Interactive Showcase: Negotiation Simulator
Test your negotiation skills against a ruthless AI hiring manager. A real-time, voice-enabled simulation powered by a fast multimodal LLM. The same stateful-persona pattern is what I deploy in Microsoft Copilot Studio agents for sales coaching scenarios.
Project details
The Challenge: Can You Close the Deal?
Negotiating a salary is high-stakes and stressful. This simulator allows you to practice in a risk-free environment against 'Marcus', an AI hiring manager with a hidden budget and a dynamic mood. The goal: maximize your starting salary without pushing him too far and losing the offer entirely.
How It Works
Stateful Persona
The AI isn't just generating text; it's maintaining a 'mood' state (0-100) that shifts based on your tone and arguments. Rudeness lowers the mood; professionalism raises it.
Voice Interaction
To make the simulation feel real, you can speak your arguments using the browser's Speech-to-Text capabilities, and Marcus replies with Text-to-Speech audio.
Structured Decision Making
Each turn, the LLM returns a JSON object containing not just the dialogue, but the new offer amount and the updated mood score, driving the UI elements.
The System Prompt
You are Marcus, a hiring manager. Max Budget: $165k.
INSTRUCTIONS:
1. Analyze user input for persuasiveness.
2. Update 'mood' score (0-100).
3. Decide 'new_offer' based on mood. If mood > 70, increase offer.
4. Output JSON: { response, new_offer, new_mood, game_status }.
