The "Guide" to AI Transformation in the Steel Industry
Recently, the State Council officially released the "Opinions on Deeply Implementing the 'AI+' Initiative" (hereinafter referred to as the "Opinions"), providing a clear direction for the deep integration and collaborative development of artificial intelligence with various industries in China. The "Opinions" emphasize that, in the area of "AI + Industrial Development," priority should be given to accelerating the cultivation of new models and business forms rooted in AI, as well as promoting intelligent transformation across all factors of industrial production. Through in-depth study and by applying these insights to my own work, I’ve gained profound insights into how "AI+" can be effectively applied in the steel industry.
Currently, against the backdrop of a complex and volatile global economic landscape, the steel industry is grappling with the dual challenges of overcapacity and sluggish demand, making it imperative to explore new paths for survival and growth. Enterprises should focus on achieving ultimate cost efficiency, operational excellence, and maximum profitability, leveraging cutting-edge technologies and innovative thinking to drive business model transformation and continuously refine production processes—thereby enabling companies to achieve high-quality development.
First is the optimization and innovation of organizational processes. Currently, the development of artificial intelligence technology has already established a clear three-stage evolutionary path: from auxiliary tools (Stage 1) to human-machine collaborative operations (Stage 2), ultimately advancing toward autonomous decision-making by intelligent agents (Stage 3). At this stage, most enterprises are at a critical window period as they transition from Stage 1 to Stage 2, and it’s anticipated that over the next five years, more companies will successfully break through to the intelligent agent-driven decision-making phase. This technological leap will trigger a fundamental transformation in organizational structures—departmental frameworks will tend toward greater flattening, while the focus of management will shift from a "human-centric" approach to an integrated "intelligent agent–human collaboration system." As a result, managers’ roles will increasingly center on designing overarching goal frameworks, dynamically allocating resources, and evaluating outcomes—while the actual execution tasks will be efficiently handled by intelligent agents equipped with self-optimizing capabilities, operating seamlessly through algorithmic feedback loops.
Therefore, future organizational restructuring must establish a three-dimensional collaborative mechanism. At the architectural design level, it is essential to clearly define the role positioning and authority boundaries of intelligent agents. At the process optimization level, we need to systematically map out three types of interaction pathways—human-to-human, agent-to-agent, and human-to-agent—while simultaneously developing dynamic feedback mechanisms and contingency response plans. Finally, at the performance evaluation level, we can introduce the OKR (Objectives and Key Results) management system, enabling intelligent agents to autonomously analyze objectives, intelligently allocate resources, and automatically provide feedback on outcomes. Of course, the integration of artificial intelligence does not mean replacing roles at the job level; instead, it creates differentiated impacts at the task level, meaning the substitution effect of AI manifests prominently at a highly granular, task-specific scale.
In the finance sector, tasks with clear rules and closed logic—such as data entry and report generation—have already achieved over 95% automation. Meanwhile, in market analysis scenarios, intelligent agents can independently handle 80% of data collection and basic modeling tasks, though complex decisions like strategic insights and anomaly attribution still require human experts to step in. This differentiated impact underscores the need for companies to build a tiered capability system: standardizing fully automated processes for routine tasks, fostering human-machine collaboration for specialized judgment, and maintaining human leadership in innovative areas. By adopting a collaborative model in scenarios such as audit compliance and customer relationship management—where "intelligent agents manage standardized workflows while human experts oversee critical decision points"—organizations can ultimately achieve an exponential boost in operational effectiveness.
Second, we must strengthen the development of our talent system. At the Smart Manufacturing Conference, Academician Zhou Ji emphasized the core principle of "smart manufacturing以人为本" (people-centered). This industrial transformation not only urgently calls for a high-caliber team that combines technical expertise with deep business insights, but will also inevitably forge a new generation of talent tailored to the intelligent era through hands-on experience and continuous refinement.
Currently, the advancement of smart manufacturing has transcended the boundaries of a single entity, giving rise to an innovative ecosystem characterized by enterprise leadership, supplier collaboration, and support from research institutions. A defining feature of this ecosystem is that business experts are increasingly stepping into the spotlight, working hand-in-hand with IT teams to drive both business insights and technological implementation.
