Optimus, the main website for listing courses within the Optimus division of Active Steps, faced a challenge with handling visitor inquiries. Prospective students had numerous questions before enrolling, and relying on the traditional "Contact Us" form was time-consuming, negatively impacting the conversion rate.
Solution Development
Leveraging LLMs for Instant Responses
To address these challenges, we harnessed the power of Large Language Models (LLMs):
1. Training a GPT Model:
○ Technique: We trained a GPT model using the organization’s course details, embeddings, and vectorized data.
○ Technology: Pinecone was employed for efficient data storage and retrieval.
2. Integration with Weebly:
○ Implementation: Since the Optimus website runs on Weebly, we developed an integration to deploy the AI-driven chatbot as a widget/plugin within the platform.
○ Deployment: The bot was seamlessly embedded on the website, allowing users to get instant responses to their queries.
3. Internal Review and Fine-Tuning:
○ Monitoring: A continuous review system was set up to monitor and adjust the bot’s performance.
○ Capability Limitation: To minimize potential abuse, the bot’s capabilities were restricted to business-related questions.
4. Human Escalation:
○ System Design: Unanswered queries automatically escalate to human agents via email, ensuring that all questions receive attention.
Administrative Tools
To enhance the management and effectiveness of the chatbot, we developed an administrative Retool application:
1. Conversation Tracking:
○ Feature: The Retool app tracks past conversations between users and the bot.
○ Benefit: This data provides valuable insights for improving the bot’s responses and performance.
2. Model Retraining:
User Control: Admins can retrain the model with updated data and FAQs using the Retool application, ensuring the bot stays current and relevant
Impact and Benefits
1. Enhanced User Engagement:
○ Instant Answers: The chatbot allows visitors to receive instant responses to their questions, making the inquiry process more efficient.
○ Personal Recommendations: Users receive personalized recommendations about courses, enhancing their experience and satisfaction.
2. Improved Conversion Rate:
○ Metrics: The conversion rate improved significantly (by x%), demonstrating the bot’s effectiveness in addressing user queries promptly.
3. Operational Efficiency:
○ Focus on High-Level Tasks: With the chatbot handling routine queries, staff can concentrate on more critical tasks, optimizing resource utilization.
○ Escalation Handling: Only complex queries that the bot cannot answer are transferred to human agents, ensuring efficient workload distribution.
Conclusion
Our innovative integration of a GPT-driven chatbot within the Optimus website has dramatically improved the inquiry process and conversion rates. By providing instant answers to prospective students’ questions and allowing for continuous fine-tuning, we have created a robust solution that enhances user engagement and operational efficiency.
