About the authors: Andrea L He is the Lawrence D and Laurie W. Fink Chair in Finance and Professor of Finance at the UCLA Anderson School of Management. Gregor Schubert Assistant Professor of Finance at the UCLA Anderson School of Management. Miao Bin Chang Assistant Professor of Financial and Business Economics at the USC Marshall School of Business.
The release of generative AI has implications for firms’ business inputs and market valuations. Anecdotal evidence is already mounting. IBM CEO Arvind Krishna told Bloomberg News this month that of the company’s 26,000 non-customer-facing jobs, “I could easily see 30% of that being replaced by AI and automation over five years.” Microsoft CEO Satya Nadella says the company needs to invest in a “major platform shift in this new era of AI,” according to a note published by Business Insider last week. The memo said Microsoft is not offering bonuses to its full-time employees this year.
This news comes as a little surprise to us. Both companies are on our shortlist of top US companies most exposed to ChatGPT, the AI chatbot released by research lab OpenAI in November. We created the list as part of a new study that builds and analyzes company-wide workforce exposure to generative AI. Our hypothesis is that the tasks performed by the workforce of companies with the highest exposure can be performed more easily by generative AI applications than the tasks performed by the workforce of companies (and industries with less exposure).
Our study builds on the idea that generative AI and related large language models will increase free cash flows at the firm level through a business effect operating through two potential channels. First, companies whose workforce can be replaced by cheaper AI-based start-up capital will see higher free cash flows through lower input costs. Second, companies whose labor inputs are more integral to generative AI will see higher cash flows due to technological improvements in one of their workforce’s complementary inputs. While we do not take a position on whether generative AI is a substitute for or complement to employment with higher exposure, our work shows that companies with a greater share of occupations exposed to AI generate experience value gains across a broad range of industries.
We constructed a quantitative estimate of the impact of technology on corporate values based on a new company-wide measure of exposure to productivity improvements from the use of generative AI. We measured the impact of the ChatGPT release on ROE at the company level. Companies whose workforces were more exposed to generative AI outperformed companies with lower exposure.
Announcement – scroll to continue
We’ve found that the impact of ChatGPT release on company values varies greatly across industries, as well as within industries across companies. While there was a large positive effect on companies with more exposure to generative AI, such as in publishing and some financial industries, there was a large negative effect in other industries, such as real estate. Value losses for incumbents are consistent with the notion that, for some industries, generative AI will drive in new entrants and displace incumbents.
To assess the impact of ChatGPT on different companies, we measured company-wide exposure to generative AI in two steps. First, we used ChatGPT to assess whether each of the 19,265 tasks currently performed by different professions could be accomplished more productively with existing ChatGPT or after investing in adapting its capabilities to business use cases. Second, we mapped occupations to publicly traded firms using a dataset on corporate occupational employment. This dataset was generated from millions of public employee profiles such as those on LinkedIn. Thus, this company-wide exposure metric captures the ability of the tasks workers in those companies are currently performing (or made more efficient) by generative AI. Our metric is the first of its kind to measure exposure to generative AI in the United States firm level.
We used transcripts of the 2022-2023 earnings calls to validate our labor-based measure of companies’ exposure to generative AI. The texts paint a similar picture. We found a strong correlation between our metric and companies’ signals of generative AI during earnings calls. We also found that Twitter call signals and AI revenue increased significantly after the release of ChatGPT.
Announcement – scroll to continue
What happens to the valuation of companies most at risk? We’ve created a model portfolio, “synthetic minus human,” that stretches long above and below AI exposure stocks. It saw daily excess returns of 0.4 percentage points higher in the two weeks after ChatGPT was released, and a cumulative 9% over four months.
Of course, assessing how big language paradigms are affecting companies is a moving target: each week brings news of more developments in what these paradigms can do or of startups and established companies adapting their capabilities to specific use cases. As a result, to the extent that these new capabilities have not yet been priced in, we are likely to see more volatility in share prices, as new companies are more likely to deploy these technologies. However, one thing is certain: given the scale and speed of technological advances we’ve seen in big language models, investors need to think creatively about how to reshape the economy — and perhaps learn how to use generative AI to their own advantage in their work in the process.
Guest reviews like this one were written by authors outside of Barron’s Newsroom and MarketWatch. Reflect the viewpoints and opinions of the authors. Send feedback suggestions and other feedback to [email protected].
#Generative #reshape #workforce #companies #exposed