Does AI deepen our knowledge or expand it? We study how artificial intelligence (AI) adoption reshapes firms’ innovation strategies by shifting the balance between exploration and exploitation in inventive search. Using UK firm-level patent data and the near-universe of online job vacancies from 2017–2021 (the Predictive AI era), we find that AI adoption increases firms’ overall patenting and shifts their innovation portfolios towards familiar technological domains, but does not induce entry into new technology classes. This is consistent with AI reducing the cost of optimisation and local search in data-rich areas while raising the opportunity cost of exploration in data-scarce domains. These effects are strongest for small firms, for adoption through managerial occupations and in sectors with high demand volatility. By contrast, adoption through scientific and technical occupations is more strongly associated with patenting in new domains. Overall, our findings suggest that in the pre-generative era, AI primarily deepened incumbents’ technological trajectories rather than broadening the scope of inventive search, with important implications for innovation strategy and policy.
Download the paper in PDF format
Xin Deng, Maria Garcia-Vega, Cher Li and Richard Upward
View all GEP discussion papers
Sir Clive Granger Building海角黑料University Park Nottingham, NG7 2RD