In contrast, traditional talent evaluation systems have historically emphasized execution capabilities, assigning a dominant weight to operational tasks—often accounting for 80% of job responsibilities—which has inadvertently stifled innovation and limited room for creative thinking. However, in the era of intelligent transformation, the reconfiguration of human-machine roles is reshaping the value landscape. Under collaborative models, AI now handles 80% of standardized, repetitive tasks, freeing up human workers to focus on the critical 20% of higher-level cognitive activities—such as strategic decision-making, groundbreaking creativity, and nuanced value judgments.
This paradigm shift not only demands a fundamental overhaul of talent development frameworks but also calls for a tripartite synergy: universities laying the theoretical foundation, enterprises providing real-world application opportunities, and specialized training institutions enhancing technical skills. By dynamically adjusting curricula, refining knowledge structures, and innovating incentive mechanisms, organizations can facilitate a strategic transition in talent capabilities—from an execution-centric mindset toward one driven by innovation. Ultimately, this approach will cultivate a new generation of versatile, industry-savvy leaders who seamlessly integrate deep expertise in industrial processes with cutting-edge digital technologies.
Third, the supply chain is evolving from intelligent sensing to intelligent decision-making. By building a fully visualized supplier network ecosystem platform, companies can achieve vertical integration and horizontal collaboration across the industrial chain, creating a digital operational闭环 that encompasses demand forecasting, supply scheduling, and risk early warning. This approach helps establish an end-to-end resilient chain, ensuring the stable operation of the supply chain even in complex market conditions.
At the supplier management level, steel companies should take material attributes as the core classification dimension, comprehensively assessing strategic value, resource scarcity, and supply risks. This will enable the establishment of a strategic procurement, value-based procurement, price-driven procurement, and transparent procurement management system. Furthermore, tailored collaboration strategies can be designed for different tiers, ensuring precise resource allocation and optimized supply flexibility.
In terms of upgrading decision-making intelligence, we take orders as the central data thread, deeply integrating internal transaction and procurement execution information. At the same time, leveraging industrial operations as the data ecosystem, we capture in real time multi-dimensional signals—including external technical services, production fulfillment, financial credit, logistics dynamics, and emerging risks—thus building an intelligent sensing network that supports both internal and external dual-loop operations. By seamlessly merging supplier-tiering strategies, performance evaluation models, and dynamic risk mapping, we establish a digital twin-based capability for simulating and refining procurement strategies. This enables us to close the loop from category demand forecasting to cost-elasticity control, effectively resolving the inherent trade-off between volume and price. Ultimately, by harnessing big data and AI algorithms, we can simulate future supply-chain trends in scenario-specific contexts, delivering forward-looking, intelligent decision-making solutions.
Fourth, we will build new business models. By implementing systematic data governance and leveraging data asset monetization strategies, companies can deeply unlock the commercial potential of data, enabling them to develop differentiated profit models within their industry ecosystems. On one hand, firms can capitalize on their unique scenario advantages by launching SaaS services—such as charging a fee based on cost savings per ton of steel—or by establishing specialized platforms in the metallurgy sector that offer paid access to expert knowledge. This approach helps turn tacit expertise, like process know-how and failure case studies, into tangible, revenue-generating assets. On the other hand, companies can explore data rights confirmation and trading opportunities, creating industry-level data products from anonymized production data and equipment performance metrics. Additionally, businesses can incubate independent technology entities, adopting a light-asset model to pursue these innovative ventures, thereby driving a strategic transformation—from traditional producers into comprehensive solution providers.
In summary, digital transformation, as a systematic undertaking, requires both sustained, large-scale financial investment for support and scientifically planned guidance, enabling practical, phased progress. Industry players should actively align with the nation’s “AI+” strategic initiatives, driving the evolution of intelligent technology applications—from isolated breakthroughs at the equipment level toward seamless, end-to-end intelligent collaboration—and ultimately establishing autonomous decision-making systems that encompass all stages, including raw material formulation, process control, and quality inspection.
Over the next 5 to 10 years, as AI algorithms become deeply integrated with metallurgical principles, innovative paradigms such as fully autonomous closed-loop control across the entire process will reshape the industry's technological architecture, paving the way for a transformative shift—from experience-driven manufacturing toward data-driven, intelligent production. To this end, the author recommends strengthening policy support, with a particular focus on backing "AI+Metallurgy" benchmark projects that demonstrate significant industry-wide impact. This should include systematic assistance in areas like computing power subsidies, data openness, and standard-setting, thereby accelerating the development of new, high-quality productivity and helping China's steel industry secure a leading position in global technological competition.
