Overview
We developed an AI-powered plugin for a global financial services company to improve client-facing conversations of their relationship managers. The solution integrates directly with Google Meet and Zoom, enabling real-time speech-to-text transcription, conversation summaries, sentiment analysis, and personalized coaching feedback.
Client
-
Company: Confidential (Financial Services Enterprise)
-
Industry: Banking & Financial Services
-
Use Case: Training and improving performance of client-facing staff
Challenge
The company needed to improve the quality and compliance of client conversations while reducing the cost of manual training. Key challenges included:
-
No objective way to measure call quality or customer sentiment.
-
Manual training relied on sample reviews, leaving most calls unchecked.
-
Difficulty scaling coaching for hundreds of employees across multiple regions.
-
Compliance requirements demanded secure storage and analysis of conversations.
Solution
We built an AI Conversational Coaching Plugin with three main pillars:
-
Real-Time Transcription & Avatar Assistant
-
Speech-to-Text: Real-time transcription of Google Meet and Zoom calls (multi-language support).
-
Text-to-Speech Avatar: Interactive digital avatar for role-play training sessions with employees.
-
Live Captions & Notes: Automatically displayed during calls to assist both employees and customers.
-
Conversation Analysis & Coaching
-
Summarization: AI generates concise call summaries for CRM and compliance records.
-
Sentiment & Tone Analysis: Detects frustration, engagement, and satisfaction levels.
-
Performance Scoring: Evaluates employee responses, empathy, and adherence to scripts/policies.
-
Post-Call Feedback: Employees receive suggestions for improvement with concrete examples.
-
Integration & Security
-
Works as a plugin for Google Meet, Zoom, and MS Teams.
-
Export of transcripts and insights into CRM (Salesforce).
-
GDPR-compliant storage with role-based access control and audit logging.
-
Encryption at rest and in transit with SOC2 alignment.
Technology Stack
-
AI & NLP: OpenAI GPT-4, Whisper, Hugging Face Transformers
-
Speech-to-Text: Google Speech-to-Text API, AWS Transcribe (fallback)
-
Text-to-Speech & Avatar: ElevenLabs, Unreal Engine MetaHuman (training avatar)
-
Backend: Node.js, Python microservices
-
Integrations: Zoom SDK, Google Meet API, Salesforce API
-
Storage & Infrastructure: AWS S3, RDS PostgreSQL, ECS with Docker, Terraform IaC
-
Security: OAuth2 SSO, encrypted transcripts, role-based access, audit trails
Results
-
100% of client calls recorded & analyzed automatically.
-
25% faster onboarding of new employees through avatar role-play training.
-
40% improvement in customer satisfaction scores (CSAT) after 3 months.
-
Compliance-ready transcripts reduced audit preparation time by 60%.
-
Scalable solution deployed globally across 5 regions.
Supporting Information
-
Key Features: live transcription, conversation scoring, avatar role-play, CRM export.
-
Compliance: GDPR, SOC2, industry-specific data retention policies.
-
Usage: as a live plugin during calls, and as a training simulator for employees.
Process
-
Discovery – workshops with HR, compliance, and customer service teams.
-
MVP – built a Zoom plugin with transcription and summaries.
-
Expansion – added coaching feedback, sentiment analysis, and Salesforce integration.
-
Avatar Training Mode – deployed role-play avatar for onboarding.
-
Rollout & Training – launched across Europe and APAC, with dashboards for HR and compliance teams.
-
Continuous Improvement – fine-tuned NLP models based on recorded calls and real feedback.
Client Testimonial
“This AI solution changed how we train and monitor conversations. Every call is now an opportunity to learn, and our team is more confident and compliant than ever.”
— Confidential due to NDA