OpenAI has published a report, How AI is expanding what people do at work, presenting evidence that AI adoption inside organizations is more often changing the shape of jobs than eliminating them outright. The company frames this as a shift in emphasis: rather than automation simply subtracting tasks from a role, it tends to add new ones — data review, prompt refinement, output verification, and judgment calls on edge cases that the automated system can't resolve on its own.
The report draws on usage patterns and workplace examples to argue that as AI tools absorb routine, repeatable components of a job, the humans in that role are frequently asked to take on higher-scope work: overseeing multiple automated processes at once, handling exceptions, or moving into more strategic or client-facing responsibilities that were previously out of reach because there wasn't time for them. OpenAI presents this as a net expansion of what a single worker can be responsible for, rather than a straightforward substitution of labor.
This is consistent with what has been observed anecdotally across customer support, sales development and operations functions at smaller companies, though it should be noted that OpenAI's report is self-published research from a company with a direct commercial interest in AI adoption being framed positively, and the specific methodology and sample details are worth reading in full before treating the conclusions as settled. The claims in the report are, as of this writing, unconfirmed by independent third-party labor research, and readers should treat the framing as directional rather than definitive.
Still, the pattern described tracks with what a lot of small and mid-sized companies are already experiencing in practice. A five-person support team fielding a fixed volume of tickets doesn't necessarily shrink to three people once automation handles the routine 60% — more often, the same five people end up covering a larger surface area: more customers, more channels, more complex escalations, because the routine load no longer eats their day. The same shows up in sales development, where reps freed from manual list-building and follow-up sequencing get reassigned to actual conversations and deal work, and in operations, where a coordinator who used to spend hours on data entry now spends that time monitoring three or four automated workflows for exceptions.
The implication for smaller B2B operators isn't that automation is a headcount-neutral exercise — cost and capacity gains are real and often the whole point of the investment — but that the simplest mental model ("automate the task, remove the person") tends to understate what actually happens on the ground. Job descriptions, KPIs and hiring criteria built around the old, narrower version of a role may need updating faster than the org chart does, since the people staying in those roles are often doing meaningfully different, broader work within months of an automation rollout.