Who adopts enterprise AI first and how do they use it?

Linking ChatGPT Enterprise account records to usage, worker roles, task classifications, and public-company financial data through March 2026, OpenAI's internal analysis — covering "1,764 organizations and 17,446,551 messages" at the six-month adoption horizon — finds enterprise AI adoption to be "a broad but uneven organizational phenomenon." Among U.S. public companies, adopters are "larger, more valuable, and more R&D- and SG&A-intensive than non-adopters." Within adopting firms, usage "spans job functions and seniority levels" but at sharply uneven intensity: "marketing and communications workers send more messages than executives, and early-career workers send many more messages than more senior employees." Output tokens grew "roughly sevenfold between June 2025 and March 2026," with about half of that growth coming from firms already using the product rather than new adopters.

The paper reads this through general-purpose-technology theory: "initial adoption does not imply effective deployment," since realizing value "requires experimentation, complementary investment, and organizational change." Larger, more intangible-intensive firms adopt earlier because they are "better positioned to identify valuable applications, support workers in using the technology, and integrate it into business processes." The breadth of tasks observed — writing, communication, and information synthesis most common, but also research, planning, data analysis, legal work, and finance — is read as evidence of AI "functioning as a general purpose technology for knowledge work" rather than a point solution. Its summary line: "adoption is only the beginning of deployment."

This complicates Does generative AI shift knowledge workers away from communication?: where Microsoft's M365 telemetry finds use narrowing toward solo documentation work, this ChatGPT Enterprise telemetry finds use spread broadly with no single task dominating — a discrepancy between two large internal-telemetry studies that may reflect product differences (Office-embedded Copilot vs. a general chat interface) more than a settled fact about AI use. It parallels Where have workers actually delegated tasks to AI? and Is AI creating common skills across jobs or deepening divisions? in tying adoption and intensity to existing firm or worker capability rather than uniform access. Against How are national lab staff actually using generative AI?, its scale — 1,764 firms, 17 million messages of vendor telemetry versus one lab's 66-person survey — shows how differently vendor-scale and small qualitative studies answer the same question.

The excerpt measures only ChatGPT Enterprise, not other AI tools, APIs, or personal accounts; job-title coverage is incomplete with no full-workforce denominator; and message classification does not capture "downstream work products, productivity effects, or changes in organizational routines." Because this is OpenAI's own telemetry on its own product, the finding that larger, R&D-intensive firms adopt earlier describes who buys and uses ChatGPT Enterprise specifically, not a general law of AI diffusion. The data support a modest conclusion: enterprise AI use is growing and broadening, but firms, in the paper's own words, "are still actively learning how to integrate AI into organizational workflows" — well short of any claim about productivity or organizational transformation.