What Holds Back Factory AI? Leadership Age and Shared Knowledge Connection
Their study covers 3,473 publicly listed Chinese manufacturing firms between 2013 and 2024, drawing on 23,469 company-year observations and 251,161 AI patent applications. The researchers examined who controlled these businesses, the average age of their top management teams and how firms combined knowledge within and across technical fields.
Developing artificial intelligence for a factory takes more than buying software or hiring people who understand algorithms. Machines, production data, engineering experience and business decisions must come together before a promising idea becomes something useful. Research into Chinese manufacturers suggests that company ownership, leadership age and the ability to connect knowledge help explain differences in AI innovation.
Xingxiu Wang and Xueshuang Xing explore these connections in "State Ownership, TMT Age, and AI Innovation in Manufacturing Firms: The Contingent Roles of Knowledge Integration Breadth and Depth," published in Sustainability. Their study covers 3,473 publicly listed Chinese manufacturing firms between 2013 and 2024, drawing on 23,469 company-year observations and 251,161 AI patent applications. The researchers examined who controlled these businesses, the average age of their top management teams and how firms combined knowledge within and across technical fields.
State ownership and older management teams were associated with lower AI innovation in the main analysis. Companies with stronger knowledge integration showed weaker versions of these negative relationships, suggesting that organizational capabilities deserve attention alongside leadership characteristics. These findings describe patterns across businesses, rather than proving that a particular ownership structure or an executive's age determines innovation.
What Ownership and Leadership Reveal
State-owned manufacturers can have access to funding and other resources that support ambitious research projects. The authors suggest that lengthy approval processes, competing responsibilities and pressure to avoid unsuccessful projects may make experimentation harder despite those advantages. Their results connect state ownership with fewer AI patent applications after accounting for factors including company size, financial performance, workforce education and differences across firms and years.
Older leadership teams showed a smaller negative association with AI innovation, which the researchers discuss in terms of familiar decision-making habits, risk preferences and readiness to explore unfamiliar technologies. Age is an imperfect indicator of those qualities, and executives of similar ages can have very different technical experience and attitudes toward change. The findings support examining leadership capabilities and decision processes, rather than treating older managers as inherently less innovative.
The Knowledge That Helps Ideas Travel
Knowledge integration describes a company's ability to bring different pieces of expertise together and use them productively. Its breadth reflects connections across technical fields, such as combining data science with production engineering, and its depth reflects specialized understanding within a particular field. The researchers estimated both dimensions using combinations of technology classifications in firms' existing patent portfolios.
Both breadth and depth softened the negative relationships involving state ownership and management age in the main models. At high levels of either capability, the estimated state-ownership relationship became positive; deeper expertise produced a similar reversal for management age, and broader knowledge weakened its negative relationship. The researchers' explanation is that shared expertise and established technical evaluation practices can make innovation less dependent on individual leaders' knowledge or willingness to take risks.
Different Factories Face Different Barriers
The researchers separated patents into AI hardware platforms, general technologies and key technologies, finding that state ownership had a negative association across all three categories. The management-age relationship was clearest for general technologies and lacked clear support in the hardware and key-technology groups. Both dimensions of knowledge integration showed supportive interaction patterns across the three categories, linking stronger capabilities with more favorable ownership and age relationships.
Industry comparisons revealed a different division, with the negative ownership relationship appearing in AI-related manufacturing and the age relationship appearing in other manufacturing industries. Breadth and depth offered broader support outside AI-related industries, where firms may need to bridge larger gaps between existing expertise and unfamiliar technologies. Within AI-related industries, the evidence chiefly connected breadth with a weaker ownership disadvantage and depth with a more favorable age relationship.
Location mattered, with the negative age relationship appearing in Beijing, Shanghai, Guangzhou and Shenzhen, and the ownership relationship appearing in other cities. Knowledge breadth and depth showed broader buffering patterns outside those four major cities, where the authors suggest firms may face greater infrastructure and institutional constraints. Among the four cities, only breadth clearly improved the relationship between state ownership and AI innovation, showing why the same organizational changes may produce different results in different places.
Building Innovation That Can Last
The study's exploratory analysis linked AI innovation with subsequent green patenting and digital transformation, with positive relationships extending across one-year and two-year lags. Green innovation was measured through patent applications, and digital transformation through an index covering strategy, technology, organizational support and other dimensions. These findings suggest potential sustainability benefits without establishing that AI directly reduced emissions, resource use or other environmental impacts.
For manufacturers, the researchers recommend simpler project approvals, longer evaluation periods and room for well-managed experiments that do not succeed. Their leadership recommendations include bringing suitable younger professionals with data-science and industrial experience into management teams, supported by expert assessment of technical proposals. Partnerships with universities and AI providers, sustained research investment, stable technical teams and documented production knowledge could help businesses develop both breadth and depth.
Policy recommendations include stronger computing infrastructure beyond major cities, training that connects AI skills with industrial experience, and evaluation systems that reward sustained innovation in state-owned firms. Support should reflect local conditions, industry needs and the technologies being developed, connecting manufacturing innovation with the sustainable industrialization goals of SDG 9. The researchers place particular value on making knowledge accessible and usable, giving firms practical support for turning technical ideas into workable projects.
The researchers checked their results using alternative measures, different models, additional research-spending controls and a sample excluding 2020–2021, finding broadly similar directions with uneven strength of evidence. The age relationship lacked clear support under a keyword-based AI measure, and some ownership-related findings weakened in alternative models, leaving important uncertainty around how consistently the patterns hold. Their methods cannot fully establish cause and effect, and patents miss unpatented improvements and the everyday sharing of practical expertise.
The sample included only listed Chinese manufacturers with sufficient existing patent activity, limiting how confidently the findings can be applied to smaller businesses, firms without patents or other countries. Green patenting and digital transformation capture only part of sustainability, leaving questions about wider environmental and social outcomes open for future research. Manufacturers need to examine how decisions are made and how expertise moves through the business when building their capacity for AI innovation.
- FIRST PUBLISHED IN:
- Devdiscourse
Google News