Under the spotlight of the 2025 World Artificial Intelligence Conference, Yang Liuchun, an AI solution expert from Digital China Tongminghu Cloud and Innovation-Driven Research Institute, raised a question that directly hit the core of the industry: "When Agent technology has become the global focus, why are PoC (Proof of Concept) emerging in waves, but the ones that can truly enter the main production channel of enterprises are extremely rare?" This question unveiled the pain point of the implementation of large models - large models are very close to people, but very far from enterprises.

Yang Liuchun's speech titled “AI-driven Process Transformation: Smart Vision Dual-Engine reconfiguring the new paradigm of the enterprise-level Agent implementation” and the two powerful tools released by the enterprise - Smart Vision Enterprise-level Agent Middle Platform and Intelligent Process Workbench-were precisely attempting to provide a systematic solution to this problem. Its core concept "AI for Process" directly pointed to the essence: the true lifeblood of an enterprise was not the efficiency of isolated tasks, but was the system logic and organizational resilience deeply embedded in the business process.

The Predicament:
When digital processes encountering the “pain of ossification”

The most fundamental aspect of enterprise operation was the Process - from IPD in product development, LTC in sales conversion, to ITR in customer service. The process was the cornerstone for large-scale operation. However, the digitalization of traditional processes had encountered an awkward situation:

• The paradox of efficiency: Although systems such as OA and ERP had achieved online and standardized processing of procedures, they had sacrificed flexibility due to the rigidity of rules, and "following the process" had become a symbol of inefficiency.

• Data silos: The switching between systems was complex, and business personnel were burdened with numerous operations. The data's value was dormant in isolated process links and was difficult to be transformed into decision-making wisdom.

• The intelligent gap: Failure in facial recognition, misjudgment in big data risk control... Traditional automation was inadequate in complex scenarios and unable to cope with the uncertainties of the business environment.

As Yang Liuchun said: "When a company has only a few employees, there is no need for processes at all. But when the scale expands, the processes will transform from ‘enablers’ to ‘restrictors’.” The past digitalization enhanced efficiency, but it also exacerbated rigidity.

Breaking Through:
AI for Process - Making Processes Come Alive

The "AI for Process" concept proposed by Digital China aims to build a new bridge between standardization and flexibility. Its core lies in leveraging Agent technology to endow processes with two key capabilities:

• Perception and Judgment: The Agent can understand the context, identify anomalies, dynamically adjust the path, and enable the process to have the “temperature” to cope with changes.

• Collaboration and Evolution: Agents interact with each other and with humans in real time, forming an intelligent collaboration network, which drives the continuous optimization and upgrading of processes.

"Innovation is the re-arrangement of production factors," Yang Liuchun quoted the famous saying of Schumpeter, "Agent-driven process reengineering is precisely the productivity revolution in the AI era." It does not simply replace human labor; instead, it restructures the underlying logic of enterprise operations and builds organizational resilience in an uncertain environment.

Powerful tool: Intelligent Process Workbench combining with Enterprise-level Agent Middle Platform,
Dual Engines Driving Process Transformation

How to turn ideas into practical actions? Digital China has introduced the Dual-Engine, which helps enterprises create "digital partners" - Agents with intelligent actions: learning quickly, performing well, and being able to cooperate. Through continuous scene innovation, it promotes process innovation in enterprises.

(1) Learning & Intelligence Process Workbench: The "Intelligent Commander" on the Business Side;

• Process visualization “battle map”: 5-level process decomposition and drag-and-drop configuration. Process managers can flexibly define and adjust business flows without writing code, and the overall status is clearly visible at a glance.

• Agent Agile Orchestrator Hub: AI capabilities is embedded at key nodes of the process by demands. For instance, the sales prediction Agent can recommend high-probability business opportunities, while the risk control Agent can assess contract risks in seconds.

• Data asset value loop: it can connect the process data chains such as LTC, record behavioral trajectories. By using AI analysis to output performance insights and optimization guidance, it enables data to truly drive decision-making.

(2) Smart Vision · Enterprise-level Agent Middle Platform: The “Super Training Camp” on the Capability Side;

• Expert cultivation system for the field: Integrating SFT, reinforcement learning, RAG, and DocOps document pipeline technologies, it transforms the general large model into a “master” which is deeply familiar with enterprise knowledge and processes, ensuring the reliability of knowledge and the explain ability of decisions.

• Centralized Control Hub for Complex Processes: it connects the enterprise knowledge base with the process engine, breaks down tasks into units that can be executed by Agents, and realizes the intelligent integration of knowledge and processes.

• Industrial-level operation guarantee: The heterogeneous integrated intelligent computing acceleration platform HICA, breaks through the bottleneck of key computing efficiency by eliminating the underlying computing power differences within the cluster. Its inference acceleration suite boosts performance by 10 times. The heterogeneous intelligent computing scheduling operation platform HISO supports the mixed networking and mixed scheduling of domestic and foreign GPUs, as well as fine isolation of computing power. Its heterogeneous computing power scheduling suite increases utilization by over 40% and supports stable operation with high concurrency.

• Evolutionary "Digital Partner": A groundbreaking long-term and short-term memory mechanism enables the Agent to possess the capabilities of personalized interaction and continuous learning, evolving from an "execution tool" to a partner with experience and initiative.

Dimensional Expansion:
From Tool to Partner, Reconstructing the Human-Machine Collaboration Paradigm

Yang Liuchun emphasized: "The goal of the Smart Vision Dual-Engine is not only to construct AI behavior, but also to reshape the human-machine relationship." When the process is redefined by AI:

• Human-machine collaboration: The intelligent workbench enables people to focus on decision-making and innovation, while leaving the cumbersome execution to the Agent; the visual and controllable design of the process Agent ensures that humans always remain in control and empower the growth of AI.

• Organizational Evolution: "AI for Process" enables enterprises to shift from "process standardization" to "process intelligence", and the core business is reengineered under the dual-drive of technology, resulting in an efficient linkage of the "service - data - computing power" resource pool.

• Strategic Pillar: The Agent Native application has become a "new partner" in the enterprise digital ecosystem. It serves as the intelligent infrastructure that underpins future competitiveness, rather than being an additional embellishment tool.

Process revolution, value marathon

Digital China has proposed "AI for Process" implementation framework. This framework is centered on the AI-native enterprise digitalization technology architecture, focusing on process intelligence. It relies on the " Smart Vision Enterprise-level Agent Middle Platform" and Smart Vision Intelligent Process Workbench" as Dual-Engine, and is supported by the collaborative efforts of the service resource pool, data resource pool, and AI cloud-native computing infrastructure, which aims to promote the closed-loop implementation of enterprise digitalization transformation.

In the long race of AI implementation, the two powerful tools launched by Digital China have pointed out a path for enterprises to break through the technological bubble and reach the core of value - allowing AI to no longer float in the sea of concepts, but sink into the soil of enterprise processes. Through reconfiguration and symbiosis, it will grow a brand-new productivity model belonging to the intelligent era. This is not only a profound insight into the trend of Agentic AI technology, but also a systematic consideration of the path for enterprises' digital transformation.

In the future, Digital China will continue to focus on "AI for Process", helping enterprises build intelligent, automated and sustainable-evolving business process systems, truly achieving the leap from "concept" to "value" of AI technology.