人工智能技术创新与企业全要素生产率

何小钢　郭晓斌

（江西财经大学）

摘要：人工智能技术是培育新质生产力的关键，对于推动企业高质量发展具有重要意义。本文结合人工智能（AI）领域前沿的谷歌公司开发的大语言模型（BERT模型）和专利摘要，较为精准地识别人工智能技术创新，考察了人工智能技术创新对企业全要素生产率的影响。研究发现，企业人工智能技术创新水平提高10%时，其全要素生产率提高约8.2%。机制分析表明，人工智能技术创新通过培育新质生产力，提升企业全要素生产率。异质性分析发现，在知识产权保护体系完善的地区、劳动密集型企业以及技术吸收能力较强的企业中，人工智能技术创新的生产率效应更加明显。此外，不同类型人工智能技术创新对企业全要素生产率的影响存在差异。本文的研究结论为中国制定人工智能技术创新政策和推动企业智能化转型提供了重要参考。

关键词：人工智能技术创新；全要素生产率；新质生产力；BERT模型；专利文本

作者简介：何小钢，江西财经大学数字经济学院教授、博士生导师，南昌，330013；郭晓斌，江西财经大学数字经济学院博士研究生。

基金项目：国家社会科学基金重大项目“人工智能对劳动力市场结构性变迁的影响趋势及政策应对研究”（25&ZD148）；国家社会科学基金重点项目“人工智能对制造业转型升级的影响研究”（23AZD085）；国家自然科学基金面上项目“机器人应用的就业极化与收入极化效应研究：理论机制、实证识别与政策设计”（72373057）

引用格式：何小钢,郭晓斌.人工智能技术创新与企业全要素生产率[J].经济与管理研究,2026,47(3):144-168.

AI Technological Innovation and Corporate Total Factor Productivity

HE Xiaogang, GUO Xiaobin

(Jiangxi University of Finance and Economics, Nanchang 330013)

Abstract: As the core driver of Industry 4.0, artificial intelligence (AI) is advancing innovative applications at a pace far exceeding traditional technologies. It spans diverse fields and exerts a profound influence on various industries. Firms, as vital microeconomic entities, serve as the main force in AI technological innovation. As more firms invest in AI research and development, in-depth research on the impact of AI technological innovation on corporate total factor productivity (TFP) will yield crucial policy implications in fostering high-quality economic development in the era of intelligent transformation. This paper combines the cutting-edge Bidirectional Encoder Representations from Transformers (BERT) model in the AI field with patent abstracts to accurately identify AI technological innovation and examines its impact on corporate TFP.

The findings indicate that a 10% increase in AI technological innovation levels correlates with an approximate 8.2% rise in corporate TFP. Mechanism analysis reveals that AI technological innovation fosters new types of workers, creates novel means of production, and develops new objects of labor, thereby cultivating new quality productive forces and enhancing corporate TFP. Heterogeneity analysis further shows that the productivity effects of AI technological innovation are more pronounced in regions with robust intellectual property protection systems, labor-intensive firms, and firms with stronger absorption capabilities. Further analysis indicates that AI technological innovation in the form of invention and utility model patents effectively boosts corporate TFP, whereas that registered as design patents fails to yield significant productivity gains.

This paper makes three marginal contributions. First, it employs the BERT model to construct an AI technological innovation database for listed companies, enabling more rational and precise identification of firms’ AI technological innovation levels and addressing existing literature gaps in recognizing emerging technologies. Second, it actively addresses concerns about the potential persistence of the Solow Paradox in the digital economy era and validates the productivity effects of AI technological innovation. Third, from the perspective of cultivating new quality productive forces, it enriches the transmission mechanisms through which AI technological innovation impacts corporate TFP, offering new theoretical insights into how AI technological innovation can empower high-quality development.

Based on these findings, this paper proposes the following policy recommendations: first, increase support for AI technological innovation to build new competitive advantages for firms in intelligent transformation; second, strengthen the cultivation of new quality productive forces within firms to enhance their overall competitiveness; third, reinforce the intellectual property protection system to enhance the safeguards for AI technological innovation; fourth, enhance firms’ absorptive capacity to improve the effectiveness of AI technological innovation.

Keywords: AI technological innovation; total factor productivity; new quality productive forces; BERT model; patent text

下载全文