In the wave of digital transformation in manufacturing, a common understanding is emerging: the true barrier to progress isn’t the technology itself, but the people using it. According to a McKinsey survey, over 70% of digital transformation projects fail to meet expectations. The three main reasons are:
- An excessive number of siloed application systems within companies, with poor data integration and flow;
- Traditional automation processes that can’t adapt to changing business needs and are costly to maintain;
- A significant shortage of digital talent.
A white paper by China’s National Industrial Information Security Development Research Center reported a shortage of 300,000 AI professionals in 2020. In the intelligent manufacturing sector, that gap is expected to rise to 5 million this year.
Intelligent manufacturing requires more people, not fewer.
Contrary to fears of automation-driven layoffs, the widespread adoption of AI is actually fueling strong demand for new skillsets and interdisciplinary talent.
In the past, AI was mainly seen as a tool to assist with inspection, data analysis, and report generation. Today, as AI models penetrate deeper into predictive maintenance, quality control, and production scheduling, they are evolving from assistants to decision participants.
This evolution is not only redefining the role of technology, it is reshaping organizational structures. Manufacturing enterprises are shifting from a one-way model of “human decision-making with AI assistance” to a two-way model of “human-AI co-decision.” AI is no longer a background tool—it is becoming an embedded intelligence that participates in business processes, drives their evolution, and triggers their reinvention.
This shift is also changing the nature of talent demand. Companies now need not only engineers who understand AI, but also AI experts who understand manufacturing. Versatile talent with cross-disciplinary knowledge, systems thinking, and business insight will become the backbone of intelligent transformation.
This trend is evident not just in cutting-edge sectors but also among smaller, more basic manufacturing scenarios. For instance, even a 1 ton gantry crane—a lightweight piece of lifting equipment—can benefit from AI-powered improvements in production planning, inventory forecasting, and delivery optimization.
However, cultivating interdisciplinary talent in manufacturing isn’t just about stacking technical skills. It requires integrating knowledge of industrial engineering, operational technologies, and AI. These professionals must understand production-line pain points and be capable of translating AI algorithms and industrial big data into practical cost-saving and efficiency-boosting solutions on the shop floor—talents that remain rare in the current market.
While many large enterprises are investing in internal digital talent development systems, these “self-sufficient” models have clear limitations:
- Long development cycles—typically two to three years from training to integration;
- High investment costs;
- Risk of attrition—digital talent often migrates to higher-paying internet or tech sectors.
This talent “bottleneck” is amplified across the supply chain. When upstream suppliers lack a solid digital foundation, downstream enterprises struggle to build integrated smart models. The shortage of talent acts like a domino effect, slowing down the entire manufacturing sector’s progress toward intelligence.
A deeper structural challenge lies in the talent pipeline itself. Traditional engineering education lacks data-centric training, while AI talent often lacks hands-on experience in manufacturing operations. There’s a significant gap between what universities teach and what businesses need, leaving the “AI + manufacturing” talent ecosystem almost completely undeveloped.
Data & Models: The Dual Engines of AI+Manufacturing—And Why They’re So Hard to Ignite
AI can only power intelligent manufacturing when its two engines—data and models—run efficiently in tandem.
Yet in practice, many companies fall into a cognitive trap: believing that once they’ve deployed AI algorithms and connected industrial data, intelligent decision-making will happen automatically. In reality, many AI pilot projects succeed, but fail when scaled—because the data and models weren’t truly activated.
1. Data Challenges: Manufacturers Have the “Most Data,” But Often the “Least Usable Data”
2. Model Challenges: Industrial Intelligence Can’t Rely on General-Purpose AI Models
Many assume the key to a successful AI project is choosing a powerful model—ResNet, YOLO, DeepSeek, GPT-4o, etc.—and that once selected, implementation is straightforward. But in manufacturing, the algorithm itself accounts for less than 30% of the outcome. The other 70% depends on:
- Whether the data accurately reflects real-world conditions;
- Whether the semantics align with process logic;
- Whether the output delivers real value to frontline teams;
- Whether it integrates with existing systems and workflows.
AI isn’t a tool engineering challenge—it’s a system engineering challenge. Without clearly defined scenarios and real operational needs, even the most advanced models will just spin their wheels.
Three Misconceptions to Avoid:
Misconception 1: AI is plug-and-play, and can instantly reduce costs and boost efficiency
Some factories believe, “GPT is powerful, visual AI is accurate—just connect to a large model and we’re done.” But real value requires contextualized, domain-specific training. For example, what does “stick drop” refer to in a cigarette factory? Is it a filter rod? A machine error? A material issue? These domain-specific terms don’t exist in general-purpose corpora. Without industry fluency, AI becomes a clever outsider—eloquent, but clueless about shop-floor realities.
Misconception 2: With AI, skilled workers are no longer needed
This “efficiency-first” view ignores the nature of manufacturing—a symbiotic relationship between humans and systems. Frontline workers hold vast amounts of unstructured knowledge that models cannot easily replicate. True intelligence means:
- AI captures experience and aids decision-making;
- Humans retain control and pace;
- The goal is not to displace people—but to elevate them.
Misconception 3: Visual AI systems can identify defects, so decision-making is automatic
Some believe that once a visual inspection system is in place, flaws will be “handled.” But identifying a defect is only the beginning. What matters is tracing the root cause, offering actionable suggestions, and preventing recurrence. AI must be capable of cross-data correlation, process understanding, and evolutionary learning—not just automated screenshot alarms.
In the past, only the Top 10 crane manufacturers in the world could afford the in-house infrastructure and domain knowledge to support such sophisticated systems. Today, with cloud-native platforms and verticalized AI tools, more SMEs are also exploring scenario-specific, lightweight intelligent upgrades.
The Future of Manufacturing Is Not Just About Products—But About Building Smart Systems
According to recent research, 95% of manufacturing companies plan to invest in AI within the next five years. This isn’t just a technological upgrade—it’s a profound systems-level transformation.
AI is becoming the launchpad for manufacturing’s second growth curve—reshaping production logic, organizational design, and competitive strategy.
In the future, a manufacturer’s core competency won’t just be making products—it will be the ability to build a system that can perceive autonomously, optimize continuously, and collaborate intelligently.
The success of this transformation doesn’t depend on whether a company uses AI—it depends on whether it can leverage AI to reconstruct a manufacturing system built for the future.
