In a very short time, AI has already transformed the way business is done, and the pace is almost certain to accelerate even more. But data shows that while almost every business uses AI, very few do it efficiently at scale. A 2026 OECD report found that only 4% of organizations have succeeded with high-intensity AI adoption in core business functions.
The common assumption is that AI will absorb the work traditionally handled by junior employees, leading to fewer entry-level roles. New data indicates that handing tasks over to AI may not be so seamless after all. In October 2026, I led a Founder Reports survey of 2,052 U.S.-based workers who use AI for their jobs at least once per week. 60% of the respondents who have experience training an employee said that getting good output from AI takes as much or more work than training a human. 33.1% said it takes more work than training a person.
What the Data Says
Of the 2,052 respondents, 1,856 said they have experience training an employee. The table below shows how they answered the question:
The data also holds up throughout different segments of the survey audience. 62.3% of individual contributors said AI takes as much or more work as training, and 58.4% of managers and leaders agreed. And there was virtually no difference across generations: Gen Z (61.0%), Millennials (60.8%), and Gen X (58.4%) are all within three points.
How This Impacts Workforce Planning
There are three specific takeaways from this data.
1. The savings aren’t where most leaders think they are. If training and using AI takes a similar effort to a junior hire, the budget doesn’t free up. Instead, that cost moves from recruiting and onboarding into supervision hours (often referred to as “botsitting”) that most companies don’t track. Our survey asked respondents if their company takes the time training, overseeing, and correcting AI into consideration when determining workload. 61.8% of workers who spend any time on AI oversight said their employer doesn’t fully count that time, or they weren’t sure.
2. The effort doesn’t convert. Junior hires, ideally, will move into more senior roles in the future. AI will not. The time and effort training a human is a long-term investment, not simply an expense. For now, AI may seem like the cheaper option, but it may come with a long-term cost when the company is forced to bring in outside candidates for management roles because there is no one in place to move up.
3. The hiring freeze is quietly building a management problem. Our survey showed that 40.7% of managers and above spend at least three hours per week training, supervising, or fixing AI’s output, compared to 25.0% of individual contributors. The botsitting burden currently falls heavier on leadership, distracting from their ability to lead.
What’s Happening in Entry-Level Hiring
In KPMG’s Q1 2026 AI Quarterly Pulse, 64% of leaders said AI agents have already changed entry-level hiring, and 57% expect humans will primarily manage and direct AI agents within two to three years. The direction is set. Stanford researchers using ADP payroll data found that employment of 22-to-25-year-olds in AI-exposed occupations is already 19% below where it would be if it had kept pace with less-exposed occupations.
None of this is an argument against AI adoption. But data indicates that the current process is inefficient for many companies and workers.
The 40% Who Disagree
While 60% of respondents told us using AI takes as much or more effort than training a human, the other 40% disagreed. And 16.7% said that AI takes much less. That’s still a sizable portion of the audience, and it gets more interesting when you drill down into some of the details.
The workers most convinced that AI pays off are the ones who use it the most. Among respondents who spend more than 10 hours a week on AI oversight, 32.9% said they would promote their AI if it were an employee, and 41.0% said they would get much less done without it. Among workers who use AI only once a week, those figures fall to 8.0% and 6.9%.
The same pattern is visible when you consider frequency of use:
Although these workers are currently in the minority, they (or their companies) have built real workflows around AI, dedicated hours to learn what it can and can’t do, and integrated it into their daily work. Their experience with AI looks drastically different than the overall picture of the larger audience.
Why It Matters
The real takeaway from this survey isn’t that AI can or can’t perform the role of a junior employee. The more significant finding is that most companies, right now, aren’t using AI in the way that produces the results they expect.
The difference between those who are getting positive results with AI and those who aren’t isn’t really about which tools they use. It’s more about how they’ve built the work around those tools.
Many companies and employees take the workflow they’re already using, look at the individual tasks in it, and ask which ones could be handed off to an LLM or an AI tool. It’s AI bolted onto a process designed for humans. It’s also what produces the headline result: effort comparable to training a human, and often less of a payoff.
The organizations getting better results took a different approach. They’ve looked at the work itself and rebuilt the process with AI already in it. Instead of trying to find random spots to use AI, they’ve adapted their entire workflow.
Companies must make that shift before assuming that AI will replace the work of a junior employee. Current headlines make it seem like AI has already reduced the need for entry-level workers. The survey data shows that AI is working well for a small number of companies, and has added supervision work almost everywhere else.
Instead of focusing on using AI to replace individual tasks, leaders should work to rebuild systems and workflows in a way that allows for efficient and effective use of AI. And that rebuilt system should still include junior employees to preserve a pipeline for future management roles. The real decision isn’t between AI or junior workers, but about developing a workflow that involves both efficiently.
The full Botsitting Report with all the survey data is available at Founder Reports.
