Virtual Agent for Travel and Hospitality
Travel & HospitalityBusiness Impacts
Personalized User Query Responses
Easy Management & Scalability
Reduced Time & Efforts
Customer Key Facts
- Location : Middle East
- Industry : Travel and Hospitality
- Technology : Conversational AI
Problem Context
The customer is one of the largest airline carriers. Their current virtual agent is placed as a widget on the company website to automate query resolution around flight status, chauffeur booking, baggage allowance, and refunds.
The virtual agent is integrated with four backend systems; CMS system, flight status system, PNR and ticket system. The solution is designed such that Genesys acts as the orchestration layer, and Dialogflow identifies the intent and entity while the response generation takes place in Microsoft Bot Framework. Virtual agent wasn’t able to handle the customer queries accurately. The customer wanted to review the conversational and visual components of the virtual agent, along with a deep dive of the Dialogflow and Microsoft Bot framework design and functioning of the agent. The customer also wanted to identify the best practices for building the virtual agent and provide suggestions on optimizations as well as team skill set and industry standard practices that should be incorporated to improve the solution.
Challenges
- Low degree of automation
- Long and staggered utterances
- Out of context questions
- Intent clashing due to similarity in phrases
Technologies Used
Google's Dialogflow
Microsoft Bot Framework
Sonarqube
Qbox
Node.js
Jenkins
Cloud Firestore
Genesys
Optimizing Virtual Agent for the Largest Airline Carrier in the Middle East
Solution
Quantiphi analyzed and reviewed the virtual agent on the following four major pillars:
1. Dialogflow Design
2. Microsoft Bot framework design
3. Conversation Design
4. End-to-end code review
We created multiple solutions for demonstrations and helped the customer improve their accuracy and enable a rich user experience. A complete report was created based on the Industry Standard best practices followed to develop the virtual agent. Also, we designed optimization techniques and evangelized suggestions to incorporate the best practices wherever applicable and necessary. Finally, a complete review of the virtual agent was captured in an industry standard documentation and an agent training optimization process was developed for the customer's perusal.
Result
- Fully deflecting 30% of calls by designing the succinct conversations and interactive screenwriting
- Increased the virtual agent success rate by 5-10% by designing the dynamic fallback responses based on the context of the user query
- Improved conversation design by ~70% by increasing coverage of the best practices across all use cases
- A complete review of the virtual agent was captured in an industry standard documentation and agent training optimization process was developed for the customer's perusal.
