OpenAI data shows ChatGPT Enterprise users used the product across a wide range of work tasks. Separately, nationally representative U.S. survey data found nearly 60% of firms reporting AI use in at least one business function used it in only one to three functions. The studies measure different populations and layers of adoption, so their percentages are not directly comparable.
Common tasks and breadth of exposure
OpenAI’s working paper, “How Organizations Use AI: Evidence from ChatGPT”, found more than half of active users performed documentation or technical-writing tasks at least once in a week, while nearly half performed technical digital work. Large shares also used the product for research, planning, data analysis, sales and marketing, legal work and financial or tax tasks.
The researchers distinguish user reach from message volume. Documentation and technical writing, technical digital work and message drafting accounted for large shares of messages. Business and market research, legal and regulatory work and financial and tax tasks reached substantial numbers of users but generated fewer messages. “The central pattern is not the dominance of a single application, but the breadth of task exposure among active users,” they wrote.
For the task analysis at the six-month adoption horizon, the researchers used a sample of 973 organizations and 8.7 million classified messages. Each message was assigned to one of 60 task categories using an automated classifier.
Variations in usage by role and industry
The mix changes with employees’ jobs. Engineering and technical practitioners were especially likely to use ChatGPT Enterprise for technical digital work and debugging, while finance and accounting workers used it for financial and tax tasks. Sales, account management, marketing and communications workers showed greater use of sales and marketing tasks.
Industry differences appeared more strongly in which tasks users tried than in the share of messages those tasks generated. Financial and tax work and business and market research were more prevalent in finance and insurance. Sales and marketing tasks appeared more often in arts, entertainment and recreation, information and retail than in manufacturing.
U.S. Census data highlights concentrated firm-level adoption
The Census data measures adoption differently, asking firms where AI is used across business functions. A Center for Economic Studies working paper using the 2026 AI supplement to the U.S. Census Bureau’s Business Trends and Outlook Survey found that, on a firm-weighted basis, 57% of firms reporting AI use in at least one business function used it in only one to three functions. Sales and marketing was most common at 52%, followed by strategy and business development at 45% and IT at 41%.
The Census researchers found similar concentration at the worker-task level: nearly 65% of firms reporting generative AI use for work tasks limited it to one to three task types. Worker task use also sometimes occurred without reported firm-level adoption, while firm-level adoption sometimes appeared without worker task use.
The measures therefore describe different layers of adoption rather than the same phenomenon. The Census worker-use measures were reported by firm respondents rather than employees themselves, so they reflect firm-level assessments of worker behavior.
Measuring exposure versus business impact
OpenAI’s telemetry covers only activity inside ChatGPT Enterprise, not competing AI systems, API-based tools, internal applications or personal accounts. Its classified messages also do not measure downstream work products, productivity or changes in organizational routines. The evidence establishes broad task exposure among active users, but not broad organizational integration or business impact.