The 4th industrial revolution

Throughout day and night, people are caught in the act of clutching a mouse or rapidly tapping keys on their keyboards, constantly inputting and querying information. The phone rings incessantly which is accompanied by a series of questions, each demanding a reply that gradually takes a toll on their voices, leading to voices becoming increasingly hoarse over time. This common scenario happens everywhere and among different people. As they gaze out the window at the traffic, they slip into daydreams, fantasizing about cloning themselves to assist with work, answer calls, gather information, and write plans. With this, all they need is to merely shift their mouth to convey the information provided by the clone. The idea of having a duplicate version of themselves to manage everything leaves them in awe.

In November 2022, a GPT application was quietly released. Within a mere two months, it became the fastest application ever to reach 100 million users. With its powerful understanding and learning capabilities, this application quickly became a global sensation, inspiring limitless possibilities. Overnight, it seemed like a storm had swept away all the “traditional” smart service enterprises. According to Valuate’s predictions, the telecom industry is expected to spend $11.2 billion on intelligent software by 2025.

Transformative changes in the contact center landscape

In the traditional attempt of applying intelligent technology to contact centers, the comprehensiveness of corpus and the difficulty of continuous training are the biggest constraints of application. Even after a perfect pre-training model goes live, a large amount of labor is required to ensure that the accuracy is maintained at a high level. As pessimists often say: “Enhancing intelligent technology requires a corresponding investment in human resources”.

With the emergence of GPT technology, it is highly anticipated that many problems will be effectively resolved. GPT has the ability to learn language patterns and rules from a vast corpus, which enables it to produce natural language texts that are coherent, logical, and semantically accurate. Given its capabilities, GPT can be utilized throughout the entire customer service process, leading to significant transformations.

As a transformer-based natural language processing model, GPT offers several advantages that can positively impact customer service.

These include:

• Natural language understanding: Deep understanding of natural language input from users, allowing for accurate identification of their intentions and requirements. This leads to more natural and seamless interactions with users, ultimately improving customer satisfaction.

• Swift Training: Reduced dependency on labeled service samples for LLM training and optimization, resulting in shorter service rollout periods.

• Automatic Generation: Support for generation tasks such as conversation summaries, which traditional intelligent models cannot perform for customer service.

• Conversation Experience: Ability to process complex open dialogues and multiple rounds of conversations, enabling better responses to multiple customer inquiries and requirements. GPT can also perform context understanding and semantic inference during conversations, leading to improved customer experience and service quality.

Opportunities come with challenges

However, there are still some concerns about the considerable gaps in the application of GPT to customer service:

• The challenge of accuracy or “serious nonsense”
• Untimely knowledge updates
• Domain-specific corpus feeding requirements
• Excessive computing power and personnel investment costs
• Intellectual property disputes
• Information security risks
• Legal risks due to incorrect answers

These issues have been partially addressed through data cleaning and filtering, pre-training in specialized domains, and operator training. Additionally, contact center professionals are inclined to prioritize the use of GPT technology for manual-assisted scenarios, such as agent prompting and conversation summary generation, while approaching end-user-facing scenarios with caution.

Harnessing innovation to drive change

Currently, the majority of contact center owners and vendors recognize the potential impact of GPT technology and are developing their own strategies and scenarios for integrating with LLMs (if they do not already have their own). They are eager to utilize GPT technology to enhance their customer service capabilities.

Huawei’s Pangu 3.0 model, one of the most highly anticipated LLMs in China, was officially launched on July 7, 2023. It includes 5 foundational models as well as multiple industry-specific and scenario-specific models. Pangu 3.0 offers a new set of capabilities, such as knowledge Q&A, copywriting generation, and code generation for NLP large models, as well as image generation and understanding for multi-modal large models, to cater to various industries.

Huawei Contact Center has been interconnected with Pangu 3.0 to provide CCaaS solution with the enhancement of LLM on public cloud, dedicating to provide end-to-end “CC + AI” solutions for contact centers with the following key features:

• Quick FAQ extraction: Automatic generation of massive FAQs, lower the period of extraction to minutes.
• Full scenario coverage: Scenario-based intelligent capabilities cover the entire contact center process, reducing the cost per call by 30%.
• Extensive domain knowledge: Industry contact center pre-training model, with a wide variety of OOTB knowledge.

In addition to interconnecting with the HUAWEI Pangu 3.0 LLM, Huawei Intelligent Contact Center also supports flexible interconnection with third-party large models.

The past 10 months have shown that change is the only constant in the world. The telecommunications field and contact center industry are calling for the next era, and now that the winds of change have arrived, we will make it happen.

Sponsored by: