New OpenAI research finds that workers are using artificial intelligence to take on tasks traditionally assigned to other occupations, raising important questions about skills, job design, training, and workplace accountability.
Editorial Note
This article is provided for educational and informational purposes and does not constitute employment, financial, legal, investment, or technology-procurement advice.
The research discussed in this article was produced by OpenAI using activity associated with its own platform. OpenAI therefore has a commercial interest in how artificial intelligence is understood and adopted.
The study is descriptive rather than predictive. It does not prove that AI increases productivity, improves the quality of work, causes job losses, or eliminates the need for specialists. It shows how a particular group of U.S. users employed ChatGPT across different work-related tasks.
Artificial intelligence may be changing more than how quickly employees complete their work. It may also be changing which responsibilities they are expected to handle.
OpenAI released new economic research on July 27 examining more than 800,000 work-related messages from U.S. ChatGPT users. The study found that workers frequently used AI for tasks traditionally associated with occupations other than their own.
Across the full sample, 16.8% of work-related messages involved tasks associated with another occupation. When researchers excluded broadly shared activities such as writing emails, summarizing information, and scheduling meetings, 43.5% of occupation-specific messages crossed traditional job boundaries.
The findings suggest that AI may not simply automate existing job duties. It may enable employees to perform work that would previously have been handed to a colleague, specialist, contractor, or separate department.
A salesperson might use AI to examine a customer dataset instead of immediately sending it to an analyst. A marketer might troubleshoot a website without waiting for a developer. A small-business owner might prepare promotional content, review a contract, or perform a basic financial analysis with AI assistance.
That flexibility could create new opportunities. It could also lead employers to expect more from individual workers without providing additional pay, training, time, or professional oversight.
What OpenAI Announced
OpenAI published the first report in a new research series called Work at the Frontier.
The series is intended to examine how artificial intelligence is changing workplace activity in real time. The first report focuses on what the researchers call task crossover.
Task crossover occurs when someone uses AI to perform work historically associated with a different occupation.
The study examined users in eight occupational groups:
- customer experience;
- design;
- engineering;
- finance;
- human resources;
- legal;
- marketing; and
- sales.
Researchers compared the work described in individual messages with occupational tasks identified through the U.S. Department of Labor’s O*NET system.
The analysis then classified messages as generic, within the user’s occupation, or outside the traditional boundaries of that occupation.
OpenAI says the results offer an early view of changes that may be occurring before employers rewrite job descriptions or create new job titles.
Nearly Half of Occupation-Specific AI Use Crossed Job Boundaries
The headline finding is that 43.5% of occupation-specific messages involved work normally associated with another field.
That figure does not include generic activities that appear in many occupations, such as drafting emails or scheduling meetings.
The crossover rate differed substantially among occupational groups.
According to the report, outside-occupation tasks represented 77% of occupation-specific messages from customer-experience workers, 75% from designers, 69% from human-resources workers, 56% from legal workers, and 53% from marketers.
These figures do not mean that three-quarters of a designer’s total job has moved into another profession.
They refer only to the occupation-specific ChatGPT messages included in the analysis after generic work was removed. The unit of measurement was a user message, not an hour of labor, completed project, or full job description.
Even with that limitation, the pattern is significant.
Workers appear to be experimenting with responsibilities beyond the traditional limits of their roles.
Marketing and Engineering Tasks Travel Across Many Jobs
The research found that certain kinds of work moved across occupational boundaries more frequently than others.
Marketing and engineering tasks appeared widely among workers outside those fields.
Employees used AI to create promotional materials, develop marketing plans, troubleshoot software, explain technical systems, and complete other tasks that might once have required assistance from a specialist.
Financial calculation also appeared frequently across all seven non-finance occupational groups examined. Technology troubleshooting similarly ranked among the most common engineering-related tasks performed by workers outside engineering.
This does not mean that an AI-assisted employee becomes a qualified accountant, engineer, lawyer, or marketing professional.
It means that AI may lower the practical barrier to attempting certain tasks.
A person may be able to draft a budget, identify a software problem, create an advertisement, or summarize a regulation without immediately transferring the work elsewhere.
The quality, legality, and reliability of that result still depend on the worker’s judgment and the complexity of the task.
Small Businesses May Benefit the Most
The report found somewhat more task crossover among typical users in smaller workspaces.
