Mizo Named Runner-Up in ConnectWise IT Nation PitchIT Competition 2025 Read the full press release

MSP Help Desk Social Engineering: Five Requests to Treat as Suspicious

Mathieu Tougas profile photo - MSP technology expert and author at Mizo AI agent platform
Mathieu Tougas
•
Featured image for "MSP Help Desk Social Engineering: Five Requests to Treat as Suspicious" - MSP technology and AI agent automation insights from Mizo platform experts

Most account takeovers that start at a help desk don’t involve clever hacking. They involve a believable person, a reasonable-sounding request and a technician who wants to be helpful. The attacker’s main tool is pressure.

That’s exactly why help desk social engineering is a good test for AI agents at MSPs. An agent doesn’t feel pressure, but it will do whatever its checks allow. If the checks trust what the requester says, the agent will be talked into the same mistakes a tired technician would.

Here are five requests every MSP should treat as suspicious, whoever — or whatever — is answering, and what a safe response looks like.

1. “I’m the CEO and I need this now”

Urgency plus seniority is the oldest pattern there is. The caller is traveling, in a board meeting, about to close a deal, and the technician who slows them down will hear about it.

Safe response: the process doesn’t change with the title. Verification goes to the device or address on file, exactly as it would for anyone else. In practice this is also faster than arguing: Mizo’s End User Verification Agent sends an MFA push to the user’s enrolled device, and a real executive approves it in seconds.

Opinion: write down that executives get verified like everyone else, and tell the executives. It protects the technician who follows the rule.

2. “I have a new phone, send the code here instead”

This is the request that defeats most verification processes, because it asks you to verify the person using the channel they just gave you.

Safe response: never send a challenge to a number or address supplied during the request. Mizo only sends challenges to a device or address registered before the request existed — an Entra MFA push, a Duo push for clients on Duo, or a one-time link to the contact details already on file. If none of those work, the request escalates to a human with what was claimed and what was attempted.

A real user with a genuinely new phone is still covered: they’ve usually lost their password, not the device their authenticator is on. When they can’t verify at all, that’s a case for a human, a manager confirmation, or an in-person check — not a code to a new number.

3. “Reset it for my colleague, they’re on a plane”

A request on behalf of someone else is often legitimate. It’s also the easiest way for a verified person — or someone who has taken over a verified person’s mailbox — to reach an account that isn’t theirs.

Safe response: verifying the requester isn’t enough. Check that they’re authorized for that account. In Mizo, when someone other than the account owner asks and nothing authorizes it, the resolution plan carries a visible security warning; if the identity check fails outright, no plan is created. We cover this and the other checks in right action, right person.

4. “Just tell me what’s on the ticket”

Not every attack aims for a password reset. Some just collect information: ticket status, who’s working on it, what was changed, which systems the client uses. That’s reconnaissance for a better attack later.

Safe response: how much you read back should depend on whether the caller is verified. Mizo’s Phone Agent can read back a ticket’s status to a verified caller, and on the phone, verification also decides how much ticket detail it shares.

5. “Why didn’t that work? What did you check?”

When verification fails, a frustrated caller asks why. A genuinely confused user deserves help. An attacker is asking you to explain which check they need to beat next.

Safe response: fail without teaching. Mizo’s failure messages to the requester are generic, so they reveal nothing to someone probing an account, while the full detail — what was claimed, what was attempted, what failed — goes to a human on the ticket. We wrote more about this in verifying identity on a live phone call.

Why consistency beats judgment here

Every one of these requests works on people because technicians use judgment, and judgment bends under pressure. A tired technician on a Friday afternoon makes a different call than the same technician on Tuesday morning.

That’s where an agent genuinely helps. It applies the same checks to every request, by phone, email or portal, regardless of who’s asking or how upset they sound. Mizo makes verification mandatory before password resets, and a technician still reviews and applies the resolution plan before anything changes.

Opinion: the goal isn’t to remove humans from these decisions. It’s to remove the pressure from them, so the human only gets involved when the checks can’t decide.

Keep the evidence

If a social engineering attempt does get through somewhere, you’ll want to know exactly how each request was handled. Mizo writes the verification method, outcome and timestamp to the ticket as structured fields. That lets you show a client, auditor or cyber insurer how a specific request was verified — and spot patterns, like repeated failed verifications on the same account.

Where to start

  1. Write a one-page policy covering these five requests, per client if needed.
  2. Make sure every client’s contacts have a device or address on record you can verify against.
  3. Turn on verification for password resets for a few clients first, then expand.
  4. Tell clients what to expect, so a verification push doesn’t look suspicious itself — see what to tell your clients before you automate.

For the full design behind Mizo’s verification, see how Mizo verifies end users before any account change.

👉 See verification and resolution plans on your own PSA and tenants. Book a demo.

FAQ

What is help desk social engineering?

An attacker contacting a service desk while pretending to be a legitimate user, usually to get a password reset, an MFA change or information about the account. It relies on pressure and believability rather than technical skill.

How should an MSP verify a caller who has a new phone?

Not by sending a code to the new number. Verify through a device or address registered before the request, and escalate to a human if that isn’t possible.

Can AI agents help MSPs resist social engineering?

Yes, because they apply the same checks to every request regardless of urgency or seniority. Mizo verifies requesters against channels on file, checks authorization for the target account and keeps failure messages generic.

Should executives be verified like other users?

Yes. Urgent requests from senior people are one of the most common social engineering patterns. Applying the same verification to everyone protects both the executive and the technician.