Industrial Private Network Agent: Powering B2B Growth with AI-Enhanced Connectivity

Over the past year, the major trend we have observed in our interactions with global carriers is the pivoting of enterprise customers from basic physical metrics like bandwidth and latency to intelligent requirements like AI inference support, seamless data integration, and collaboration with intelligent agents. This shift underscores the transformation of carriers' B2B value propositions from simple transmission to intelligent collaboration. Private networks are transforming from mere transmission pipes into trusted intelligent production partners.

This evolution is part of the wider industry trend of digital and intelligent transformation. According to GSMA, by 2030, the average ROI for enterprise digital transformation will hit 200%, with a payback period of around 4.5 years. Investments will largely focus on AI and 5G. Major economies are setting policies and targets for the integration of AI with sectors like communications and manufacturing. Technologically, we are at a pivotal juncture. The rollout of 5G-A is accelerating, enabling new paradigms like integrated sensing and communications (ISAC) and passive IoT. These capabilities allow networks to connect to massive numbers of labels and really sense the physical world. On the AI front, foundational models are becoming more advanced, and agents are evolving from perception and cognition to autonomous decision-making and execution. Private networks are unlocking new growth avenues as smart technology is integrated into telecommunications. The convergence of market trends, policy support, and technological advancements signals a new chapter for carrier B2B business.

Huawei's Cloud Core Network Product Line has prioritized the integration of AI and private networks. Collaborating with carriers and industry partners, Huawei delivers practical solutions for the government, education, healthcare, and manufacturing sectors. The industrial private network agent exemplifies this approach in the manufacturing industry.

Industry trend: What does an AI-era private network look like for customers?

Enterprises have diverse needs for AI deployment on private networks. Large enterprises with strong digital foundations have established their own AI data centers, industry foundation models, and agent platforms. Their aim is to seamlessly integrate private networks into their existing AI systems. For this, they need schedulable capabilities and readable network status. Their intelligent agents need to be able to access terminal service status directly and invoke private network capabilities. 5G private networks need to be fully integrated into enterprise AI workflows, rather than operated as black boxes. Medium-sized enterprises need a more lightweight approach. With their smaller business scale, building AI platforms in-house would likely have low ROI. Public cloud agent services are expensive and require data transfer outside the private network. These enterprises want carriers to offer out-of-the-box, one-stop services that bundle private networks, intelligent offices, and automated Q&A, enabling affordable and comprehensive smart upgrades for office and manufacturing environments.

Private networks need to be aligned with these needs. Enterprises need to interact with private networks using AI-based methods like natural language and tokens. Service systems and agents should be able to invoke network resources directly, rather than through passive API responses. Beyond connection and positioning, private networks must be able to handle sensing, inference, and data processing. Inference computing power is moved to production sites to enable local data processing within the private network. Private networks will connect an increasing variety of terminals on the endpoint side. This will transform private networks from mere transmission pipes into intelligent production partners (see Figure 1).

Figure 1: From transmission pipes to intelligent production partners

Technological innovation: Building an industrial agent inside the private network

To meet these needs, Huawei has partnered with leading global carriers and industry partners to develop the overall architecture of the 5G-A private network's industrial agent. This architecture focuses on three functions.

• Making private networks easy for junior O&M staff to operate

  In traditional private networks, tasks like terminal configuration, 5G LAN VN group orchestration, and policy updates rely heavily on senior network engineers at each step. This leads to huge maintenance barriers and long response times for enterprises. The private network agent transforms this process. In factories with frequent production line changes, managers can simply issue commands like "add the five new robots in workshop 3 to the production group." The agent, which automatically understands the manager's intent, completes group identification, orchestration, and policy delivery. This allows junior operations staff and even service managers to use the private network directly, lowering the usage threshold and reducing O&M costs.

• Enabling networks to understand enterprise knowledge

  Core enterprise assets like production knowledge, equipment ledgers, material labels, and technique specifications are often scattered across different systems. These assets include fragmented documents, operation manuals, and structured service data. The private network agent connects to the enterprise knowledge base, service system data, and IoT tag information to ensure the network understands services and can execute tasks independently and efficiently. For instance, in warehousing, management can simply say, "check the inventory of drainage pumps for all drum washing machines in warehouse area A." The agent then associates with passive IoT tags and position data to complete the inventory check instantly. For the first time, knowledge previously locked in complex tables or accessible only to experienced personnel can now be directly read and utilized by the network.

• Orchestrating complex tasks with network and service agents in concert

  In the past, network and service systems primarily connected via APIs. This fixed interaction limited the ability of the systems to handle complex, multi-step, cross-system tasks. Now, using industry-standard protocols like A2A and MCP, the private network and enterprise intelligent center can communicate at the agent level. The enterprise's production scheduling agent can send requests directly to the network agent. The network agent responds autonomously and schedules resources based on the current status. For example, for a batch of quality inspection tasks, a scheduling agent can request low-latency links, and a private network agent completes end-to-end resource orchestration.

These three functions transform private networks into an agent that can understand service languages, grasp enterprise knowledge, collaborate with intelligent systems, and proactively participate in services. In an enterprise’s cloud-edge-device architecture, the industrial private network agent resides at the edge (as shown in Figure 2). Using A2A and MCP protocols, the agent handles diverse enterprise services. Downstream, 5G-A connects a range of terminals from low-power IoT to device AI. This streamlines data, computing power, and AI capabilities at the production site.

Figure 2: Enterprise cloud-edge-device architecture

From pilot to scale: Business practices

Only in real production environments can we see the value of technical concepts. In 2025, Huawei's Cloud Core Network Product Line collaborated with China Mobile Research Institute, China Mobile Jiangsu, and industry customers to launch the first 5G-A industrial private network agent project. This initiative won the 5G-A Industrial Private Network Intelligent Agent Award 2025 from CWW.

The project used a 5G-A + MEC + AIGC architecture, moving inference capabilities to the factory using local computing power. It integrated an enterprise knowledge base, 5G private network knowledge base, and mainstream foundation model ecosystem to deliver high-quality network assurance for major production services like AGV navigation and programmable logic control. The tested network offers 1 Gbit/s downlink speeds, 400 Mbit/s uplink speeds, end-to-end latency under 20 ms, and programmable logic control latency below 10 ms.

Network capabilities are turned into directly usable services when the industrial private network agent enters real-world factory settings. When an AGV encounters an issue, the system locates the root cause and offers handling suggestions within 10 seconds. O&M personnel can say "perform a network-wide inspection" and receive a structured report covering all domains within 15 minutes. This significantly shortens the risk check times for latency-sensitive services like programmable logic control. When employees encounter process problems on the production line, they can access the standard operation guide within 3 seconds. Previously scattered knowledge on R&D, manufacturing, and quality inspection is now readily available to employees.

The project covers diverse production settings and boosts overall service efficiency by about 30%. These solutions are being standardized and implemented in more manufacturing settings.

Working together to explore new B2B growth in the AI era

As AI integrates with private networks, carriers can expand the scope and value of their B2B services. Moving beyond traditional connection monetization, carriers are now seeing growth in payments for computing power, new applications, and new services. This extends the value chain from a mere pipe to a comprehensive combination of networks, computing, and applications.

Looking ahead, Huawei will work with carriers and industry partners to develop industry models in more real-world production settings. We aim to convert the integration of AI and private networks into business value for customers, and to work with partners along the industry chain to embrace new forms of B2B growth.