Page 336 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

speech-to-text model transcribes the separated speech into text for each individual, which
serves as input to LLM.
• Customer profile visualization and intent recognition: Hui Xin performs semantic
understanding and extracts key information, such as customer’s phone number. By
using this phone number as input field, the AI agent automatically calls the customer
information query Application Programming Interface (API) from backend systems and
displays the customer profile to the service representative for enhanced customer insight.
Concurrently, Hui Xin autonomously identifies the customer’s service intent from the real-
time conversation. It automatically opens the corresponding system interface and may
reduce average customer waiting time by 10%.
• Telecom products recommendation and gold-standard sales pitch generation: After
identifying the customer's service intent, the smart assistant dynamically recommends
relevant telecom products and generates a gold-standard sales pitch to guide the service
representative in customer marketing. The average marketing success rate of service
representatives has increased by over 30%.
• Instant retrieval and recommendations of telecom service knowledge: Based on the
progress of the conversation between the service representative and the customer,
smart assistant also provides real-time prompts about product details or solutions
based on dialogue context. Representatives may additionally perform Retrieval-
Augmented Generation (RAG) searches for specialized queries [2]. This supports service
representatives in better answering customer questions, especially important for the
novices.
• AI-powered autonomous orders filling: After successful marketing, it may proceed to the
system order placement phase and service representatives are no longer constrained
by the tedious tasks of the traditional GUI interface, such as extensive manual mouse-
clicking to select information and keyboard typing to input customer details. Hui Xin
extracts critical information from the conversation and, integrated with Robotic Process
Automation (RPA), autonomously populates order forms in the system interface. This GUI
form-filling paradigm frees up 20%-30% of sales representatives' time for high-value tasks
such as customer communication and marketing, cuts key process error rates by over
50%.

2�2 Benefits of the use case

The Smart Assistant enables the creation of an inclusive work environment where all staff at
business halls, including novices and individuals with manual dexterity impairments, can receive
comprehensive AI assistance. And, the entire process eliminates the need for staff to manually
operate cumbersome systems, allowing them to focus on high-value tasks such as customer
engagement, marketing, and communication. This ensures that staff can consistently deliver
high-quality telecom services without errors. The initiative aims to create more income and
employment opportunities for all staff at business halls, including novices and individuals with
manual dexterity impairments, thereby advancing decent work and economic growth.

2�3 Future Work

In the initial phase, the smart assistant for telecom business acceptance has been deployed
on the computers of two China Telecom service halls and demonstrated the ability to support
a work environment where all service representatives at service halls, including novices and
individuals with manual dexterity impairments, can deliver error-free, high-quality telecom
services consistently. Subsequent phases plan to expand the deployment to an additional 100
service halls, with gradual nationwide deployment to all service halls in the future.

300