JobMarketHealth
Research / AI & Work
AI & Work / Evidence through October 4, 2026

AI is spreading at work. What should you do about it?

About one in four U.S. employer businesses reported using AI in the Census period covered here. Job ads mention it more often too. For workers, the useful next step is to understand where it fits into their own role.

Census survey through October 4, 2026 · Indeed postings through September 30, 2026

Published November 14, 2026 · By JobMarketHealth

Start with the work you do

Our read: get familiar with how AI could help with tasks in your role, and check what employers actually expect. Growing adoption is a reason to pay attention. These figures do not measure jobs lost or show that everyone needs to become an AI specialist.

What businesses report

About one in four used AI

Share of U.S. employer businesses reporting use for any business function during the survey's two-week period.

November 16, 202517.3%
October 4, 202624.6%
Census BTOS. Dates are the ends of the two-week periods. Survey estimates; comparison uncertainty is explained in the source notes.
What job ads mention

AI appears in more postings

Share of U.S. Indeed postings mentioning AI, measured separately from the business survey.

October 31, 2025About 4.1%
September 30, 20267.3%
Indeed Hiring Lab, seven-day trailing averages. An AI mention can describe a company or its work; it is not necessarily a required skill.

The two panels measure different groups: businesses and job postings. Compare each with its own earlier value. Explore the underlying AI evidence.

What has changed

AI is becoming part of more workplaces

In the Census periods shown here, reported business use rose from 17.3% to 24.6%. That moves the estimate from roughly one in six employer businesses to roughly one in four. The question covers any business function, so a company can count as a user even if AI plays a limited role.

For an employee, the practical question is how that use reaches their own work. A business might use AI in one team while another team barely encounters it. Ask which tasks are changing and what people are expected to check, decide or deliver.

Read the requirement behind the AI mention

AI appeared in 7.3% of U.S. Indeed postings on September 30, compared with about 4.1% at the end of October 2025. That makes the wording in a job ad worth reading closely.

For example, an ad describing an AI product, an ad asking someone to use an AI tool and an ad hiring someone to build AI systems describe different expectations. Those are examples of distinctions to look for, not categories counted by this report.

Put it to use

Three useful steps for your own work

  1. Look for a specific task.

    Choose a small part of your work you can evaluate: drafting a first version, organizing notes or explaining unfamiliar code. Start with a task where you know what a good result looks like.

  2. Check the result as well as the time saved.

    Using a tool approved for your work and material you are allowed to share, compare its output with your usual approach. Include the time spent correcting errors. Keep a useful example if it improves the result.

  3. Connect the skill to a real expectation.

    Read postings for roles you want, or ask your manager where AI is being used. Be ready to explain the task, your judgment and the result. A concrete example gives you more to discuss than a broad claim that you “know AI.”

These are practical suggestions from JobMarketHealth, not outcomes tested by the Census or Indeed data.

A small experiment you can learn from

Illustrative example: a marketing role

Take a public product description and ask an approved tool for a first draft of a short campaign brief. Check whether it preserves the facts, suits the audience and says anything you would need to remove. Compare the total drafting and editing time with your usual process.

The useful result is an honest account of what worked and what still needed your judgment. If the tool adds work or errors, that is useful to learn too.

What if my occupation has high AI exposure?

Exposure research asks which tasks AI could potentially assist with. Use it to identify parts of a job to investigate. It does not tell you how many employers have adopted the tools or how many jobs have disappeared.

Before changing career plans, bring that task information together with actual employer requirements and the occupation's labor-market conditions. This report's adoption and posting figures cannot establish a link between AI and a change in employment.

Sources and how to read them

Survey uncertainty, question wording and coverage

BTOS is an experimental survey. We have not assessed the statistical significance of the change between the two estimates shown. Census changed the question's terminology for later collections; this edition retains the earlier series while the newer wording's comparability is reviewed.

The Census percentages describe businesses, not the share of workers using AI. Indeed measures language on one posting platform, not all vacancies, required skills or completed hires. Its figures are rounded here. The two sources do not identify the same businesses or workers, and neither comparison measures job losses.

Technical exposure estimates potential task assistance. Observed use, employer adoption and employment outcomes are separate measures and should be interpreted separately.

Evidence through October 4, 2026 · Edited for clarity November 14, 2026. The evidence periods are unchanged; this edition does not include later observations.