Among users in the middle half of message activity, cross-occupation messages represented 18.9% of work-related use in workspaces with two to five seats. The figure fell to 16.3% in workspaces with more than 100 seats.
The difference may reflect the structure of small organizations.
A large company may employ separate specialists in finance, marketing, human resources, law, technology, communications, and data analysis.
A small business may have only a few employees. The person who identifies a problem may also be the person expected to solve it.
AI can function as a general-purpose support tool when specialist resources are limited.
A small-business owner could use it to prepare a social-media campaign, organize customer information, compare expenses, troubleshoot a website, or draft a policy.
That flexibility may help smaller organizations compete with companies that have larger teams.
It can also encourage owners and employees to attempt work that requires professional expertise they do not possess.
AI May Create More Workplace Generalists
Modern organizations often divide work into specialized roles.
One employee handles marketing. Another manages technology. A third oversees contracts, and someone else analyzes finances.
Artificial intelligence may make it easier to combine some of those responsibilities.
Employees could become more like generalists who use AI to complete an expanding mixture of tasks.
For some workers, this could be empowering.
They may gain greater independence, solve problems faster, and participate in projects previously outside their reach. Employees may also develop broader portfolios that help them qualify for promotions, leadership positions, or new careers.
For employers, task crossover could reduce delays created by departmental handoffs.
However, broader roles can also become heavier roles.
An employee who once performed one job may gradually be expected to handle portions of several jobs because AI makes those tasks appear easier.
The workplace question is therefore not only whether AI makes more work possible.
It is whether employees will receive the training, authority, compensation, and time required to perform that work responsibly.
Broader Job Duties Should Not Mean Unlimited Expectations
AI may allow employees to complete tasks more quickly, but organizations should not assume that every AI-assisted task is effortless.
A worker still has to define the problem, provide useful information, evaluate the output, correct errors, communicate results, and accept responsibility for the final decision.
Those responsibilities require time and judgment.
Employers should be cautious about expanding job descriptions without reviewing workloads and compensation.
An administrative employee using AI to produce marketing materials is still doing marketing-related work. A salesperson using AI to interpret financial information is still taking on analytical responsibility.
If these duties become routine rather than occasional, the employee’s position may need to be reclassified.
AI should not become a justification for quietly combining multiple jobs into one while leaving pay and staffing unchanged.
Specialists Will Still Matter
The research does not show that specialists are becoming unnecessary.
OpenAI explicitly cautions that making a task easier to attempt does not eliminate the need for expert review. The study measures shifting divisions of work, not which occupations will gain or lose jobs.
A marketer may use AI to create a basic website script, but a developer may still be needed to ensure that it is secure and reliable.
A manager may use AI to summarize a contract, but a qualified attorney may still be necessary to identify legal risks.
An employee may use AI for preliminary financial calculations, but a finance professional may be required to verify the assumptions and ensure compliance.
The difference between generating a plausible answer and producing professional-quality work remains important.
AI can help people enter another field’s territory. It does not automatically give them the experience, credentials, ethical duties, or professional judgment associated with that field.
The Risk of Responsibility Without Expertise
Task crossover creates a serious accountability question.
Who is responsible when an employee uses AI to perform work outside their training and the result is wrong?
A human-resources worker might use AI to interpret an employment rule. A salesperson might analyze confidential customer information. A marketer might change website code. A small-business owner might rely on an AI-generated contract.
Errors in these areas can create financial, legal, privacy, security, or reputational consequences.
Organizations need clear review rules.
Low-risk work may require only ordinary employee judgment. Higher-risk work should receive specialist approval before it affects customers, employees, contracts, finances, public communications, or regulated decisions.
Employers should also define which activities employees are prohibited from completing through unapproved AI systems.
Expanding access without establishing accountability could turn convenience into institutional risk.
Privacy Remains an Important Concern
The report was based on more than 800,000 work-related messages from U.S. users.
OpenAI states that researchers used models to classify messages anonymously and did not manually read the underlying conversations. The users’ occupations were connected through self-reported role information from ChatGPT Business accounts, while the analyzed messages came from those users’ individual ChatGPT accounts.
That methodology deserves careful attention.
Even when messages are classified anonymously, workers should understand how workplace-related activity may contribute to research and product development.
