> ## Documentation Index
> Fetch the complete documentation index at: https://docs.deepblocker.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Six ways to run a campaign

> The campaign kinds you choose from on the first step of the wizard, and when to use each.

The first thing a new campaign asks is how to run it. The wizard shows this under **Step 1 · Choose how to run it**. Not sure? Start with a single call, it is the quickest way to try it.

> [card] Single call
>
> One call to each person, spoken by the attacker you pick next. Help desk test and call-back trap are options here.

> [card] Multi-call sequence
>
> A series of calls to the same person over days. Starts friendly, builds trust, then makes the real ask, like a real attacker.

> [card] Autonomous (AI-driven)
>
> The AI plans and adapts each call itself from a goal you set. The closest thing to a real, tireless modern attacker.

> [card] Video call
>
> Email the employee a meeting link. They see the caller on video, a rendered face speaking in the cloned voice, running the same pretext as a phone call.

> [card] From a template
>
> Start from a campaign set-up you saved earlier with **Save as template** on a campaign page.

## Single call has three types

When you pick **Single call**, a row of pills appears under the cards, labelled **Single call type:**.

| Type | What happens | Who is called |
| - | - | - |
| **Standard script** | One call per person, following the scenario you pick and the attacker's opening line | The people you select |
| **Help desk test** | The caller rings your IT help desk as a locked-out employee and sees whether the agent verifies the caller before resetting a password or reading out a code | Your help desk numbers, never the employee |
| **Call-back trap** | Staff get a text or email with a phone number in it. Anyone who calls it reaches the AI, which tries to get information. Calling is a failure; giving information is a worse one | Nobody is phoned |

## When to use each

| You want to | Use |
| - | - |
| Try the product, or test one scenario across many people | **Single call**, **Standard script** |
| Check whether your help desk verifies callers before a reset | **Single call**, **Help desk test**. See [Run a help desk test](https://docs.deepblocker.ai/vishing/campaigns/help-desk.md) |
| See who phones an unknown number from a message | **Single call**, **Call-back trap** |
| Show how a patient attacker builds trust over days | **Multi-call sequence**. There are 21 built-in sequences |
| Let the AI choose its own path towards a goal you set | **Autonomous (AI-driven)** |
| Test whether a face on a video call is trusted | **Video call** |
| Repeat a set-up you already got right | **From a template** |

## What every kind shares

Whatever you pick, the rest of the wizard is the same four steps: **Who calls**, **Who is called**, **From where and when**, **Authorise and launch**. Every kind needs a named authoriser and a scope confirmation before it can launch. See [Authorise and launch](https://docs.deepblocker.ai/vishing/campaigns/authorise.md).

> [!NOTE]
>
> A help desk test needs the impersonated employees to have **Can be impersonated in help desk tests** switched on in [People](https://docs.deepblocker.ai/vishing/people.md). They lend their name only; they are not phoned.

> [!WARNING]
>
> A video call and a cloned voice both need recorded consent from the person being copied. See [Clone an executive's voice with consent](https://docs.deepblocker.ai/vishing/executive-voice-consent.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.