AI has made phishing harder to spot, which means MSPs have a bigger responsibility to help clients understand what they are up against.
Most clients are not equipped to keep up with these threats on their own. They may only know the old signs of phishing. AI has made it easier for attackers to craft cleaner messages, use publicly available information, imitate trusted voices, and create content that feels natural enough to lower someone’s guard.
MSPs are in the best position to help clients understand this before the lesson comes from an incident.
That was the focus of a live session with Empath Co-founder Wes Spencer, Head of MSP Success Dean Trempelas, and Huntress Community Growth Strategist Tom Lawrence.
Keep reading for a practical guide to starting that conversation with clients, including the AI phishing methods worth explaining first and the simple questions that can help their teams slow down before they trust the wrong thing.
AI phishing is defined by a notable shift in how attackers make scams feel more personal and believable.
Here are the methods clients need to understand first.
AI makes it easier for attackers to write messages that sound closer to normal business communication. Tom explained that attackers who may not understand a client’s industry or speak English as a first language can now use AI to phrase a request better, craft a lure, and make the message sound credible.
The old advice was often, “Look for spelling mistakes.” The better advice now is, “Look at what the message is asking you to do.”
Is it asking for credentials? A payment change? A file download? A login? A password reset? An urgent approval? A new tool install?
A phishing email can be polished and still be dangerous.
Attackers do not need private access to make a message feel personal. A lot of useful context is already public.
LinkedIn profiles, company websites, event pages, podcasts, social media posts, community groups, public records, and old online activity can all help an attacker understand who someone is, what they care about, who they work with, and what kind of message might get their attention.
Personalization does not prove trust. It may only prove that the attacker did enough research.
Clients are used to treating voices, faces, and videos as proof. If they hear someone’s voice on a call or see someone’s face in a video, it feels natural to believe the request is legitimate.
That habit is becoming risky.
Tom pointed out that phishing can occur on LinkedIn, Discord, Facebook, Instagram, via text, or on any other platform where people communicate.
People often behave differently outside of email. They may be more relaxed in a community chat, more responsive to a text, more trusting in a social DM, or less suspicious when something comes through a platform they use casually.
Dean described searching for help with a UniFi issue and finding a result linking to a Tom Lawrence video with commands that appeared to solve the problem. He trusted it because Tom is a known authority in that space.
Then Dean asked the scarier question: what if someone deepfaked a trusted expert, optimized the fake tutorial for search, and gave commands that solved the immediate issue while also opening a door into the environment?
Clients need to understand that search results, YouTube videos, Reddit posts, and AI-generated summaries can all become part of the attack path.
Clients are already using AI, and many will continue to use it because it helps them move faster. The better goal is to help clients become more careful users.
This is where MSPs can take a more proactive role in the client relationship. Protecting clients from cybersecurity issues entails helping the people inside the client’s business understand how modern attacks reach them.
They need to understand enough to pause before acting on a request that creates risk. Phishing awareness training and broader cybersecurity awareness training have become more valuable.
MSPs cannot take all the technical “nerdery” from a conversation like this and hand it directly to clients. They need to translate it into something clients can use. Wes emphasized that the goal is not perfection. It is getting “little bits better.”
MSPs can teach better questions.
Was the person already waiting for this message, invoice, file, approval, download, or login prompt? Unexpected requests deserve extra attention, even when they look professional.
Would this person usually ask through this method?
If a vendor normally uses a portal, why are they texting? If a manager usually sends approvals through email, why is this request coming through a social platform? If a colleague usually shares files through the company’s approved system, why is this link somewhere else?
Clients should be trained to slow down when a request involves credentials, payments, access, software installs, confidential information, browser extensions, or command-line instructions.
Does the message mention a real vendor? A recent event? A personal interest? A company project? A colleague? A familiar tool? A public detail from LinkedIn?
If the message feels believable because it knows something, that does not automatically make it safe.
For high-risk actions, clients need a second path.
That could mean calling a known number, checking the official portal, asking through an existing thread, confirming with the MSP, or using an internal approval process.
The key is that verification should not happen only inside the suspicious interaction.
AI phishing is not going to get easier for clients to understand on their own. The methods will keep changing, and the content will keep getting more believable, which is why MSPs should bring this topic into regular client education before there is a problem.
The goal is to give them language they can repeat inside their own teams and practical habits they can use when something feels convincing.
MSPs have the opportunity to show value beyond the tools they manage in the background. Clients may not fully understand every detection, policy, or configuration change, but they can feel the value of an MSP that helps their team understand what is changing and what to do about it.
This blog is a high-level, practical version of the conversation. The full Empath and Huntress live session goes deeper into the examples, including public recon, personalized lures, deepfake risk, fake tutorials, and how attackers exploit trust.
At Empath, we are also working on ways to help MSPs offer AI training as a proper client-facing service, so teams can educate clients without having to build every resource from scratch. If you want to stay in tune with updates about it, subscribe to the Empath newsletter.