Organizations should also train employees not to place confidential, personal, proprietary, or regulated information into unapproved AI tools.
A worker attempting a task outside their usual role may not recognize the data protections that normally apply to that function.
For example, an employee helping with human resources may mishandle personnel information. Someone assisting with financial work may enter private account data. A worker troubleshooting technology may expose passwords or internal system details.
Broader capability must be paired with stronger data literacy.
Why Education and Training Must Change
The findings have major implications for schools, universities, and workforce-development programs.
Traditional career preparation often assumes that each profession has a relatively stable group of responsibilities.
Students study marketing, finance, law, technology, education, or business and then enter roles associated with those fields.
AI may make occupational boundaries less predictable.
A marketing graduate may need basic data-analysis skills. A human-resources employee may need technology literacy. A teacher may need to understand AI-assisted content creation, privacy, and digital evaluation. A small-business owner may need working knowledge across finance, marketing, law, and technology.
Education systems should not attempt to turn every student into an expert in every field.
They should help learners develop enough cross-disciplinary knowledge to recognize when they can proceed independently and when specialist support is required.
That distinction may become one of the most valuable workplace skills of the AI era.
AI Literacy Is More Than Prompt Writing
Many AI-training programs focus heavily on how to write prompts.
Prompting can be useful, but it is only one part of responsible AI use.
Workers also need to know how to:
- verify factual claims;
- recognize hallucinations and missing context;
- protect confidential information;
- identify bias;
- cite or document sources;
- understand professional boundaries;
- evaluate the quality of outputs; and
- escalate high-risk work to qualified specialists.
The ability to obtain an answer from an AI system is not the same as the ability to judge that answer.
As workers move into unfamiliar tasks, evaluation becomes even more important.
A person operating within their own field may quickly recognize a weak response. Someone crossing into another occupation may not know enough to detect the mistake.
Future training should therefore emphasize judgment, not merely generation.
Entry-Level Jobs Could Change
Task crossover may also reshape entry-level employment.
New workers have traditionally learned by completing routine assignments under the supervision of more experienced employees.
Those assignments might include drafting basic materials, organizing information, conducting preliminary research, preparing reports, or reviewing simple data.
AI can perform or accelerate many of those activities.
Employers may respond by expecting entry-level workers to handle broader and more complex responsibilities from the beginning.
That could create opportunities for well-prepared graduates.
It could also make it harder for students to enter careers if employers remove the lower-risk tasks through which beginners traditionally gained experience.
Schools and employers may need to create new apprenticeship models that teach judgment, verification, client communication, and responsible AI use rather than relying only on routine assignments.
Job Titles May Lag Behind Real Work
The report suggests that workplace activity may change before official job titles do.
An employee may continue to be called a salesperson while performing regular marketing, financial, and analytical tasks.
A designer may increasingly troubleshoot technology or create business strategies. A customer-service employee may complete work associated with sales, policy interpretation, or marketing.
Labor-market statistics and job descriptions may not immediately reflect those changes.
This matters for compensation, promotion, hiring, and professional development.
Employers should periodically examine what workers actually do rather than relying on outdated position descriptions.
Employees should also document the additional responsibilities they perform.
That record can support performance reviews, promotion requests, résumé updates, and conversations about compensation.
What the Research Does Not Prove
The study has several important limitations.
The sample does not represent the entire American workforce. It covers users connected to eight occupational groups and relies on people who used ChatGPT for work-related activity.
The researchers analyzed messages rather than completed tasks.
They did not observe whether workers used the AI-generated output, whether the result was correct, whether it saved time, or whether a specialist later reviewed it.
The study also does not estimate job losses, wage changes, productivity improvements, or the number of employees whose formal responsibilities have changed.
OpenAI describes the findings as evidence of evolving work patterns, not a forecast that particular professions will disappear.
Those limitations do not make the study unimportant.
They mean its findings should be treated as an early signal rather than a final conclusion.
A Positive Opportunity With Real Risks
The research offers a more complicated picture than the claim that AI will simply replace workers.
People appear to be using AI to expand what they can attempt.
That may help entrepreneurs, educators, nonprofit employees, freelancers, and small-business teams operate with fewer resources.
Workers may become more creative, independent, and capable of contributing across departments.
The risks are equally real.
