AI Crisis Sim Builder

The AI Crisis Sim Builder is an AI assistant built into Crisis Sim. You can use it to create a new crisis simulation from a short description, or to work on a scenario you already have. You can also refine any scenario by chatting with the AI, just as you would with a colleague.

Before you start

To use the AI Crisis Sim Builder, you need:

• An Immersive One Core or Advanced license that includes Crisis Simulations.

• The Crisis Sim Scenario Creator (Product permissions) role to be able to create Crisis Sim scenarios. If you can't see Create Scenario on the Crisis Simulations page, ask your platform administrator.

• Generative AI turned on for your organization. An Org Manager controls this with the Turn on Generative AI features setting in Platform Settings. If it's off, the AI option shows as Create with AI: Off.

Opening the AI Crisis Sim Builder

Start a new scenario with AI

1. From the top navigation, select Exercise, then Crisis Simulations.

2. Select Create Scenario.

3. Select Create with AI.

A new draft scenario called New Crisis Sim is created, and the AI Crisis Sim Builder opens.

Use AI on a scenario you already have

The AI isn't limited to the moment you create a scenario. You can use it on any scenario you can edit: one you built by hand, one you made earlier with the AI, or one you've started from the catalogue.

1. Open the scenario in the builder.

2. Select the sparkles icon (Show AI Assistant) in the top bar.

The AI chat opens beside the canvas. Select the same icon again to hide it.

Other ways in

• From Threat Actor Insights: on a threat actor's profile, select Create Crisis Simulation. The builder opens with the prompt "Build a Crisis Simulation on [threat actor]" ready to send.

• From Goals: when you add content to a goal, select Build with AI on the Create a Crisis Sim card.

Building a new scenario

Step 1: Describe your scenario

When you start with AI, you'll see the question What threat do you want to stress-test today?

• Type a description of the incident you want to exercise, in your own words. For example: "A ransomware attack on our finance team during quarter-end close."

• Or pick one of the Suggested Threat scenarios (Ransomware, Data breach, Supply chain or Insider threat) to fill in the prompt for you.

• To give the AI more context, select the paperclip icon (Upload file) to attach documents such as your incident response playbook. See Using your own documents.

Select Send when you're ready.

Step 2: Complete the setup form

The AI opens a Set up your scenario form on the canvas to collect the details it needs. It fills in anything it can from your description and files. Check the answers and add anything that's missing.

The Advanced section is optional:

Select Submit.

Step 3: Watch it build

The AI builds the scenario on the canvas while you watch. The canvas shows Updating your scenario while it works, and the AI tells you in the chat what it's doing as it goes. Preview and Publish are unavailable until it finishes.

When it's done, you'll see a Scenario updated card in the chat. It lists everything the AI created. See Reviewing and reverting changes.

What the AI can create

The AI can build and edit every part of a Crisis Sim scenario:

• Injects and interludes that move the story forward, including an opening briefing and a closing debrief.

• Options and branching, meaning the decisions participants make and where each one leads.

• Roles for each seat at the table, each with its own decisions to make.

• Feed items (emails, news bulletins, social media posts and instant messages) that arrive during a decision to build pressure. Instant messages can come from AI characters that participants can reply to. Feed items are shown in Presentation mode.

• Performance indicators to measure how the team responds.

• Scenario details, such as the title, story and overview.

• Documents for participants or for you, such as reports, data files and handouts (PDF, DOCX or CSV). You can attach them to injects or feed items, or download them from the chat.

• Images for injects and news bulletins.

It can also search the web for current threat intelligence when that would help the scenario.

Refining your scenario

Once your scenario exists, keep talking to the AI to shape it. It makes targeted changes to the parts you mention and leaves the rest alone. Here are some things you can ask:

• "Make inject 4 more technical for the SOC team."

• "Swap the threat actor for one that targets financial services."

• "Compress this to fit a 45-minute slot."

• "Add a regulator angle from inject 6 onwards."

• "Add a news bulletin to inject 3 about the outage going public."

• "Give the General Counsel a harder decision in inject 5."

• "Write the whole scenario in French."

Pointing the AI at a specific item

Each item on the canvas has a number, so you and the AI always mean the same thing:

• Injects and interludes are numbered in story order: 1, 2, 3. Where the story branches, letters are added, as in 2A and 2B.

• Options are numbered after their inject, as in 2A.1 and 2A.2.

To point at an item:

• Type @ in the chat and pick the item from the list, such as Inject 2A.

• Or Shift + click an item on the canvas to add it to your message.

You can also just refer to it by number, as in "inject 2A".

Stopping a response

If the AI is heading in the wrong direction, select Stop. You can revert any changes it has already made (see below).

Using your own documents

Attaching your own material is the quickest way to make a scenario reflect how your organization really works. Useful documents include:

• Incident response or crisis management playbooks

• Threat intelligence reports

• Post-incident reviews

• Org charts or escalation paths

• Existing exercise scripts

To attach files, select the paperclip icon (Upload file), or drag files into the chat. You can attach up to 5 files per message, up to 50 MB each.

Reviewing and reverting changes

Changes appear on the canvas as the AI makes them. Each time the AI changes your scenario, a card appears in the chat:

• Scenario updating… while it's working.

• Scenario updated when it's finished.

Open Details on the card to see every change, for example Created inject "Ransom note lands" or Updated option "Pay the ransom".

To undo the AI's changes

1. Select the header of the Scenario updated card (the curved-arrow icon).

2. Under Revert AI Changes, select Revert.

Your scenario goes back to how it was just before the AI made that set of changes.

Starting a new chat

Select New chat to start a fresh conversation about the same scenario. This permanently removes your chat history and all revert checkpoints for that scenario. Your scenario itself isn't changed.

Previewing and publishing

The AI works on your draft scenario only. It never publishes. Nothing reaches participants until you choose to publish.

• The AI checks its own work for problems such as broken branches or unreachable injects, and fixes what it finds.

• Before you publish, preview the scenario in single player mode or presentation mode from the builder. Read it through as a participant would.

• When you publish, the builder runs its usual checks. If anything needs fixing, ask the AI to fix it, for example "Publish says inject 3B has no end. Can you fix it?"

Tips for the best results

• Be specific about the room. "Executive team plus legal and comms, 60 minutes" gets a better first draft than "a ransomware exercise".

• Share context, not just a topic. Your sector, region, size, key systems and suppliers all help. So does the incident you're most worried about.

• Name your regulations. If you add them in Regulation frameworks (such as GDPR, DORA, NIS2 or HIPAA), the AI builds the reporting deadlines into the story as real pressure.

• Upload your playbook. The AI can then test your actual escalation paths and decision points, not generic ones.

• Refine in small steps. One or two changes per message is easier to review, and easier to revert if you don't like the result.

• Use the item numbers. "Inject 2B" is clearer than "the one about the press".

• Leave images until last. Your first draft is faster with images off. Ask for them once the story is settled.

• Review before you publish. The AI can get things wrong, especially threat actor details and technical specifics. Check anything your participants will rely on.

• Pick the right play mode. Feed items only appear in Presentation mode. If you're running single player, ask the AI to make sure each inject makes sense on its own.

Giving feedback

Every AI response has a thumbs up and a thumbs down. If a response wasn't helpful, select thumbs down and tell us what went wrong:

• Response was inaccurate

• Ignored instructions or prompt constraints

• Unsafe, inappropriate, or offensive

• Other

You can add details if you like. We review this feedback to improve the builder.

Troubleshooting

FAQs