LitWare Inc. is building a Copilot Studio agent with a distinct brand persona — the agent should always respond in a professional, concise tone, refer to itself as 'Aria', and avoid mentioning competitors. Where in Copilot Studio should the developer configure these persona and behavioral instructions?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Setting an agent's persona is like writing a character brief for an actor — it goes in the script's front matter (the system prompt/instructions), not in individual scene directions. Copilot Studio's 'Instructions' field is exactly that system prompt. The correct answer is B.
Full explanation below image
Full Explanation
## Why B is Correct In Copilot Studio, every agent has an Instructions field (sometimes called the system prompt or custom instructions) accessible in the agent's Settings or on the Overview page. Text entered here is sent to the underlying language model as the system-level context for every conversation. This is where persona definition, tone instructions, behavioral constraints (like 'do not mention competitors'), and role definition belong.
## Why the Distractors Are Wrong A (Trigger phrases in Greeting topic): Trigger phrases are used for intent-matching, not for instructing the AI model. Adding instructions to the Greeting topic would display them as a user-visible message, not set the model's behavior.
C (Topic named 'Persona'): A topic fires only when its trigger is matched. You cannot force a topic to run at the start of every conversation reliably for model behavior configuration. The Greeting system topic does run at start, but it affects the conversational message flow, not the LLM's system-level instructions.
D (Power Platform admin center): AI model settings in the admin center govern tenant-level policies and data governance — not individual agent personas.
## Exam Tip Agent Instructions / System prompt = agent-level behavioral configuration. Think of it as the 'meta-prompt' that shapes all responses. It is distinct from topic-level prompts used in generative answer nodes.