Generative AI and
Firm Values

Andrea L. Eisfeldt, Gregor Schubert, Bledi Taska, and Miao Ben Zhang

UCLA Anderson · UCLA Anderson · SkyHive · USC Marshall

Data Repository
  • 1
    Generative AI occupation exposure scores (core and overall)
  • 2
    Generative AI firm exposure scores
  • 3
    Firm data asset measures

About This Project

The release of ChatGPT in November 2022 marked a watershed moment for generative AI adoption. This research examines how financial markets responded to this technological shift, focusing on firms' differential exposure based on their workforce composition.

We construct novel measures of generative AI exposure at both the occupation and firm levels, leveraging detailed task-level data to assess which jobs and companies face the greatest potential for AI-driven transformation. Our findings reveal significant market reactions: firms with higher AI exposure experienced positive abnormal returns following ChatGPT's release, suggesting investors expect profitability gains from AI adoption. AI can boost firm profitability in two ways: replacing occupations with core tasks exposed to generative AI, or enhancing occupations with supplemental tasks exposed to generative AI. Importantly, the market reacted positively only to core task exposure, consistent with investors expecting labor cost savings from AI-driven automation of primary job functions.

A−H

The "Artificial Minus Human" Portfolio

We construct a zero net investment portfolio that goes long on firms with the highest generative AI exposure ("Artificial") and short on firms with the lowest exposure ("Human").

Daily Return:+0.45%(t = 3.53)
Event Period Return:~5%

Data Downloads

Occupation-Level Generative AI Exposure

Measures the proportion of tasks in each occupation that can be enhanced using Generative AI tools, assessed in March 2023 based on GPT-4 capabilities.

Variable Structure (4 columns):

  • •SOC 2010 occupational codes
  • •genaiexp_estz_total- Total occupational GenAI exposure
  • •genaiexp_estz_core- Exposure from core tasks only
  • •genaiexp_estz_supplemental- Exposure from supplemental tasks
Download CSVStata Format

Firm-Level Generative AI Exposure

Measures expected weighted task share per firm that could benefit from Generative AI, based on March 2023 assessments and LinkedIn occupational employment structures.

Variable Structure (4 columns):

  • •Firm GVKEY codes
  • •firmgenaiexp_estz_total- Total firm GenAI exposure
  • •firmgenaiexp_estz_core- Core task contribution
  • •firmgenaiexp_estz_supplemental- Supplemental task contribution
Download CSVStata Format

Firm Data Asset Measures

Two complementary measures of firms' data assets that indicate readiness to leverage AI capabilities.

Variable Structure (4 columns):

  • •gvkey - Compustat firm identifier
  • •name - Company name
  • •dm3_sh- AV Data Assets: workforce share in data management roles
  • •gptdata_overall- 10K Data Assets: GPT score (0-3) from 10-K analysis
Download CSVStata Format

Citation

Suggested Citation

Eisfeldt, Andrea L., Gregor Schubert, Bledi Taska, and Miao Ben Zhang. 2026. "Generative AI and Firm Values." Journal of Finance, forthcoming.

BibTeX Entry

@article{eisfeldt2026generative,
  title={Generative AI and Firm Values},
  author={Eisfeldt, Andrea L and Schubert, Gregor and Taska, Bledi and Zhang, Miao Ben},
  journal={Journal of Finance},
  year={2026},
  note={Forthcoming}
}

Key Findings

Stock Market Reaction to ChatGPT Release

Cumulative Abnormal Returns by Generative AI Exposure

Figure 1: Cumulative Abnormal Returns by Generative AI Exposure. The graph shows cumulative abnormal returns for portfolios sorted by firms' labor-based generative AI exposure. High-exposure firms ("Artificial") outperformed low-exposure firms ("Human") following ChatGPT's release on November 30, 2022, with the spread reaching over 2% during the event period.

Generative AI Exposure Across Industries

Generative AI Exposure Within and Across Industries

Figure 2: Generative AI Exposure Within and Across Industries. This figure displays the average and standard deviation of firms' generative AI exposure within each NAICS 3-digit industry sector. Professional services, information, and finance sectors show the highest average exposure, while significant within-industry variation exists across most sectors.

Core vs. Supplemental Task Exposure

Core-Task vs Supplemental-Task Generative AI Exposure by Occupation

Figure 3: Core-Task vs. Supplemental-Task Exposure to Generative AI. Each dot represents a SOC 6-digit occupation, plotted by its core-task exposure (x-axis) and supplemental-task exposure (y-axis). Occupations below the 45-degree line have primarily core-task exposure (e.g., Word Processors, Proofreaders, Information Security Analysts), while those above have primarily supplemental-task exposure (e.g., Pharmacy Aides, Agricultural Inspectors). This distinction matters: stock markets responded positively only to core-task exposure.