Reader question: What does OpenAI’s September work research suggest teams should measure about AI use?
Primary event: 2026-09-16 — https://openai.com/index/unlocking-new-ways-of-working/
Thesis: OpenAI’s research frames AI adoption as a change in the mix of tasks people do, so the useful management question is how work is redesigned rather than how often a tool is opened.
OpenAI’s September 16 post, “How workers are unlocking new ways of working,” describes new economic research on how people use AI for tasks outside their typical occupation and which new tasks they add to their workflow. The focus is more useful than a simple adoption count.
Usage is not the same as impact
A person can use an assistant often without changing the work that matters. OpenAI’s framing points toward a richer question: what tasks are being moved, combined or newly attempted because the tool is available? That shift can reveal where AI changes a role and where it merely speeds up an existing step.
The post is a research summary and a vendor publication. It does not establish a causal effect on productivity, wages or customer outcomes for every organization. The claims should be read as findings from the cited analysis, not as a promise about a particular workplace.
New tasks need new review
When workers begin doing tasks outside their usual occupation, the organization may need to revisit training, ownership and quality checks. A new AI-supported task can create value, but it can also move an important decision into a process nobody has clearly assigned. The research makes task design a management issue, not only a software issue.
That suggests better questions for teams: Which tasks are new? Which decisions still require a specialist? How is quality checked when the role boundary moves? Those questions can be answered without claiming that AI has already transformed a business.
Takeaway
OpenAI’s work research is a reminder to measure what people do, not only whether they have access to an AI tool. The September 16 source supports a task-centered lens and describes the research direction. It does not provide a universal productivity number or a customer result, so organizations still need evidence from their own work.
Source: the official publication linked above, dated 2026-09-16. Vendor statements are attributed and interpretation is labelled; no independent test or personal experience is implied.