20% of full-time workers in the U.S. say AI is already part of their job
A new survey released last Thursday just dropped a number worth pausing to think about: 1 in 5 full-time workers in the United States already have part of their job being done by artificial intelligence. The study also found that half of American adults used AI in the past week, whether for personal or professional purposes.
The study was conducted by Ipsos in partnership with Epoch AI, a nonprofit organization recognized for its data-driven research on the development and impact of AI worldwide. In total, about 2,000 American adults took part in the survey, answering questions about how and when they use AI tools in their daily lives. And the results paint a picture that is far more advanced than most people realize.
We are not talking about some distant trend or something confined to the tech sector. Automation is already happening inside everyday work routines, across a wide range of roles, and the pace of adoption is surprising even public policy experts.
At the same time, a curious finding pops up in the middle of all this: while some workers lost tasks to AI, another 15% started performing new duties that simply would not exist without it, with a margin of error of plus or minus 2.5%. In other words, AI takes away on one side and creates on the other, but the speed at which each of those things is happening is what really stands out here. 👇
What the survey actually revealed
When Ipsos and Epoch AI went into the field to understand how artificial intelligence is entering people’s work, they did not expect to find such a significant level of penetration. The finding that 20% of full-time workers have tasks automated by AI is especially relevant because it did not come from a tech-centric or niche sample. These are everyday workers, across a variety of industries, who now live with AI tools as part of their routine — not as a novelty or an experiment.
Caroline Falkman Olsson, who helped lead the research at Epoch AI, said the results confirmed widely held assumptions about the growing impact of AI in the workplace. According to her, when you look at what people report about their AI use, it becomes clear that there are both augmentation and automation effects at play. However, Olsson made an important caveat: more detailed research is needed to understand exactly which tasks are being affected and how people’s work environments and routines are actually changing.
The survey also shows that AI use in the workplace goes well beyond simply typing a prompt into ChatGPT and copying the response. Many respondents reported that AI already participates in structural stages of their roles, such as data screening, report generation, automated customer responses, document analysis, and even real-time decision suggestions. This represents a profound shift in how work is organized and carried out, because now it is no longer the employee doing everything and using the tool as support. In many cases, the tool executes, and the employee reviews.
Another point the survey highlighted clearly is the difference across industries. Professionals in fields like legal, finance, marketing, tech support, and customer service were the ones who most frequently reported having tasks performed fully or partially by artificial intelligence systems. And what is interesting is that many of these professionals do not see this as an immediate threat but rather as a shift in their role within their own positions. They are still employed, but they are doing different things than they were two or three years ago.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Frequency of use and AI user profiles
One revealing aspect of the survey has to do with frequency of use. Among adults who used AI in the past week, nearly 50% said they used artificial intelligence tools between two and five days during the week. However, when the study analyzed the intensity of that use, the picture showed that most — about 62.5% — completed only one or two quick tasks on their heaviest day of use. This contrasts with the roughly 6% of respondents who reported intense, continuous AI use.
This finding is interesting because it shows that even though adoption is widespread, usage is still relatively shallow for most people. AI is already part of many people’s daily lives, but for a large share of users it works more as a quick reference tool than as a system fully integrated into their work routines. The so-called heavy users, who represent a small slice, are likely the ones experiencing the most transformative effects of the technology in their professional day-to-day.
Another standout finding involves who is paying for the tool. The survey revealed that roughly half of American adults who used AI for work in the past week relied on personal subscriptions or free versions of AI services, rather than accounts provided by their employer. This suggests that many companies still have not formalized AI adoption in their operations, even though their employees are already actively using these tools. It is a scenario where adoption is being driven from the bottom up, by the workers themselves, rather than from the top down, by organizations.
AI agents are starting to show up on the radar
The survey also covered the growing use of so-called AI agents, which are systems capable of performing tasks independently, going beyond simply answering questions. Although agent technology has only gained broad industry attention in recent months, early signs of adoption are already showing up in the data.
According to the survey, 8% of AI users (with a margin of error of plus or minus 1.5%) used an AI agent in the past week. For comparison, 49% of AI users (with a margin of error of plus or minus 1.6%) used AI systems to search the web. In other words, information search still dominates AI usage, but agents are already carving out space even though the technology is so new.
Renan Araujo, program director at the Institute for AI Policy and Strategy, a nonprofit organization, pointed out that this finding about agents is remarkable. According to him, 1 in 12 Americans has already used an autonomous AI agent — meaning software that does not just answer questions but takes actions on behalf of the user. Araujo emphasized that this capability simply did not exist two years ago, and it is striking to see how quickly usage has grown.
What people are using AI for
The Epoch AI study also investigated the most common uses of artificial intelligence among American adults. The results show that AI is being used in quite varied ways, but a few categories stand out clearly.
Among adults who used AI in the past week, 80% said they used the services to look up information or get recommendations. Right behind that, 59% reported using AI to write or edit text, and 53% said they used the tools for brainstorming ideas. These three uses lead by a wide margin and show that AI is solidifying its role as a kind of intellectual assistant in people’s everyday lives.
