Red teams have always been good at email. The phishing frameworks are mature, the reporting is familiar and the playbooks are well worn. The gap is everything that happens after the email: the phone call that follows it, the voicemail that asks for a callback, the Teams message from a colleague, the video meeting where a familiar face asks for an exception. That is where adversaries now close, and it is where most red team toolchains run out.

We run authorized vishing and phishing campaigns for enterprise security teams, so here is the short answer. The red team tools used for vishing and phishing campaigns fall into five categories: pretext and OSINT research, email and landing page frameworks, voice and telephony, collaboration and video channels, and an orchestration and evidence layer. Breacher.ai OSES™ runs all five as one platform, with a live AI voice agent that also answers the inbound callback.

What do red team tools for vishing and phishing campaigns need to do?

That definition sets the bar for the tooling. It has to reach every channel, carry context from one stage to the next, hold a conversation and produce evidence a CISO can act on. For the field view of what the voice leg looks like in practice, our AI vishing red team field findings cover what we see on the callback log.

The five categories of red team tooling

Every vishing and phishing campaign uses something from each category. The difference between programs is whether those pieces are stitched together by hand or run as one system.

Where a stitched toolchain breaks

A strong operator can build a campaign from separate tools, and many excellent red teams do. The friction shows up at the hand-offs. The email tool does not know the target replied. The telephony tool cannot answer when the target calls back an hour later. The Teams message runs from another account with another log. The evidence ends up in four places, and the one question the CISO asks, whether the help desk ran the verification step, takes a day to reconstruct.

What an orchestrated vishing and phishing campaign looks like

Orchestration is the difference between a series of tests and a campaign. Here is a typical sequence against a vendor banking change, one of the consequential actions most organizations protect least.

The callback is the stage that matters most. A target who calls a number back has already decided to trust it, and a tool that cannot answer that call never tests the moment trust is granted. OSES™ answers it with a live agent, which is why the orchestrated callback sits at the center of our campaigns.

The best red team tool is the one that follows the request all the way to the verification step.

Rules of engagement every campaign needs

Vishing and phishing campaigns that use live voice and likeness carry more responsibility than an email test. The tooling should make the guardrails easy, not optional.

• Written authorization and scope. The consequential actions in play, the people and numbers in scope, the time window and the abort conditions, agreed before anything is built.
• Executive consent for voice and likeness. Any cloned voice or face is used only with the written consent of the person it represents.
• Telephony rules confirmed. Calling and caller identification practices reviewed with counsel for every jurisdiction in scope.
• A hard stop at the verification boundary. No payment leaves, no credential changes and no access is granted. The campaign records the decision and ends.
• Findings named against procedures. The report describes which step held or slipped, never which person to blame.

What to measure from a red team campaign

A red team campaign is only as useful as its findings. These four measures, each defined by its denominator, turn a campaign into a decision about your controls.

How Breacher.ai runs vishing and phishing red team campaigns

The OSES™ simulation platform was built by security practitioners to replace the stitched toolchain with one orchestrated engine. In our sixteen-capability comparison, it is the only platform verified on all sixteen, including the AI callback agent and interactive video avatars.

• Every channel in one sequence. Email, SMS, live two-way AI vishing, the inbound callback, Microsoft Teams and interactive avatars in Teams, Zoom and Google Meet.
• Conditional branching. Each stage responds to what the target did at the last one, so the campaign behaves like a real adversary.
• Agent personas and editable prompts. OSINT-informed pretexts tuned to your organization, with cloned voice and video used only under written consent.
• Findings that drive training. Every slipped procedure becomes a module generated in minutes by the Secure Behavior Management platform, written from your own policies.
• Run it your way. Your team runs campaigns on the platform, or our practitioners run a full AI social engineering red team engagement end to end.

Why the campaign measures the procedure, not the eye

No bank measures whether a teller can eyeball a counterfeit note. They measure whether the teller ran the check. The bill got better. The check did not have to.

Our campaigns use the most convincing synthetic voices and faces we can build, because that is the pressure a real adversary brings. We never grade whether someone recognized the voice as synthetic. Human perception does not improve as generation quality does, which is why deepfake detection training is a decaying control. A callback to a known-good number returns the same answer against a crude fake and a flawless one. A workforce that invokes process defeats our red teams, and that is the outcome every campaign is designed to prove.

This is why the red team tooling has to be convincing. A better fake does not break a verification step directly. It raises the pressure to grant an exception, which is why exception rate is in every report. Procedures are executed by people, so people still matter, and a procedure only protects the paths where it exists, which is why coverage comes first. We also welcome machine identity verification, such as liveness and document checks in the hiring path, and test whether those controls hold too.

Frequently asked questions