Question

A financial services firm is planning an agent in Copilot Studio that helps relationship managers respond to client requests. Regulatory policy requires that any advice touching investment products must be reviewed by a licensed human advisor before being sent, that the agent must never quote specific product performance figures, and that all interactions be logged for audit. As the solution architect, how should you incorporate these requirements into the agent design during the planning phase?

Accepted Answer

Define explicit rules and constraints—including topic restrictions, human-in-the-loop approval for regulated advice, and mandatory interaction logging—as governing guardrails that shape the agent's scope, instructions, and orchestration before development begins

Answer

Rely on prompt engineering alone to instruct the model not to give investment advice, since a well-crafted system prompt is sufficient to satisfy regulatory obligations

Answer

Build the agent without constraints first to maximize capability, then add compliance filtering only if auditors raise findings after deployment

Answer

Replace the agent with a fully custom-trained language model, because only a bespoke model can enforce regulatory rules reliably