Implementing a GenAI and Microsoft PVA chatbot for a major airline

Project Summary

A major airline faced inefficiencies in managing employee communications and customer updates during telecommunication outages, relying on cumbersome email-based methods. Avec, in partnership with Microsoft, proposed a Proof of Concept (POC) for a GenAI and Microsoft Power Virtual Agents (PVA) chatbot to shift communications to web-based platforms and assess its feasibility for customer updates.  The POC was completed successfully, demonstrating the chatbot’s viability and effectiveness in improving operational efficiency and laying the groundwork for a full-scale roll-out.

The challenge

A major airline faced significant challenges in managing employee communications and customer updates, particularly during large-scale telecommunication outages. The existing methods, primarily email-based, were inefficient and cumbersome. Previous attempts to implement chatbots within the organisation had failed due to the high administrative workload required for training models with the technology available at that time. The airline needed a more efficient, scalable, and secure solution to streamline communications and improve operational efficiency.

The solution

Avec, in partnership with Microsoft, proposed a Proof of Concept (POC) for a team travel GenAI and Microsoft Power Virtual Agents (PVA) chatbot. The POC aimed to divert employee communications from email to web-based platforms and evaluate the feasibility of using the technology as a customer-facing channel for updates during outages.

Key components of the solution included:

  • Connection to Azure OpenAI GPT-4: Leveraging the advanced capabilities of Generative AI and Large Language Models to provide accurate and human-like responses while ensuring data privacy and security.
  • User Interface: An airline-approved architecture designed to mimic the final production version, allowing end users to thoroughly test the solution.
  • Minimal Model Training: The solution required minimal model training and maintenance efforts, significantly reducing the administrative workload.
  • Data Security and Transparency: Ensuring transparency regarding the destination, storage, and duration of airline data, with the ability to reduce static storage of customer data to zero days.
  • Customisable Chatbot Persona: Aligning the chatbot persona with the airline brand as designed by project stakeholders.
  • One-Click Launch: Simplifying the user experience with a one-click launch process to begin interacting with the chatbot.
  • Context Retention: The chatbot retained the context of previous parts of the conversation to provide coherent and relevant responses.
  • Privacy Notice: Displaying a privacy notice to users each time the Teams chat was opened for use.

The result

The POC was successfully completed within the stipulated timeline, demonstrating the viability and effectiveness of the proposed solution. The POC provided the airline with a practical route to divert employee communications from email to web-based platforms and demonstrated the feasibility of using the chatbot as a customer-facing channel during telecommunication outages. The solution offered clear measurements of the time and effort required to set up and maintain the chatbot, ensuring minimal administrative workload. Additionally, the chatbot’s ability to handle multiple concurrent conversations, retain conversational context, and provide accurate responses significantly improved operational efficiency.

The successful implementation of the POC laid the groundwork for a full-scale roll-out of the production-ready chatbot, with a documented path to understand all associated operating costs and efforts.