

The key detail is "on file." Calling the number in the email or the caller ID on the phone defeats the purpose. Attackers control those.

Payment Change Controls

Vendor bank account changes are a common target because they are routine and high value. Require a callback to a known contact, a second approver, and a short waiting period before the change takes effect.

The procedure should apply to every change, including ones that look legitimate. Attackers count on exceptions.

Credential and MFA Requests

No legitimate IT person needs a password or a one-time code. Teach employees to refuse the request, hang up or close the message, and report it. Then give them a simple way to reach the real service desk.

Use phishing-resistant MFA for high-risk accounts. Passkeys and hardware keys are much harder to trick than a code typed into a fake page.

Executive and Urgent Requests

Urgency is a manipulation tactic. A request that must be handled immediately, combined with a familiar name, is more likely to bypass judgment. Teach employees that urgency is a reason to slow down, not speed up.

Give assistants and finance staff a clear escalation path for unusual executive requests. The path should be easy to use and should not require challenging the executive directly.

Build Training Scenarios That Match the Threat

Use scenarios that reflect current attack methods: a vendor invoice with changed bank details, a voicemail from a known executive asking for a callback, a text message about a package, a chat message from a colleague asking for a code.

Run the scenario as a decision exercise, not a quiz. Ask employees what they would do next, where they would find the number, and who they would tell. Practice the procedure until it is familiar.

Include the awkward moments. What does an employee say when a caller claiming to be the CEO pressures them for a payment? Rehearsing the response makes it easier to use under pressure.

Make the Safe Path Fast

Verification fails when it is slow. If the employee has to search for a phone number, find an approval form, or wait for a manager who is in meetings, they will skip the step. Put the contact information where the work happens and make the approval process short.

A verified vendor list, a payment change form, and a direct line to the service desk are not glamorous controls. They are the ones that get used.

Review and Repeat

Threats change, and so do the messages employees receive. Review phishing simulations and real reports every quarter. Look for new themes, new sender patterns, and new pressure tactics.

Update the training examples and the procedures when something changes. A yearly course cannot keep up with a threat that evolves monthly.

What Leadership Should Do

Leadership sets the tone. When executives follow the verification procedure themselves, employees see that it applies to everyone. When a leader gets impatient about a callback, the procedure loses credibility.

Fund the technical controls that reduce reliance on human judgment: email filtering, phishing-resistant MFA, endpoint protection, and monitoring. Then train employees on the decisions that technology cannot make.

ATS helps businesses build phishing defenses that combine filtering, authentication, and practical training. If your team is still being taught to look for typos, contact our team to update the program for AI-era threats.

Posted in: Cybersecurity