Business Impacts
~11.3K
calls were handled by the Voicebot during the event
66.7%
of calls were deflected from live agents saving manual efforts.
25,000
queries resolved within the first 2 months of deployment.
Customer Key Facts
- Country : Qatar
- Industry : Telecommunications
Problem Context
Ooredoo is a Qatari multinational telecommunications company headquartered in Doha. Ooredoo has a customer base of more than 100 million and provides mobile, fixed, broadband internet and corporate managed services tailored to the needs of consumers and businesses across markets in the Middle East, North Africa, and Southeast Asia. Ooredoo is focused on building its Supernet network to make Qatar one of the best-connected countries in the world.
Ooredoo, being an official telecommunications partner of a global sporting mega-event which took place in the year 2022, anticipated footfall of over one million customers to attend the event and subsequently opted for a unique SIM offering, the “Hayya SIM.” Expecting a surge in call volume, Ooredoo wished to deploy a virtual agent to provide immediate assistance to the global customers in their preferred language with frequently asked questions. The voicebot was to be the first line of support to handle user queries related to Hayya SIM in seven different languages.
The client expected a voicebot that would enable Ooredoo to deflect a high volume of repetitive queries incoming to the contact center while providing a best-in-class customer experience. It would also help Ooredoo to ensure round-the-clock support and ease of convenience for the end customer.
Business Requirements
- Prepare for a large incoming volume with more than one million customers expected to purchase Hayya SIM and attend the event
- Maintain 75% or greater accuracy for the English voicebot and 70% or greater accuracy to accommodate user phrase variations in Spanish, French, Japanese, Korean, German, Brazilian and Portuguese
- Maintenance of FAQs post-handover without in-depth Dialogflow technical skills
- Manage language change feature for seven languages through Dialogflow and Cisco Telephony integration
Technologies Used
Data Studio
Dialogflow ES
Cloud Firestore
GKE
Secret Manager
BigQuery
HTTP Load balancer
Cloud Armor
Google Source Repository
Cloud Build
Container Registry
Cloud IAM
Solution
- Quantiphi developed a virtual agent using Dialogflow ES for approximately 53 FAQs related to “Hayya SIM” in seven different languages, namely English, Spanish, German, French, Portuguese, Japanese and Korean.
- To ease pressure on the Ooredoo Business Team, Quantiphi integrated the virtual agent solution with a Content Management Sheet which can be leveraged to dynamically update the responses to the questions without requiring Dialogflow ES expertise.
- Live agent handover using Cisco Telephony as well as the ability to switch languages from English to other languages (vice versa) are implemented in Dialogflow virtual agent.
- Analytics Dashboard with standard and additional Ooredoo need-based KPIs to track the performance of the virtual agent.
- As part of the handover, a detailed knowledge transfer session on Virtual Agent, Dashboard and GCP infrastructure setup was provided to the Ooredoo Team.
- During the support phase between October and December 2022, Quantiphi monitored the call logs and trained the virtual agent on user variations for respective FAQs and performed false training at certain places to improve the overall accuracy.
Results
- Achieved more than 98% accuracy of voicebot during the global sports events in Qatar 2022.
- Achieved English accuracy of 98.6%, with six additional languages achieving accuracy greater than 90% during the event.
- Around-the-clock (24x7) virtual agent availability and quick call containment, with an average handle time of two minutes.
- Achieved more than 98% accuracy of voicebot during the global sports events in Qatar 2022.
- Achieved English accuracy of 98.6%, with six additional languages achieving accuracy greater than 90% during the event.
- Around-the-clock (24x7) virtual agent availability and quick call containment, with an average handle time of two minutes.
