At the recently held AMA event, Zhao Binqiang, Vice President of Technology and Product Center at Lingyi Wanwu, and Han Wei, General Manager of Solutions and Delivery in China, jointly shared the company’s forward-looking insights into the development of China’s enterprise agents in 2026. They systematically put forward six major forecasts and a “three-step” implementation path, and unveiled the latest progress of the “Wanzhi 2.5 Enterprise Multi-Agent” platform to the public for the first time.
Zhao Binqiang pointed out that from the model era to the Agent era, and then to the upcoming enterprise multi-agent era, AI is profoundly reconstructing enterprise operation models. Based on its own technological exploration and practices with multiple leading industry clients, Lingyi Wanwu has summarized the following six major forecasts:
Forecast 1: Agents Evolve from “One Person, One Tool” to “One Person, One Team”
AI will no longer be merely a tool empowering individuals, but a comprehensive AI driver that promotes the entire architecture of enterprises, from top-level managers to executive teams, enabling the “one person, one team” work model.
Forecast 2: Multi-Agents Must Possess the Three Core Elements of TAB
Team (Collaborative Operation): Achieve human-machine and machine-machine intelligent collaboration, AI-enable business operations, and make business capabilities software-defined and flexible.
Auto-pilot (Business Fission): Conduct intelligent planning based on business scenarios to realize controllable, traceable business fission and flexible growth.
Business (Business Reconstruction): Optimize business models by extracting, disassembling, and replicating team capability modules, and realize “Capability as a Service”.
Forecast 3: China to Become the Global “Super Engine” for Multi-Agents
Leveraging its advantages in data, manufacturing foundation, and rich business scenarios, China is expected to upgrade from the “world factory” to an “agent factory”, becoming a core driving force for the development of multi-agents.
Forecast 4: “Top-Down Leadership Initiative” is the Key to Capturing AI Dividends
The AI transformation of enterprises is not just partial optimization, but a holistic reshaping. Only with top-down promotion and overall planning can enterprises avoid “data silos” and truly realize the “transformation dividends”.
Forecast 5: Agents Feed Back to Unlock the “Autonomous Evolution” of Digital Infrastructure
Multi-agents can activate the “capillaries” of enterprise digital infrastructure. Through technologies such as automatic annotation and data cleaning, they can convert fragmented business data into an AI-understandable knowledge system, forming an enterprise “memory bank”.
Forecast 6: 2026 to Become the First Year of Enterprise Multi-Agent Deployment
Enterprises with sophisticated digital infrastructure, high management complexity, and long business chains will take the lead in adopting multi-agents. The focus of enterprise competition will shift from “how many people to recruit” to “how many silicon-based teams to command”, and give birth to emerging positions such as “Enterprise Agent Operations Officer”.
The “Three-Step” Implementation Path and FDE Characteristic Model
To support the implementation of these forecasts, Lingyi Wanwu proposed a “three-step” practical path:
Top-Down Leadership Guidance: Integrate into core business chains and promote the AI-First strategy.
FDE Empowerment at the Frontline: Leverage the Frontline Deployment Engineer (FDE) team to bridge organizational gaps, delve into business scenarios, and achieve technology implementation.
Collaborative Evolution to Cross the Technology Gap: Rely on the “Wanzhi 2.5 Enterprise Large Model One-Stop Platform” to support multi-agent R&D and the standardization of capability modules, effectively transforming AI into productivity.
Zhao Binqiang emphasized that Lingyi Wanwu’s FDE model is not only about technology delivery, but also about growing together with clients and providing full-process support from top-level strategy to business implementation. This model has been verified in the transformation projects of the international energy giants it serves.
Han Wei further summarized the methodology of “Top Leadership Defines Goals, Five Steps for Implementation, Two Measures for Assurance”: from disassembling business goals and designing Agent roles, to connecting data tools and selecting platform architectures, and then to continuous monitoring and optimization, forming a closed-loop delivery system.
In response to the media’s question about “how to achieve large-scale growth with the FDE model”, Zhao Binqiang elaborated on the “One Horizontal and Multiple Vertical” strategy: “One Horizontal” refers to building a scalable AI operating system with the Wanzhi Platform as the core; “Multiple Verticals” means focusing on key industries, co-building scenario-based solutions with lighthouse clients through the “Top-Down Leadership Initiative”, accumulating industry know-how, and promoting the evolution of products towards standardization and low thresholds.
Two demos, “AI Replacement for Marketing Department” and “AI Replacement for HR Department”, were showcased at the event, demonstrating the application potential of multi-agents in cross-departmental collaboration and recruitment efficiency improvement.
Zhao Binqiang introduced: “Taking this online AMA event as an example, 85% of the materials, such as dubbing, background music, poster production, and manuscript polishing, were independently generated by AI agents based on the enterprise knowledge base and customized training. Finally, only minor adjustments by the marketing team were needed before delivery.”
Enterprise-Grade Agents: Beyond Models, Focus on System Capabilities Outside the “Brain”
In response to the view that “models are equivalent to Agents”, Han Wei pointed out that in enterprise scenarios, models are only the “brain” of Agents, which must also possess four key supporting capabilities:
Secure, controllable, and compliant enterprise-level permission and audit mechanisms.
Stable tool and system invocation capabilities.
Intelligent planning and dynamic adjustment capabilities oriented to business goals.
Collaboration and verification capabilities across multiple models and roles.
He made an analogy: “Having an engine is not equivalent to having a fleet. What enterprises need is safe, stable, and reliable logistics services, not isolated engines.”
Regarding model selection, Zhao Binqiang stated that enterprises should “abandon the obsession with top-ranked models”. Instead, they need to build a hierarchical, collaborative, and cost-controllable hybrid model system based on scenario requirements, integrating lightweight, high-performance, and industry-customized models to achieve the leap from “usable” to “user-friendly”.
In addition, Han Wei revealed that Lingyi Wanwu’s revenue in 2025 achieved several times the growth of 2024, and the company plans to further expand its scale in 2026. The company is building a healthy and dynamic revenue structure through the combination of high-value overseas projects and large-scale domestic orders, while carrying out industrial intelligence cooperation with local governments, paving a sustainable path for AI commercialization.
