How should we measure the benefits of AI implementation in firms?

Assessing the business value of AI Does adopting artificial intelligence improve firm performance, and does moving first confer an advantage? Research by Professor Cheng-Kui Huang and his graduate student, Jheng-Siang Lin, at National Chung Cheng University finds stronger performance on selected financial indicators and higher market valuation among AI adopters. The evidence for productivity gains and first-mover advantages is less uniform. By distinguishing these outcomes, the study provides a more specific account of where the business value of AI is observed.   Linking adoption evidence with financial data The study focuses on non-IT firms in the S&P 500. It identifies AI implementation through news published between 2010 and 2017 and evaluates performance using fiscal-year 2017 data from Compustat. The final sample comprises 50 confirmed AI adopters and 250 comparison firms for which relevant AI news was not identified. Financial indicator regressions, Cobb–Douglas production functions, and a Tobin’s q-based valuation model assess financial performance, productivity, and market valuation. Additional comparisons examine adoption timing and industry groups. Figure 1 Study design   Financial and valuation benefits differ from productivity results In analyses controlling for industry differences, AI adopters show significantly stronger results for selected financial indicators, including labor productivity, profit margin, and receivables collection efficiency. Market valuation is also significantly higher. However, estimated differences in value added and output in the production-function models are not statistically significant. A favorable result on an individual financial indicator therefore should not be interpreted as evidence of improvement across every dimension of firm performance. Figure 2 Main findings   Early adoption does not guarantee an advantage Among the 50 adopters, 23 implemented AI before 2017 and 27 did so in 2017. Earlier adopters show significantly higher output, but no significant advantage in financial indicators or market valuation. Drawing on prior industry reports, the study also compares retail and finance/insurance firms with other AI adopters. This group shows higher value added without a broad financial advantage, and its market valuation is significantly lower. These findings indicate that adoption timing and industry membership alone do not guarantee benefits across all outcomes.   Implications and scope The study connects AI investment with observable firm outcomes and supports evaluating financial, operational, and market measures together. Its cross-sectional design uses a single fiscal year, and missing financial inputs substantially reduce the effective samples in the productivity models. The results describe associations in the sample, rather than establishing causal or long-term effects, and should not be treated as estimates of returns to today’s generative AI investments. Multi-year and cross-industry research can further examine how AI benefits develop over time.