Building Trusted Multidimensional Agent Services with AI+ Messaging

Gartner predicts that by 2025, generative AI will be embedded in 80% of conversational AI offerings. As a basic telecom service centered on message-based dialog and interaction, 5G messaging is a typical entry point for conversational AI. In China, there are already more than 1.3 million ToB Chatbot applications, a figure estimated to grow to tens of millions in three years. These applications are characterized by strong demands for general AI dialogs, industrial knowledge, enterprise applications, and domain agents, offering great development value. For ToC, the focus is on building high-frequency, must-have personal agents. All of these trends present important opportunities for AI+ messaging.

Three Layers and Two Planes: Creating Three Types of Portals for AI+ Messaging

By introducing an intelligent plane and seamlessly integrating it with the messaging plane, the network-side messaging platform forms a three-layer, two-plane AI+ messaging architecture. It creates three service portals—AI+ messaging service, AI+ information service, and AI+ application service—serving individual (ToC), business (ToB), home (ToH), and other (ToO) domains, while guaranteeing the reliability and trustworthiness of message content (see Fig. 1).

The architecture uses the intelligent control layer as its brain, aggregating and scheduling capabilities across all layers and planes.

• Data layer: Collects raw CDRs, data, and corpus from the daily operation of the network, platform, and BOSS. After data processing and cleaning, it provides core, long-term corpus for the intelligent layer.

• Intelligent layer: Includes large and small models and algorithms. Through corpus training, fine-tuning, or continuous learning, it natively generates various large AI model applications, which are tightly integrated with the messaging plane to complete the service process.

• Intelligent control layer: Acts as the front-end brain, offering capabilities such as intent recognition, task orchestration and distribution, template adaptation, and service subscription. It is the control center of AI+ messaging.

• Messaging plane: Provides mobile communication messaging services, capabilities, and channels, offering services such as SMS, MMS, 5G messaging, and industrial messaging.  

• Intelligent plane: As a new capability plane, it includes the intelligent control layer, intelligent layer, and data layer, integrating with the messaging plane to enable the AI+ capability upgrade, and providing trusted multimodal agent services for end users through three main service portals.

MLLM-Enhanced Anti-Fraud Service for Trusted Messaging Communication

According to the 2024 report from the Global Anti-Scam Alliance (GASA), consumers lost about US$1.03 trillion to to online scams in 2023. As scams continue to evolve and fraud incidents increase, they cause huge economic losses and social impacts that traditional governance solutions cannot address. The multimodal anti-fraud model application for AI+ messaging enable the upgrade from the traditional governance model to an AI-driven approach, guaranteeing the credibility of new communications.

Built on large language models (LLMs), computer vision (CV) models, and multimodal LLMs (MLLMs), this anti-fraud solution can identify and process various media types, including text, images, audio, video, and files. It features powerful semantic analysis, emotional recognition, intent analysis, and logical reasoning capabilities, greatly improving the accuracy of identifying malicious messages.

The multimodal anti-fraud LLM is highly efficient, precise, and comprehensive. With a well-curated corpus, fine-tuning, and prompt design, its accuracy consistently exceeds 95%, far surpassing traditional network governance solutions. This greatly improves the user experience, creating reliable, trusted, and worry-free communication services.

Endogenous Multimodal Agent: Messaging as an AI Service (MaaAIS)

To address the poor user experience of current 5G messaging services, as well as the high threshold, high cost, and long cycle of enabling AI applications in industry chatbots, the AI+ messaging solution, based on the platform's endogenous multimodal agent, implements the MaaAIS to enable ubiquitous AI services.

For ToB services, network-wide chatbots and domain agents offer the following features: seamless activation with no need for development or reconstruction; instant and batch commissioning; QoS guarantees; high security with data remaining within the network; and minimal cost.

For ToC services, long-term accumulation and learning from platform-level corpus and data allow the creation of personal agents that understand users best.

The endogenous multimodal agent also supports the AI-powered dialog functions across various types of messages, including SMS, MMS, and 5G messaging, bidirectional multimodal interaction, automatic web search generation, document summarization, high-precision intent recognition, and automatic card template adaptation, providing an optimal user experience.

Future Prospects

The upgrade and transformation of AI+ messaging will accelerate the development of multimodal agent entry services that are terminal-native, accessible via phone numbers, and highly reliable, driving advances in technology, services, and value:

• Technology upgrade: Integrates and develops the intelligent plane to complete the technological evolution and build next-generation messaging infrastructure.

• Service upgrade: Transitions from basic messaging communication to AI+ information service and AI+ application service.

• Value upgrade: Shifts from the traditional per-message charging model to a differentiated AI+ service-based charging model, creating new and incremental business models.

With the evolution of communications networks to 6G, AI+ messaging will pave the way for a a smarter world through messaging.