Multi-persona roleplays: how personas interact
A multi-persona roleplay agent simulates a buying committee: multiple AI personas in a single session, each with its own identity, role, personality, and priorities. This article explains how the personas interact during a session and what to expect as a team member.
<iframe src="https://www.youtube.com/embed/liLnav3iPV4" frameborder="0" allowfullscreen></iframe>
Watch: How to Train Reps for Multi-Stakeholder Discovery Calls with AI Roleplay
For setup instructions, see Create a multi-persona agent and Intro to multi-persona roleplay.
How turn-taking works
During the call, one persona speaks at a time. The active speaker changes based on the flow of conversation: how you address the group, which persona you direct a question to, and how the scenario is configured. You do not control the order; the AI decides which persona responds based on what was said and each persona's defined role in the conversation.
The call interface shows which persona is currently active. Each has its own name, title, and avatar so you can track who is speaking at any moment.
The primary persona
Every multi-persona agent has one designated primary persona. It typically leads the conversation, introduces the other participants, and drives the overall direction of the meeting. Other personas contribute from their own areas of concern: a technical evaluator raises implementation questions, a financial stakeholder challenges ROI, and so on.
How personas are configured
When a manager builds a multi-persona agent, each persona gets a name, title, and company; a distinct personality profile and communication style; its own set of objections, priorities, and concerns; and a gender and language setting. This means each persona can behave quite differently within the same session. A Dominance-type CFO and an analytical IT director push back on different things, in different ways.
Scoring
A multi-persona session produces one scorecard result, not a separate score per persona. The scorecard evaluates how well you handled the conversation overall: discovery, objection handling, stakeholder management, and closing. The same scorecard logic applies as in single-persona sessions.
Practical tips
- Address personas by name when you direct a question at one of them; this signals to the AI who should respond.
- Do not ignore quiet personas. In real buying committees, a silent stakeholder who does not feel heard can derail a deal.
- If the conversation stalls or one persona dominates, re-engage the others with a direct question.
- Use multi-persona sessions after you are comfortable with single-persona roleplays. Managing multiple stakeholders at once is a significant step up in difficulty.
Hope this article was helpful. For any queries, please write to support@outdoo.ai