00:00

AI Chatbot Assistant for Better Customer Experience

It’s our in-house chatbot case study based on the company's request to improve interactions with clients with AI for customer experience.

CHI Software is a global leader in software development, delivering cutting-edge solutions for nearly two decades. We specialize in creating sophisticated and innovative solutions across artificial intelligence, IoT, cloud computing, and data science.

Project background

The goal of this AI assistant case study is to provide a solution helping website visitors with a more dynamic, engaging, and interactive method to obtain information. Designed with ease of use in mind, this AI chatbot for customer engagement is intuitive and accessible, making it easy for visitors to interact with. With machine learning, the chatbot continually improves its responses based on user interactions, ensuring it becomes more effective over time.

The integration of this AI chatbot significantly enhances the website experience by reducing response time, as users no longer need to search through static FAQs or wait for human assistance. By meeting user needs efficiently and effectively, the chatbot contributes to a more positive user experience and higher satisfaction levels.

Duration: 3 months, ongoing
Location: Global
Industry: IT Consulting & Development Services

Services:
AI Assistant Development

Business needs

Enhance Online Information:

Providing comprehensive and accessible information by AI virtual assistant about the company's capabilities and advantages on the corporate website to allow potential clients to independently research and compare with competitors.

Optimize Inquiry Handling:

Implementing a system to filter and manage inquiries on the website to reduce time spent on repetitive questions by the marketing and sales teams.

Improve Response Time:

Ensuring that leads from the Contact Form are contacted through a chatbot for FAQ within 1-2 days to maintain their interest and increase the likelihood of converting them into clients.

Product features

Implemented features:

– Natural Language Processing (NLP) Engine: Ability to understand and interpret user queries in natural language by the AI assistant for better customer experience.

– Question Analysis: Analyze user questions to identify intent and extract critical information.

– Knowledge Base Integration: Integration with a knowledge base to retrieve relevant answers to user queries.

– Response Generation: Generate responses with the AI assistant to user queries based on the question’s analysis and the knowledge base’s content.

– Error Handling: Capability to handle misunderstandings, ambiguities, or unsupported queries gracefully and provide appropriate feedback to the user.

Features in development

– Realistic avatar integration to represent the AI chatbot assistant visually.

– Gestures and Movements: Implement animated gestures and movements to make the avatar interactions more engaging and human-like.

– Lip Syncing: Synchronize the avatar’s lip movements with the spoken responses for a more realistic experience.

– Voice Synthesis: Enhance the avatar’s responses with voice synthesis technology to make interactions more lifelike.

– Learning and Adaptation: Implement machine learning algorithms to enable the assistant’s avatar to learn from user interactions and adapt its responses over time for a more personalized experience.

Visual Block

Solution

This complex project is divided into two main Phases:

Phase 1: Text-only view

Phase 2: Realistic avatar

As of today, our developers have implemented Phase 1, the text-only view, on our corporate website and are working on Phase 2, implementing a realistic avatar.

Phase 1: Text-Only View

Objective: Develop a basic Q&A system that can respond to user queries with text-based answers.

Prerequisites: A set of documents (or public URL links) with the information that would be useful for a potential customer: company information, expertise, case studies, available resources, blog articles, etc.

Enhancing AI Assistant Performance Through Integration and Adaptation

Our AI Assistant integrates APIs:

Our AI chatbot assistant connects with various external services and databases through APIs. This integration empowers the tool to perform a wide range of tasks, from fetching real-time information to executing complex actions seamlessly.

Ensuring Data Security:

Our AI assistant employs robust encryption and access controls to safeguard user information and sensitive data. We prioritize data integrity and privacy, implementing stringent measures to prevent unauthorized access and data breaches.

Leveraging Datasets:

Datasets are pivotal in training and enhancing the FAQ chatbot's capabilities, enabling it to understand and respond effectively to user queries. We utilize diverse datasets sourced from reputable sources, ensuring relevance to different user interactions and scenarios.

Self-learning Capabilities:

Our AI Assistant continuously utilizes machine learning algorithms to improve and adapt based on user interactions. By learning from past interactions, the AI chatbot assistant can provide increasingly accurate responses, enhancing the overall user experience over time.

Phase 2: Realistic Avatar

Objective: Enhance the FAQ chatbot system by introducing a realistic avatar that can visually respond to user queries by syncing lip movement and speech.

Deepfake pipeline

Inference-optimized DiNet-based lip sync pipeline

Person-specific post-training and face landmark weighted tracking algorithm:

1. Choice of source video from the library

2. Initializing with input audio stream

3. Face detection and tracking

4. Audio-based lip zone generation

5. Lip zone in-painting

6. Real-time video and audio rendering

Our technology stack

NLP
Langchain
Qdrant
Azure OpenAI Services
HTML/CSS
JavaScript
AWS
Adobe Animate
Photoshop
Illustrator
Azure TTS
Amazon Polly
Montreal Forced Alignment
DeepVoice
Vosk
Adobe After Effects
Blender
Wav2Lip
LipGAN
SadTalker

Key achievements delivered

On implementing Phase 1 of the AI virtual assistant, we have already received valuable feedback from website users and noticed the following improvements:

– Achieved a 20% boost in efficiency and responsiveness by implementing the FAQ chatbot on the corporate website;

– Delegating replying questions to the chatbot reduces account managers’ manual workload, resulting in 35% time savings and reduced administrative tasks.

– Enchanted general user satisfaction.