OpenAI launched ChatGPT on July 27, 2026 How do you use it in your work? I released a new study that I analyzed.
Part of what stands out in this study in particular is not productivity or speed of response. The point is that the person who uses AI goes beyond the scope of their job and starts directly requesting and trying to do things to ChatGPT that would have been handed over to other experts in the past.
A marketer solves a simple website error, a salesperson analyzes customer data, and a small business owner takes on everything from writing promotions to doing basic financial analysis.
So is AI really pushing the boundaries of work?
| Analyzed over 800,000 work messages
The OpenAI Economic Research Team randomly analyzed more than 800,000 work-related messages from US ChatGPT users.
The job family information used the information entered when signing up for ChatGPT Business, and the analyzed message is a work-related message from the user's personal ChatGPT account.
Eight job categories are analyzed: Customer Experience, Design, Engineering, Finance, Human Resources, Legal, Marketing, and Sales. We compared the job family information entered by the user with O * net, the job database of the U.S. Department of Labor, and classified whether each message originally corresponded to their job or was close to the job of another job family.
In this process, tasks common to many professions such as writing emails, summarizing documents, and managing schedules were separated into ‘general tasks’.
The researchers stated that they did not read the individual messages directly, but used the model to classify the type of work in an anonymous state.
| Key numbers are 16.8% and 43.5%
Of all the work-related messages 16.8%was classified as a task outside the user's original job family.
However, aside from general tasks such as email creation and scheduling, this ratio is 43.5%Climb up to.
AI use cases with clear occupational colors More than four out of ten cases are jobs that originally exceeded their job responsibilitiesIt means it was.
However, you can't accept 43.5% as half of all work messages here. The total standard including general work is 16.8%, and 43.5% is recalculated after excluding general work.
There was a big difference by job group. The proportion of other job groups in the message except general work was as follows.
Customer Experience Category: 77%
Designer: 75%
Human Resources: 69%
Legal occupations: 56%
Marketer: 53%
In particular, Customer experience, design, and human resourcesgoes above and beyond helping ChatGPT in its original role. Tools to make up for a lack of expertiseIt was used a lot as.
| The most widespread were marketing and engineering jobs
The tasks that have spread the most to other occupations Marketing and EngineeringIt was.
Designers create promotional materials, sales reps write marketing statements, and customer service reps solve software problems. Financial calculations and computer troubleshooting have frequently appeared in almost every other job family subject to analysis.
The interesting thing here is that different job families used AI in different directions.
The designer was close to bringing a lot of work from other fields. 35.2% of the messages from designers were from different job groups, but only 1.7% of the design work appeared in messages from different job groups.
On the contrary, engineering was used a lot by other occupations. The external jobs accounted for 18.5% of the engineer's messages, but the engineering jobs accounted for an average of 7.4% of user messages in other fields.
Marketing showed both personalities. Marketers have tried a lot in other fields, and marketing has also spread to various occupations.
| Smaller organizations took on more roles with AI
Expanding jobs with AI Small WorkspaceIt was a little more pronounced in.
Based on users with medium message usage, the proportion of work in other occupations was 18.9% in workspaces with 2 to 5 seats. In workspaces with more than 101 seats, it's down to 16.3%.
In large organizations, there are specialized departments such as development, legal, and finance, but in small organizations, it can be interpreted that the person who encountered the problem right away is likely to solve it directly with the help of AI.
However, the number of seats in the workspace is not the same as the actual total number of people in the company. In the group of users with high usage, there was no constant decrease in the proportion depending on the size. Therefore, it cannot be concluded that “the smaller the company, the more reliant on AI”.
| But that doesn't mean AI eliminates experts
With this result “AI is replacing jobs”It's hard to conclude.
What the study confirms is what tasks people have requested from ChatGPT. We don't know if we actually used the deliverables, how much time we saved, if the answers were good enough, or if the experts reviewed them again.
The subject of analysis is also limited to eight job families of American users, so it does not represent the entire labor market. OpenAI also clearly explained that this study did not measure employment reduction or productivity improvement.
After all, thanks to AI, even non-professionals AttemptIt has become easier to do, but the professionalism of judging and taking responsibility for the results has not disappeared.
It means that even if you can leave it to AI to draft contracts or create simple code, it is difficult to pass important legal judgments or even security reviews.
| What's important in the future is more than the ability to deal with AI Ability to review
The change this study shows is not that jobs disappear all at once. Close to a sign that the combination of tasks of one person is changing.
In the future, “What is my job?” “How far can we scale with AI”is likely to become important. At the same time, it is necessary to have the ability to judge how far you can believe the results that are not your area of expertise.
If AI is a tool that lowers the barriers between jobs, then human roles can move further towards reviewing outcomes, contextualizing decisions, and taking responsibility.
To summarize, ChatGPT is moving from a simple job aid to a general-purpose tool that helps one person take on multiple roles. The name of the job isn't going away right away, but the name that goes into it The actual to-do list is already starting to look differentYou can go and see it.
It's been really hot lately. Take a break, bring water often, and have a healthy summer!
Thank you for visiting today!!
Source of Data
OpenAI, “How AI is expanding what people do at work,” July 27, 2026, https://openai.com/index/how-ai-is-expanding-what-people-do-at-work/
OpenAI Economic Research, 「Work at the Frontier: How AI is expanding what people do at work」, https://cdn.openai.com/pdf/work-at-the-frontier-report.pdf
#ChatGPT #OpenAI #Artificial Intelligence #Generative AI # AI Business #Work Automation #AI Productivity # Future of work # Future Job #IT News