As for the most popular tools, ChatGPT leads by a landslide, used by 31% of the sample. Next comes Gemini, from Google, at 21%, and Copilot, from Microsoft, at 10.5%. ChatGPT’s lead is no surprise, considering that the OpenAI tool was the one that popularized the concept of generative AI for the general public starting in late 2022.
The automation that creates while it replaces
One of the most surprising findings in this survey is precisely the one that seems contradictory at first glance. While the dominant narrative about AI and work revolves around replacement, the data shows that 15% of workers have started performing roles that literally did not exist before automation. This includes activities like monitoring and fine-tuning language models, curating AI-generated outputs, training systems with business-specific data, and managing automated workflows, among others. These are roles that emerged because AI emerged, and they depend on a person to work well.
This phenomenon is not exactly new in the history of technology. When electronic spreadsheets arrived, accountants who did manual calculations moved on to analyzing more complex scenarios. When the internet became mainstream, entirely new professions appeared that nobody even had names for yet. The difference now is the speed. The impact of artificial intelligence on the labor market is happening at a much faster pace than previous technological transformations, and that is what is making public policy experts genuinely concerned. Not about the direction of the change, but about the ability of people and institutions to adapt in time.
Nicholas Miailhe, an AI policy expert and leader at the Global Partnership on Artificial Intelligence (an international initiative made up of 46 countries and the European Union), said the results should serve as a wake-up call for workers and policymakers alike. According to him, when 1 in 5 workers says AI is already replacing parts of their job, you can talk about labor market restructuring happening in real time.
Miailhe also called attention to the fact that replacement appears to be advancing faster than the augmentation that AI provides. In his view, the window of opportunity for governments to help shape how AI transforms work is likely closing faster than most governments realize.
The automation that creates while it replaces also raises an important question about who benefits from that creation. The new roles that emerged with AI generally require a level of digital literacy and familiarity with systems that is not evenly distributed across the population. This means that while some workers naturally transition into these new roles, others fall behind — not because of a lack of effort, but because of a lack of access to training, time, or resources. This is the more complex side of AI’s impact that adoption numbers alone cannot capture.
Wall Street reports back up the picture
The Epoch AI survey does not stand alone. It arrives at a time when major investment banks are also publishing analyses on the impact of AI on the labor market. Economists at Goldman Sachs published new findings this week indicating that AI is eliminating roughly 16,000 jobs per month in the United States, when factoring in both the automation and augmentation effects caused by artificial intelligence.
Goldman Sachs researchers had previously estimated, back in March, that AI has the potential to automate tasks that account for about 25% of all work hours. Reports from Morgan Stanley have also been addressing the growing role of AI in the labor market. These data points coming from Wall Street serve as additional validation of what the Epoch AI survey is capturing directly from workers.
Who is Epoch AI
For those unfamiliar, Epoch AI was founded in 2021 as a volunteer project dedicated to collecting and analyzing data on trends in artificial intelligence development. The research center gained prominence in 2022 when it began studying the amount of computing power that leading AI companies were using to train their models. Since then, the organization has expanded its scope and now also studies the pricing of AI services, the construction of data centers around the world, and the types of chips used in AI model development.
The latest survey was conducted between March 3 and 5 through the Ipsos online research platform, ensuring a representative sample of the adult population of the United States.
What these numbers mean for the future of work
The impact of artificial intelligence on work is no longer a future projection — it is a snapshot of the present. And the Ipsos and Epoch AI survey is one of the first major systematic efforts to put that snapshot into concrete numbers, with solid methodology and a representative sample. What these numbers tell us is that the transformation is underway, that it is not uniform, and that the conversation about how to deal with it needs to happen now, not five years from now.
For companies, the signal is clear: automation through AI has stopped being a long-term strategic bet and has become a short-term operational reality. Organizations still in the phase of evaluating whether to adopt artificial intelligence in their processes are, in practice, falling behind those that are already measuring the impact and adjusting their teams accordingly. The question that leaders and managers need to be answering is no longer whether AI will change work, but how that change will be managed so it does not leave people on the wrong side of the equation.
For workers, the landscape holds real opportunity but also calls for real attention. The survey shows that those already using AI at work tend to report productivity gains and an opening to new responsibilities. But that does not happen automatically. It requires a willingness to learn, to experiment with new tools, and to understand that professional value today is less and less about the mechanical execution of tasks and more and more about the ability to think critically about the results that automation delivers. This shift in competencies is perhaps the biggest individual challenge that this new landscape puts on the table. 🤖
Key data highlights
- 50% of American adults used AI in the past week, for personal or professional purposes
- 20% of full-time workers in the U.S. already have tasks being performed by artificial intelligence
- 15% started performing new roles that emerged because of automation
- 62.5% of users completed only 1 to 2 quick tasks on their heaviest day of AI use
- 8% of AI users have already used autonomous agents in the past week
- 80% used AI to search for information or recommendations
- 59% used AI to write or edit text
- 53% used AI for brainstorming ideas
- ChatGPT is the most popular tool (31%), followed by Gemini (21%) and Copilot (10.5%)
- Goldman Sachs estimates that AI is eliminating roughly 16,000 jobs per month in the U.S.
