Docs

Simulation Prompts

The Simulation Prompts tab is where Super Administrators manage the prompt configurations used by the platform’s simulation engine. Simulation prompts define the prompt templates that power agent simulations — they are the instructions the simulation LLM follows when acting as a synthetic user during a chat or voice simulation session.

Note

The Simulation Prompts tab is only visible to Super Administrators, and only when simulation is enabled for your Neo deployment. If you are a Super Administrator and can’t see the tab, ask your platform team whether simulation has been switched on.

Prompt Configurations

Each prompt configuration has three fields:

  • Type — the category of simulation this prompt applies to. Types are selected from a predefined list, or a custom type can be entered by checking Custom type. The type is used by the simulation engine to select the appropriate prompt for a given simulation context.
  • Name — a descriptive label for the configuration, used to identify it in the list.
  • Prompt — the full prompt template text. This is the system instruction given to the simulation LLM, defining how it should behave when acting as a synthetic caller or user.

Managing Prompt Configurations

Creating a prompt configuration

Click + Create (the plus icon in the top-right corner) to open the prompt configuration form. Fill in the Type, Name, and Prompt fields, then click Create.

Editing a prompt configuration

Click the Actions menu (three-dot icon) on any row and select Edit to modify an existing configuration.

Deleting a prompt configuration

Click the Actions menu on any row and select Delete. A confirmation dialog will appear before the configuration is removed.

Best Practices

One prompt per simulation type : Keep prompt configurations focused and specific to a single simulation type. Broad or multipurpose prompts produce less predictable simulation behaviour.

Test before setting as default : Before relying on a new prompt configuration for agent testing, run it against a known agent to verify the simulation behaves as expected.

Version naming : Use descriptive names that indicate the prompt’s purpose and iteration (e.g., voice-inbound-v2). This makes it easier to track changes when multiple configurations exist for the same type.