Will AI Agents Replace Employees? What the Data Says

So far, the best survey data we could open shows little net change in employment from AI overall, and no study we opened counts jobs replaced by AI agents specifically. Executives in a major 2026 study report essentially no net employment effect from AI over the last three years.

That is not the same as "no jobs will change." This guide is part of our AI agents for business hub. It separates what has been measured from what is only predicted, and it was checked on 9 November 2026.

What executives report

The strongest source we found is NBER Working Paper 34836, issued in February 2026 and revised in October 2026. It says: "We survey nearly 9,000 CFOs, CEOs and senior executives at firms in the US, UK, Germany, Australia, Canada, Mexico and South Africa in Summer 2026 on AI adoption and its effects on jobs and productivity."

On jobs, its abstract says: "Third, executives report that AI has had essentially no net effect on firm employment over the last 3 years, and predict little net impact over the next 3 years."

Do not stop reading there. The next sentence says: "These small net effects, however, mask considerable job churn: more than 40% of firms anticipate positive or negative employment effects that roughly offset only in the aggregate."

In plain English, the total stays flat, but individual companies may add or cut. These are executives' reports and predictions across all AI use, not a count of jobs replaced by agents.

What workers report

The Census Bureau asked workers directly. Its August 2026 story says: "About 56% of U.S. workers said they have used Artificial Intelligence (AI) on the job for at least one of 11 tasks asked about on the U.S. Census Bureau's March 2026 Household Trends and Outlook Pulse Survey (HTOPS)."

Among workers who used AI in the last week, Census reports this self-reported time saving:

Those figures describe saved hours, not eliminated jobs. They also cover AI in general, and they are self-reported. Census's story does not report on job losses.

A separate Census story puts the national rate at 19.8% of businesses for the period ending 3 May 2026. That measures businesses, not workers, so it should not be compared directly with the 56% figure.

What the lab tests show

If agents were ready to replace whole jobs, you would expect strong results on realistic work. Tests so far show otherwise, with a large caveat about age.

The Remote Labor Index authors write: "AI agents perform near the floor on RLI, with the highest-performing agent achieving an automation rate of 2.5%." TheAgentCompany's authors say a good portion of simpler tasks could be solved autonomously. Harder, long-horizon tasks were out of reach of the systems they tested.

Both tests are from 2025, so the models are older than today's. Treat them as dated snapshots. Stanford's 2026 AI Index reports agents rising from 12% to about 66% task success on a different test, OSWorld.

Our guide on what AI agents can actually do goes further. Check the current picture on our reliability leaderboard before you make a decision.

What this means for staffing

Nobody can tell you what your headcount should be in three years. The data does support a few careful conclusions.

Plan by task, not by job. The Census list of top uses covers searching for information, writing, generating ideas, summarizing and administrative tasks. Those are parts of jobs, not whole jobs.

Count the review time. Someone has to check an agent's work. If you remove the person who knows what good looks like, you also remove your quality control.

Expect uneven change. The NBER study points to churn. Some firms in your market may cut roles while others add them, and the total will not tell you which you are.

Questions to ask before you cut a role

1. Which specific tasks would the agent take over?
2. What share of the job are those tasks?
3. Who reviews the work, and how long does that take?
4. What happens on a day when the agent is wrong or down?
5. Have you tested this on your real work?

Our comparison of an AI agent and hiring an employee walks through the numbers for one role. Use your own figures, and be wary of any vendor's headline savings.

A better starting question

Ask what your team would do with the time. An agent that takes over routine work may let staff handle more customers or tougher problems.

That is a planning choice, not a forecast. The NBER executives report that AI raised annual labor productivity growth by 0.2% over the last 3 years. They expect 0.6% per year over the next 3, and both figures are their own estimates.

Next step

List five repetitive tasks on your team and mark who reviews each one. Our guide on what to automate first can help you pick. For a custom build, see our AI agent development service.

Frequently asked questions