-
-
Tulia AI in actions
How we built it:
We designed and developed tulia using a modern tech stack that includes React for the frontend, Django REST Framework for the backend, and PostgreSQL for the database. We integrated Celery and Redis for task management and real-time data processing. For AI-powered intent recognition, we leveraged OpenAI and Claude APIs to handle natural language processing and identify customer needs.
The project involves building a seamless connection with WhatsApp via the Whatsapp API to deliver automated responses and ecommerce functionality like ordering and payment through M-Pesa. The customer dashboard provides SMEs with full control over their product listings, pricing, and analytics, all tied to the chatbot's interactions. The backend is designed to handle a wide range of customer interactions, from simple queries to transactions, while the system escalates more complex cases to human support agents.
Challenges we ran into:
AI Memory Integration: The biggest hurdle was integrating memory into the platforms AI and conversation module to ensuring smooth chat experience where the bot remembers previous interactions.
Intent Recognition: Training the AI to accurately identify customer intent in diverse and sometimes ambiguous conversations was a complex task. We had to fine-tune our Architecture and how we interacted with LLM models to ensure that even with varied wording, the bot could understand the queries.
Payment Gateway Integration: Integrating M-Pesa payments into the system posed a challenge, especially when ensuring secure payment processing via whatsapp while maintaining a seamless user experience.
Scaling: Managing high volumes of customer interactions, while ensuring quick response times and efficient task execution (such as handling multiple customer queries simultaneously), required careful attention to our backend architecture.
Accomplishments that we're proud of:
Real-time AI-Driven Conversations: Successfully built an AI-driven WhatsApp bot that can handle customer inquiries, process transactions, and forward unresolved queries to human agents.
Ecommerce Integration: Enabled complete ecommerce functionality within WhatsApp, allowing customers to browse, order, and pay for products and services directly in the app via M-Pesa, an achievement we’re proud of given the complexity of building non-hallucinating bots and mobile payment systems in East Africa.
Customizable Dashboard: Developed a powerful and intuitive client dashboard for SMEs to manage their bot, track customer interactions, and monitor sales analytics in real-time.
Operational Efficiency: Reduced average customer response times, shortening the sales cycle and improving customer satisfaction for SMEs.
What we learned:
User Experience is Key: The simplicity and user-friendliness of the chatbot interface were essential to user adoption. We learned that focusing on clean design and intuitive flows for both customers and SME admins helped drive engagement and ease of use.
Data Is Crucial: We learned the importance of gathering and utilizing data. By collecting insights from customer conversations, we can continuously improve the bot’s performance and train it to be more accurate in handling specific business needs.
Security is Non-Negotiable: Integrating M-Pesa as well as the multi tenancy aspect of the system taught us valuable lessons on securing financial transactions and protecting customer data.
What's next for Tulia AI:
AI Training & Optimization: We plan to continue training the AI to improve its understanding of customer queries and enhance its ability to handle complex requests autonomously.
Advanced Analytics: We aim to provide even more in-depth analytics and predictive capabilities for SMEs, helping them forecast sales trends, optimize product pricing, and understand customer behavior.
Multilingual Support: Expanding Tulia’s language capabilities to support multiple languages, particularly local languages, so that businesses can cater to a wider customer base.
Expansion of Payment Gateways: We’ll integrate more payment methods and expand beyond M-Pesa to make it easier for businesses to serve customers from different regions with varying payment preferences.
Scaling: We’ll work on optimizing our infrastructure to scale efficiently for larger SMEs with growing customer bases.
Tulia AI is just getting started, and we’re excited to continue innovating and helping SMEs streamline their customer support, drive sales, and enhance overall customer experience, all from a simple WhatsApp chat.
Log in or sign up for Devpost to join the conversation.