Company Name:
TDCX (MY) SDN BHD (0555268-P)
Sector
Digital Services, Digital Cities, Digital Health, Digital Finance, Digital Trade, Digital Content, Digital Tourism, Transportation, Government
Tech Enabler
Data Centre & Cloud, Artificial Intelligence (AI) and/or Big Data Analytics (BDA)
Overview:
A customer service team supporting an airlines corporation with a wide range of products and services faced challenges in efficiently addressing customer inquiries. This was due to the complexity and volume of queries received daily. The issues were heightened with regular protocol and promotion updates to the information bank. We then utilised AI and its generative capabilities to expedite information retrieval and retention. Aiming to provide accurate and timely resolutions to customer enquiries, agents would have to search a knowledge base for updated information and appropriate protocols.
Pain Points:
- Accessing, identifying, and processing the information to be suited as responses had proven to be a time-consuming task.
- Existing knowledge base lacked organization abilities, making the search for answers an overwhelming activity.
- Agents had to review multiple suggested articles and compose resolutions in the shortest time.
- Accuracy of resolutions provided was compromised especially for new agents who were unfamiliar with the business know-hows.
- Significant time would be spent to consult seniors within the team for answers, leading to reduced productivity.
- Overall longer resolution times and decreased customer satisfaction
Solution/Services:
Conversational AI KB
Works as an agent-facing copilot to solve customer queries.
Key Features and Functionalities
- Primarily utilizing Generative AI, our KB is driven by a Large Language Model (LLM) that enables conventional knowledge bases to be humanlike.
- Searches can now be done in a Q&A interaction with questions and answers written in natural language.
- Agents can also prompt the AI to generate ready-to-be-used conversational responses for quick replies to customers.
- AI will recommend prompt questions based on search histories to bridge the gap of learning for new agents.
- Multilingual translations of up to 99 languages and voice read responses for better information retention
Implementation:
- Data from product descriptions, FAQs, and support documents is transformed and ingested into the knowledge base to ensure access to relevant information.
- With the help of subject matter experts, the system is then directed to a rigorous testing cycle to ensure effectiveness of the data transformation and the AI accuracy.
- Further customizing the KB to existing answer flow, it is enriched with interactive text abilities such as generating answers in table visualizations, icon-rich document links, and active hyperlinks to minimize future information update efforts
Benefit:
A test was conducted in a real-life operation with the client’s data for 60 days. 779 customer support agents conducted 54,748 searches with an average time of 30 seconds per search.
1. Efficiency Improvement
- More than 66% of search time saved per talent.
- Average search time reduced from 90 seconds to 30 seconds.
2. Accuracy enhancement
- Achieves a high accuracy rate of 97% for generated answers
- Ensures reliable and precise information provided to customers, enhancing overall service quality.