AgentSync AI — Contact Center Copilot 🥈

🥈 2nd Place Winner — Akbank Case Competition
Solution developed for Akbank Contact Center Operations & Customer Experience

AgentSync AI is a real-time conversational AI copilot designed for banking customer service representatives. It analyzes live customer audio and speech in real time to monitor caller sentiment, dynamically guide call resolution, prevent premature cross-selling during stressful inquiries, and trigger high-probability Next-Best-Offer (NBO) recommendations when customer sentiment stabilizes.

---

🎯 The Problem

In direct banking contact centers:

• Representatives face immense multitasking pressure: active listening, navigating core banking menus, adhering to strict compliance protocols (KVKK, 2FA), and attempting cross-selling.
• When an anxious or stressed customer calls (e.g., lost debit card, suspicious account activity), robotic cross-selling pitches damage customer trust and escalate churn.
• Conversely, when positive sentiment and financial inquiries arise, representatives often miss the exact window to propose relevant financial instruments.

---

💡 The Solution: AgentSync AI Architecture

AgentSync AI operates as a non-intrusive, live HUD (Heads-Up Display) for representatives:

[ Caller Audio Stream ]
         │
         ▼
[ Real-Time STT & Acoustic Analysis ]
         │
    ┌────┴───────────────────────────┐
    ▼                                ▼
[ Sentiment & Stress Curve ]    [ Intent & Topic Classifier ]
    │                                │
    └──────────────┬─────────────────┘
                   ▼
       [ Prescriptive Copilot Logic ]
                   │
    ┌──────────────┴─────────────────┐
    ▼                                ▼
[ Immediate Action Guardrail ]  [ Adaptive NBO Engine ]
(e.g., Freeze Card, Hold Pitch)  (Recommend Loan / Fund @ 78%)
                   │
                   ▼
     [ Live Representative Cockpit ]

---

✨ Key Capabilities

1. Live Speech & Sentiment Analytics:
   Calculates moving sentiment scores across 3-minute intervals. Visualizes sentiment curves (falling vs. rising trends) to alert representatives to customer emotional distress.
2. Emotion-Tagged Real-Time Transcription:
   Labels live dialogue turns with psychological markers (e.g., Neutral, Anxiety, Stress, Opportunity).
3. Context-Aware Action Guardrails:
   Automatically locks and defers marketing offers when distress is detected, instructing the representative to prioritize security and reassurance.
4. Adaptive Next-Best-Offer (NBO) Engine:
   Evaluates customer demographic data, credit score (Findeks), and recent spending patterns to predict relevant products with quantified acceptance probability (e.g., 78% acceptance probability for an education loan once anxiety resolves).
5. Real-Time Regulatory & Fraud Checks:
   Monitors KVKK consent, two-factor authentication bypass status, and fraud risk score simultaneously on a single unified screen.

---

🚀 Interactive Live Prototype Demo

An interactive, self-contained demonstration prototype is included in this repository.

Quick Start

# Clone the repository
git clone https://github.com/bederd/agentsync-ai-copilot.git
cd agentsync-ai-copilot

# Open in browser
start index.html       # Windows
open index.html        # macOS

Or serve via any static HTTP server:

python -m http.server 3000
# Visit http://localhost:3000

Scenario Walkthrough

• Scenario 1 (Lost Card & High Anxiety): Caller Sinem A. reports a misplaced card. The copilot locks cross-sell offers, prioritizes card block, and queues an education loan offer for post-resolution follow-up.
• Scenario 2 (Investment Inquiry & Opportunity): Caller Mehmet B. inquires about yields. The copilot detects high positive sentiment, unlocks technology mutual fund recommendations, and suggests an immediate in-app confirmation workflow.

---

📂 Repository Contents

.
├── index.html                             # Standalone interactive Copilot cockpit demo
├── akbank_canli_temsilci_destek_ekrani.html # Core UI layout mockup component
├── Adsız tasarım- Sunum.html               # Case competition presentation archive
├── Adsız tasarım- Sunum_files/             # Presentation assets & slide styles
└── README.md                              # Technical & business documentation

---

🏆 Award & Presentation

• Competition: Akbank Case Competition
• Award: 🥈 2nd Place Winner
• Evaluation Dimensions: Technical Innovation, Feasibility of Integration with Akbank Core Banking APIs, Customer Experience Impact, and Business Case.