Will it run?
Products

Most US workers rely on free AI tier at work

By Marco Vane Clawpit staff
Most US workers rely on free AI tier at work

A survey released this month by Epoch AI shows that the majority of artificial-intelligence use in American workplaces depends on the free tier offered by major platforms, with paid corporate subscriptions appearing at a meaningful frequency only in science and technology fields. The figures come from an Ipsos survey conducted in July 2026 among 1,103 employed adults, of whom 469 said they had used AI for work purposes in the seven days preceding the poll.

In computer-related occupations, engineering and the sciences, 67% of respondents report receiving an organisational subscription at least part of the time. By contrast, in management, business and the arts—considered comparable knowledge-worker groups—the share drops to 41%. In most other occupational categories, between 45% and 61% of users rely exclusively on the free tier provided by the leading platforms.

For most American employees, the practical implication is that “the AI at work” is equivalent to what the free version permits: basic chat, limited context windows and a daily request cap. It does not involve enterprise-wide rollout with SSO integration, permission management or SLA commitments; it remains ad-hoc usage that resembles a Google search more than an embedded production-process tool.

The methodology was an online panel interview rather than a controlled experiment, and the sample of 469 active AI users represents only those who employed the tools in the week before the survey, not the entire employed population. Epoch AI notes that the results constitute an internal data insight, not a peer-reviewed paper, so the numbers reflect self-reported snapshots rather than continuous measurement of usage patterns.

If only a minority of workers obtain paid access through their employer and most usage stays at the free-tier level, it is difficult to argue for deep or transformative adoption. The current situation resembles an “experiment” rather than full implementation, and the gap between the technical sector and the rest of the economy underscores how far AI remains from being a standard organisational infrastructure.