Inspiration

In today's fast-paced world, customers expect prompt and accurate responses to their inquiries. Traditional customer support systems often struggle with delays, inconsistencies, and high operational costs, leading to customer dissatisfaction and churn. Existing chatbots frequently fail to understand and respond correctly to customer requests, resulting in frustration and poor customer experiences. With the growth of LLM (Large Language Models) and AI models, auto-generated responses have introduced the problem of hallucination, where the AI generates incorrect or nonsensical information. As a business or brand, you don't have control over what LLMs will generate, and if they respond with something conflicting, it can damage your business and brand reputation.

HelpMate AI was inspired by the need to enhance customer support experiences while addressing these common pitfalls. By leveraging pre-verified answers, we aim to eliminate the risks of hallucinations and inaccuracies that often accompany AI models. Our solution ensures that only approved responses are sent, providing businesses with control and peace of mind. Our goal is to create a dependable and scalable AI assistant that not only improves customer satisfaction but also optimizes support operations for businesses.

What it does

HelpMate AI is an advanced AI-powered assistant designed to provide accurate, context-aware responses to customer queries. By leveraging a pre-verified knowledge base, HelpMate AI ensures that all responses are consistent and reliable, eliminating the risk of incorrect or misleading information. The assistant supports video call interactions with pre-recorded agents, creating a more engaging and interactive customer support experience.

Our infrastructure is built to be versatile, allowing HelpMate AI to be adapted for various use cases. Whether it's customer support, medical assistance, education, finance, or retail, HelpMate AI can seamlessly integrate and enhance the user experience across different industries.

How we built it

We built HelpMate AI using a combination of Microsoft Azure services to ensure scalability, efficiency, and low latency. Here’s a breakdown of our process:

Azure OpenAI: Utilized for generating responses and making decisions based on user input, providing advanced natural language processing capabilities.

Azure Cosmos DB: Employed to store our pre-verified knowledge base, ensuring quick access to accurate and reliable information.

Azure Blob Storage: Used to store multiple pre-recorded video clips, converted to HLS (HTTP Live Streaming) format, for seamless video interactions.

Azure Speech: Integrated to convert user speech to text, enabling accurate and real-time transcription of user queries.

When a user interacts with HelpMate AI, their speech is first converted to text using Azure Speech. The text input is then processed using Azure OpenAI to determine the appropriate response. The corresponding video clip is retrieved from Azure Blob Storage and played back to the user, creating an engaging and interactive experience.

For backend development, we used C# to ensure robust and efficient server-side processing. On the frontend, we leveraged Flutter to create a seamless and engaging mobile app experience, providing users with a smooth and intuitive interface. This combination of technologies and careful design ensures that HelpMate AI is both powerful and user-friendly.

Challenges we ran into

Building HelpMate AI presented several challenges, including:

-Ensuring Accuracy and Reliability:

Challenge: Maintaining the accuracy and reliability of AI-generated responses. Solution: Implemented a pre-verified knowledge base for the AI to draw from, minimizing the risk of incorrect or inconsistent answers.

-Transition to In-House Video Encoding:

Challenge: Designing a reliable alternative to Microsoft Media Services, which is reaching its end of life, by developing our in-house video encoding system. Solution: Optimized server infrastructure and managed the encoding process to handle this transition effectively.

-Low Latency and Performance:

Challenge: Ensuring low latency and high performance, particularly during video calls. Solution: Strategically chose server locations and optimized infrastructure to reduce lag for users in different regions.

Accomplishments that we're proud of

We are incredibly proud of everything we have achieved in a short period. Starting development in early May and reaching significant milestones by the end of June, we have accomplished the following:

Robust AI Assistant: Successfully developed a powerful AI assistant that consistently provides accurate and context-aware responses.
Video Call Integration: Integrated video call support with pre-recorded agents, enhancing the customer support experience with interactive and engaging interactions.
Versatile System: Created a scalable and adaptable system that can be applied across various industries, including customer support, medical assistance, education, finance, and retail.
Rapid Progress: Demonstrated rapid progress and innovation in a short timeframe, highlighting our team's dedication and capability to deliver effective solutions quickly.

These accomplishments underscore our commitment to revolutionizing customer support through advanced AI technology and seamless user experiences.

What we learned

Throughout the development of HelpMate AI, we learned several key lessons:

Combining AI with Human Oversight: Ensuring accuracy and reliability requires a blend of advanced AI capabilities and human oversight.
Server Performance Optimization: We gained valuable insights into optimizing server performance and managing latency issues to provide a smooth user experience.
In-House Video Encoding: Transitioning to our in-house video encoding system taught us a great deal about video processing and storage solutions.
User-Centric Design: The importance of user feedback was underscored, as it was crucial in refining the AI's responses and overall functionality. Real user feedback played a significant role in shaping the final product

What's next for HelpMate AI

The next steps for HelpMate AI include:

-Expanding IT Support: Currently, HelpMate AI can assist with internet connection issues or email login problems. We aim to expand these capabilities so that app users can quickly and easily resolve a wider range of IT issues.

Web Application Development: Adding a web application to complement our existing Android and iOS apps. We faced challenges with web HLS integration, which is why it was outsourced in the current deployment.
System Integration: Enhancing our system to integrate with existing platforms. For instance, if a user has an email issue, the AI Agent should connect to the email service, check if the account is locked, and unlock it if necessary. In the medical field, the AI should pull user data from the system and provide recommendations, or send the conversation between the user and the AI to a GP for review and prescription, thereby reducing the GP's workload.
Sentiment Analysis Integration: Implementing sentiment analysis to better prioritize urgent issues and improve customer support efficiency.
Analytics Dashboard Development: Creating a comprehensive analytics dashboard to provide insights into performance and customer satisfaction.
Expanding Capabilities: Enhancing support for multi-channel interactions to provide a seamless user experience across various platforms.
Exploring Additional Use Cases: Expanding into the medical, educational, financial, and retail sectors to make HelpMate AI an even more versatile and powerful tool for businesses.

By focusing on these areas, we aim to make HelpMate AI an indispensable tool for businesses, providing efficient, reliable, and intelligent customer support.

Current Store Status

We are excited to participate in the Microsoft AI Hackathon and showcase our innovative HelpMate AI app. Here is a quick update on our current release status and how you can access the app:

Android App Status: Our Android app is currently awaiting approval for Public and Open Test releases. In the meantime, you can download the app directly from our AppCenter public link: Android App Centre. This allows you to experience the full functionality of HelpMate AI on your Android device while we finalise the public release.

iOS App Status: We are actively working towards a public release for our iOS app on the Apple App Store. While this process is underway, you can access the open TestFlight link to download and test the app on your iPhone: AppStore. This gives you early access to HelpMate AI and enables you to explore its features and provide valuable feedback.

We are committed to making HelpMate AI available to all users as soon as possible and appreciate your patience and support. Thank you for your interest in HelpMate AI, and we look forward to your feedback!