Picture a 25-person accounting client on a quiet Tuesday. Someone in finance gets a text from the “IT help desk” about an expiring password, then a phone call to confirm it. No email anywhere in the chain. The quarterly simulation you built around a fake invoice would have missed every step of it.
That gap is the real story of AI and security awareness. Most vendor pitches for MSP security awareness training skip it and go straight to scary AI headlines. The data points somewhere more useful.
The 62% Nobody Can Shake
Verizon’s 2026 Data Breach Investigations Report studied more than 22,000 confirmed breaches across 145 countries. People played a part in 62% of them, up from 60% a year earlier.
You’d expect the next paragraph to blame AI. The findings are stranger than that. Verizon worked with Anthropic’s Safeguards team on 793 threat actors flagged for misuse. The median one used or researched AI help across 15 documented attack techniques, and a few pushed it to 40 or 50. Yet the AI-assisted malware mostly rebuilt old tricks. The median technique already had 55 known malware examples doing the same job, and under 2.5% of observations touched techniques with one or fewer known examples. Verizon’s own summary line is “refinement, not revolution.”
So the habits you already teach still hold up. Check the sender, distrust urgency, report anything odd. AI just writes the lure faster, in cleaner grammar, at higher volume, and agents now run whole attack campaigns for pocket change.
The bigger shift is where people get hit.
Phones Beat Inboxes
In the phishing simulations Verizon reviewed, median click rates on voice and text ran 40% higher than on email. The sample was small, just 35 campaigns, so treat the exact number loosely. The direction matches what help desks already hear, though. Staff have years of practice squinting at suspicious emails. A mid-afternoon call from someone who sounds like the vendor’s accounts team catches them flat-footed.
Pretexting, the slow build of a fake trusted relationship, now accounts for 6% of breaches as a way in, usually by phone. Consumer fraud shows the same drift. UK National Trading Standards lists cloned voices and error-free phishing at scale on its first AI scam watchlist.
Here’s how those findings map onto a training program:
| What Verizon 2026 found | What the program should change |
|---|---|
| Human element in 62% of breaches | Keep training on the contract as a standing service |
| Mobile click rates 40% above email | Rotate voice and text scenarios into simulations |
| AI-assisted attacks mostly reuse known techniques | Refresh the lures, keep the core habits |
| 45% of employees use AI regularly on work devices | Add a short module on what stays out of AI tools |
Nothing in that table says to scrap the curriculum. Verizon’s data argues the opposite. The genuinely new topic is your clients’ staff using AI themselves. Regular AI use on corporate devices jumped from 15% to 45% in a year, and 67% of users reach those tools through personal accounts. Shadow AI is now the third most common non-malicious insider action in Verizon’s data loss prevention records, and source code is what people paste most. A five-minute module on what never goes into a chatbot costs little and closes a real hole.
The Small-Business Wrinkle
Verizon’s numbers for small and mid-sized businesses look different from the headline. People were involved in 45% of SMB breaches, well under the 62% overall. Third parties showed up in 55%. The top way in was vulnerability exploitation at 26%, ahead of credential abuse at 13% and phishing at 9%.
Training still matters for these clients. But “your people are your biggest risk” oversells it, and an owner who reads the report will notice. Pitch training as one layer next to patching and MFA. And remember that an MSP counts as a third party too, which deserves an honest conversation of its own. Clients forgive a modest claim far faster than a stretched one.
Building the Service
What goes in the package
A yearly session gets forgotten by spring. Monthly delivery works better, and it matches how you already bill for backup and patching. The core has three parts: short modules, simulated phishing, and an easy way for staff to flag suspicious messages.
Two or three tiers are plenty. An entry tier could pair annual modules with quarterly simulations. Higher tiers add role-based content, policy sign-off tracking, executive briefings, and the voice and text tests above. Sell it alongside email security, endpoint protection, and backup so clients see one security stack rather than a loose add-on.
Put the boring parts in writing: the minimum term, the headcount that changes the price, and how leavers come off the roster. Spell out that incident response sits outside the training fee, too. That sentence saves an argument later.
Onboarding
Nothing looks sloppier than training sent to someone who quit in March. Before launch, gather the staff list, departments, email platform details, and the names of executives and finance people.
Allowlist your simulation domains and test delivery with the client’s IT contact. Directory sync handles new hires and leavers automatically. If phone or text tests are in scope, agree with leadership on whose devices they’ll reach.
Have the boss announce the program, since staff accept it better from their own leadership. Decide who approves templates, who fields “was that a test?” questions, and who contacts staff after a real attack. Then run one baseline simulation before any training, so you have a starting line to measure against.
Picking the platform
Technician time is the cost that quietly kills margin. Favor tools built for providers, with one dashboard across many client accounts and modules short enough to finish between meetings.
Trial it with one friendly client first. Do the simulations look convincing? Does the report button play nicely with their mail system? Can a non-technical manager follow the report? Ask about language options, accessibility for assistive tools, and how fast vendor support actually replies.
Keeping Campaigns Unpredictable
Staff learn patterns fast. If the test always lands on the first Tuesday of the quarter, they’ll learn your calendar instead of the threat.
Send a short simulation every month or two, plus a brief module each quarter, and shift the template, difficulty, and day. Follow the seasons: tax lures in spring, parcel scams before the holidays, benefits emails in autumn. Mix in calls and texts. Schedule a round after a staffing change or software rollout, because both give attackers fresh openings.
Anyone who clicks should see quick, specific feedback right away. Thank the people who report. Repeat clickers get private coaching, since public correction only teaches people to hide mistakes. Recycle real incidents as well. If three clients see fake document-sharing requests this month, that’s next month’s scenario. A simple per-client log of templates, dates, and results keeps you from repeating yourself and hands account managers material for the next review.
Reporting and Pricing
The quarterly review is where the service earns its renewal. Keep the report readable in two minutes by someone outside security:
- Click and report rates against the baseline
- Completion by department
- Real suspicious messages staff caught
- Repeat-risk groups and the planned fix
- Next quarter’s scenarios, including phone and text
Lead with one plain line, such as “reporting rose steadily and staff caught two real phishing messages early.” Show the weak months too. An explained dip earns more trust than a glossy chart that later turns out to be spin.
Per-user, per-month pricing grows with the client and keeps invoices predictable. Cover licensing, onboarding hours, campaign management, and report prep, and set a monthly floor so tiny accounts don’t drain your margin. Some providers fold the entry tier into their core security bundle and sell the rest on top. Revisit prices every year, and bill custom templates or extra executive briefings rather than quietly absorbing them.
If this is new for your team, pilot it with two or three clients who’ll tell you the truth. Tune the templates and reports, find out what your techs can sustain, and only then roll it out across the base. Record each client’s cadence and results along the way, so the next technician picks up where the last one left off.
Related: AI Hackers Are Getting Faster. Is Your Penetration Testing Keeping Up?
