Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
How Organizations Use AI: Evidence from ChatGPT
Citation. Chatterji, A., Holtz, D., Rakholia, N., Tambe, P., & Weeratunga, G. (2026). How Organizations Use AI: Evidence from ChatGPT (Working paper). arXiv:2608.12236.
Summary. This working paper links ChatGPT Enterprise account, usage, role, task-classification, and public-company financial data through March 2026 to describe organizational adoption and use. The authors find rapid but uneven enterprise diffusion across firms, worker groups, and knowledge-work tasks, and argue that organizations are still learning to integrate the technology into workflows.
Key claims
- Aggregate output tokens from ChatGPT Enterprise customers grew roughly sevenfold from June 2025 to March 2026; nearly fourfold growth within a consistent adopter cohort accounted for about half of the increase.
- Among U.S.-based public companies, ChatGPT Enterprise adopters are larger, more valuable, and more R&D- and SG&A-intensive than non-adopters; the authors interpret this as an association with prior intangible and organizational capabilities.
- At the six-month adoption horizon, the worker-level sample covers more than 1,500 organizations and 17 million messages, with observed use spanning job functions and seniority levels but showing higher intensity among early-career workers.
- The most common observed uses are writing, communication, and information synthesis, while research, planning, data analysis, legal and regulatory work, and finance also appear across organizations.
Evidence & limitations
- The study combines privacy-preserving ChatGPT Enterprise account and message-level task data with job-title and public-company financial data through March 2026; it is observational rather than causal.
- Job-title coverage varies across firms, and observed active-user composition is not a role-specific adoption rate because workforce denominators by role are unavailable.
- This is an August 2026 working-paper version on arXiv; results are explicitly subject to change, and its enterprise-product data may not generalize to non-adopting firms, other AI products, or the full workforce.