The AI Agent Platform includes a built-in Simulator that allows testing the orchestration before publishing a new version. The Simulator replicates real conversation flows - through both chat and voice channels - so it can validate agent behavior, routing logic, and responses directly from the canvas.
- Accessing the Simulator
- Choosing a Channel
- Defining Simulation Variables
- Running the Simulation
- Test Scripts
- Related Articles
Accessing the Simulator
To open the Simulator, open your orchestration in AI Agent Orchestration and click the Test button in the top-right corner of the canvas.
A panel will appear on the right side of the screen, prompting you to choose how you want to interact with the Simulator.
Choosing a Channel
The Simulator supports two testing channels: Chat and Voice.
Chat
Selecting Chat starts the simulation immediately. A conversation window opens where you can type messages and interact with the AI Agent in real time, exactly as a customer would.
Voice
Selecting Voice takes you to the voice configuration screen before starting. Click Configure to adjust the voice settings for a realistic testing experience. Available settings include:
- Call Audio Quality: Choose between phone line quality or high fidelity audio to simulate different call conditions.
- Language: The language used during the voice interaction. You can add multiple languages to support multilingual scenarios. When multiple languages are configured, you can also define the language detection strategy:
- Most frequent: Uses the language detected most often throughout the interaction.
- Last used: Switches to the most recently detected language.
- First used: Locks to the language detected at the start of the interaction.
- Voice: The specific voice persona to use.
- End of Speech Sensitivity: Controls how quickly the system detects that the speaker has finished talking.
- Speaking Speed: Adjusts the pace of the AI Agent's speech during the simulation.
- Phrase List: List of words or phrases to improve recognition accuracy for domain-specific or uncommon terms.
Once configured, click Save to apply your settings, then Start to begin the voice simulation.
Defining Simulation Variables
Both chat and voice modes support Application Input variables - context data that would normally be injected into the orchestration by an external application (such as Autopilot, Copilot, or the API).
To set these variables before starting a simulation:
- Click the { } button (available from either channel selection screen).
- Under System Variables, add the value you want to pass into the simulation (up to 5,000 characters) in the Application Input field.
- Click Save.
This is particularly useful when your orchestration relies on variables collected earlier in a Studio flow, allowing you to replicate those conditions during testing.
Running the Simulation
Once the simulation starts, a Test Orchestration panel opens. You will see:
- The Interaction ID and Session ID for the current session (useful for cross-referencing logs).
- The AI Agent's opening message.
- A message input field at the bottom, where you can continue the conversation.
Type a message and send it to interact with the orchestration. The agent will respond based on its configured logic, routing conditions, and skills.
Insights and Thought Process
Each agent response includes an Insight indicator. Clicking on it reveals two layers of transparency:
- Agent work stream — shows the tool calls made during the response (e.g., skill executions, knowledge base lookups).
- View Thoughts — exposes the internal reasoning process, including communication between the Coordinator and individual agents, showing how routing decisions were made.
This makes it easy to diagnose unexpected behavior, verify that the correct agent was selected, and understand why a particular response was generated.
Test Scripts
If you want to simulate structured or repeatable scenarios rather than typing messages manually, you can use Test Scripts. Access this option from the Test panel menu.
Test Scripts support two modes: Scripted and AI-Generated. Please see the sections below for more details on each.
Scripted
Define each customer message manually, in sequence. This gives you full control over the exact inputs used in the simulation. You can add multiple messages to build out a multi-turn conversation, or upload a pre-defined script via Upload JSON.
AI-Generated
Describe the scenario and the goal of the interaction in natural language. The AI will automatically generate a realistic conversation script based on your description, which can then be used to run the simulation.
Test Scripts are especially useful for regression testing, onboarding new team members, or documenting expected conversation flows.