Empathy - AI Simulation Platform
Training simulator for insurance agents that won 1st place at the PASHA Hackathon v5.0. It lets new hires practice high-pressure text and voice calls with emotional AI customer personas, giving them a safe space to learn from mistakes before handling real clients
Overview
Empathy is an AI-powered training platform designed for insurance support teams and call centers. Instead of building another generic chatbot for customers, this project focuses on the people behind the phone lines. It acts like a flight simulator for support agents, letting them practice high-pressure conversations with realistic AI customers before they ever handle a live call.
Role: Product Designer (User Flows, UI Design, Strategy)
Context: PASHA Hackathon 5.0 (We won it btw)
Target Users: Interns, New Hires, and Support Teams at PASHA Insurance
The Main Goal: Replace boring, text-heavy training scripts with a safe, interactive tool that builds real communication habits and emotional confidence.
Our Team
Product Designer - Fuad Isgandarov
Machine Learning Engineer - Anar Dashdamir
Frontend (+ some Backend) Engineer - Hacaga Hasanli
Frontend Engineer - Arif Hasanov
The Problem with Traditional Training
When companies talk about AI innovation in insurance, they usually build tools for customers—like basic chat assistants. However, forcing customer habits to change rarely works well, and users often prefer human involvement and genuine empathy when dealing with complex insurance problems.
True innovation starts with training the staff. Traditional support training looks like this:
A new agent (like Dilber, a 22-year-old recent graduate) joins the team.
She sits next to an experienced employee to watch, listen, and take notes.
If the senior agent gets too busy, the training stops.
The company ticks a box and says training is done, but nobody knows if she can handle a real crisis.
The Real-World Friction:
When a real customer calls after a stressful event like a car accident, they are panicked or frustrated. Reading text-dense manuals or memorizing static scripts does not prepare an agent for messy human emotions. The agent freezes, brand reputation suffers, and the customer feels ignored. Knowing what to say is not the same as knowing how to say it under pressure.
The Solution: Empathy AI
We designed Empathy as a safe space for agents to practice real, emotional conversations based on actual historical call logs. The platform allows users to make mistakes, fail safely, and learn without risking real customer relationships or corporate revenue.
How It Works
Training Setup & Persona Selection
When an intern or new hire logs into the platform, they select their training mode and pick a specific scenario.
Live Interaction & AI Evaluation
As the agent types or speaks their responses to the simulated problem, the underlying AI judges the interaction in real time against key corporate benchmarks:
Review, Scoring, and Custom Retraining
Once the simulation ends, the platform generates an instant scorecard (e.g., 7.8/10) with a clear visual breakdown of weak points.
Smart Suggestions: The system instantly highlights errors, explaining what could be improved (e.g., "You should have explained this clause using simpler terms").
Automated Scenarios: Based on the agent's mistakes, the platform updates their dashboard to recommend the exact scenarios they need to practice next.
Technical Strategy
Building a live-assist tool that hooks directly into real phone lines requires immense backend engineering, strict security clearings, and endless development cycles.
Focusing purely on a Training Simulator allowed us to build an incredibly strong MVP fast:
Frontend: A clean, component-driven Web App built using standard layouts for easy developer handoff.
Backend & AI: Powered by Node.js/Python FastAPI utilizing the GPT-4 engine alongside advanced Text-to-Speech and Speech-to-Text APIs for the voice features.
Business Value: It allows insurance firms to test and onboard agents safely, catching compliance mistakes early before they cost the company real money or trust.