Employees may be asked to perform specialized work without adequate expertise. Organizations may reduce staffing too aggressively. Errors may go undetected because AI output appears confident. Workers may take on additional responsibilities without additional compensation.
The most beneficial outcome will require thoughtful job design rather than uncontrolled task expansion.
How New To Education Connects AI With Career Readiness
New To Education covers artificial intelligence because it increasingly affects how students prepare for careers, how employees perform their jobs, and how organizations develop talent.
The new research shows that career readiness can no longer mean learning only the traditional responsibilities of one occupation.
Workers may need technical literacy, communication, critical thinking, ethical judgment, data awareness, and the ability to collaborate with specialists across several fields.
New To Education supports learners, educators, professionals, and businesses through career resources, educational content, professional development, consulting, and a growing community focused on lifelong learning.
Key Takeaways
OpenAI released new workplace research on July 27 based on more than 800,000 work-related messages from U.S. ChatGPT users.
The study found that 16.8% of all work-related messages involved tasks historically associated with another occupation.
After generic activities were excluded, 43.5% of occupation-specific messages crossed traditional job boundaries.
Customer experience, design, human resources, legal work, and marketing showed especially high levels of task crossover.
Marketing, engineering, financial calculation, and technology troubleshooting appeared frequently across multiple occupations.
Typical users in smaller workspaces showed somewhat more cross-occupation activity than users in larger workspaces.
The findings do not prove that AI is eliminating jobs or improving productivity.
They suggest that AI may broaden employee roles and change how organizations divide work.
Employers will need clear rules involving training, compensation, specialist review, privacy, and accountability.
Frequently Asked Questions
What happened on July 27?
OpenAI released a new economic-research report examining how workers use artificial intelligence to perform tasks beyond the traditional limits of their occupations.
What is task crossover?
Task crossover occurs when someone uses AI for work historically associated with another occupation.
How many messages were studied?
The research examined a random sample of more than 800,000 work-related messages from U.S. users.
Does the study show that AI is replacing specialists?
No. The researchers say the results show that work may be shifting across occupational boundaries. They do not estimate which jobs will be eliminated or created.
Which tasks crossed job boundaries most often?
Marketing work, technology troubleshooting, financial calculations, customer communication, and policy-related tasks appeared frequently across occupational groups.
Why was task crossover greater in small organizations?
Smaller teams may have fewer specialists available, causing workers to use AI to handle problems they might otherwise delegate.
Does the research prove that AI improves productivity?
No. The study did not measure time savings, output quality, wages, completed projects, or employment effects.
What should employers do?
Organizations should establish clear rules for data protection, specialist review, workload, training, compensation, and accountability.
Final Thoughts
Artificial intelligence may be changing the meaning of a job before it changes the name of the job.
Workers are already using AI to cross traditional occupational boundaries, taking on responsibilities that may once have required another employee or department.
That could be especially valuable for small businesses and organizations with limited resources.
It could also create a workplace where employees are expected to do more simply because AI makes more tasks appear accessible.
The difference between empowerment and exploitation will depend on how organizations respond.
Workers need training, fair compensation, reasonable workloads, and clear professional boundaries. Employers need systems that distinguish between low-risk assistance and work requiring expert review.
Education systems must also prepare students for careers in which responsibilities are broader, less predictable, and increasingly supported by artificial intelligence.
The future of work may not consist only of humans competing with machines.
It may consist of humans using machines to move into unfamiliar territory—and learning when they are qualified to continue and when they need help.
Support New To Education
New To Education publishes independent coverage of artificial intelligence, career readiness, workforce development, education, business, and the changing economy.
Readers can support this work by sharing the article, subscribing to New To Education updates, joining the New To Education community, exploring our educational and professional services, or using the support options available through the website.
Related Articles
New Education Research Shows Students Need More Than AI Access. They Need AI Readiness
https://www.newtoeducation.com/view-blog/new-education-research-shows-students-need-more-than-ai-access-they-need-ai-readiness-6a4f6fa07c40d
A New $500 Million Initiative Aims to Prepare Workers for the AI Era
https://www.newtoeducation.com/view-blog/a-new-500-million-initiative-aims-to-prepare-workers-for-the-ai-era-6a3e4cc4b023e
Sources
OpenAI — How AI Is Expanding What People Do at Work
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
