AI at work: adoption and scepticism

Artificial intelligence tools are spreading through offices faster than routine trust in them. (wral.com) A Gallup Workforce survey of more than 22,000 United States workers, published in January 2026, found 12% of employed adults use artificial intelligence daily on the job, about one-quarter use it at least a few times a week, and nearly half use it at least a few times a year. (wral.com) Gallup said 21% of workers were using artificial intelligence at least occasionally in 2023, and its April 13, 2026 workplace update said half of United States workers now use artificial intelligence, even as adoption still depends on manager support, workflow fit, and whether workers see value in the tools. (gallup.com) Use is uneven across jobs. The Associated Press report on the Gallup data said about 6 in 10 technology workers use artificial intelligence frequently and about 3 in 10 use it daily, while majorities in professional services and education say they use it at least a few times a year. (wral.com) The split inside companies is not just about interest. Microsoft and LinkedIn said in their May 8, 2024 Work Trend Index that 75% of knowledge workers globally were already using artificial intelligence at work, 78% of users were bringing their own tools to work, and 60% of leaders said their company lacked a vision and plan to implement it. (microsoft.com) That leaves many organizations with workers experimenting on one side and management systems lagging on the other. Microsoft said 79% of leaders viewed artificial intelligence adoption as critical to staying competitive, but 59% worried about measuring the productivity gains. (microsoft.com) The governance problem gets sharper with artificial intelligence agents, which are software systems that can take actions across apps instead of just answering questions. Okta said 91% of organizations are already using artificial intelligence agents, 80% have experienced unintended agent behavior, 44% have no governance in place, and 23% report credential exposure through agents. (okta.com) Okta’s pitch is that companies now need to know where their agents are, what they can access, and what they can do, and to assign every agent a human owner with tightly limited permissions. That framing treats agents less like a single tool and more like a new class of worker inside the company network. (okta.com) The result is a workplace where adoption can rise without becoming routine, and where routine use can spread before rules, oversight, and measurement catch up. Gallup’s latest updates describe that gap in practical terms: workers use these systems more when managers support them and when the tools fit the job in front of them. (gallup.com) That is why the next phase of workplace artificial intelligence looks less like a race to install chatbots and more like a slower effort to test them, limit them, and decide where they belong. (microsoft.com, okta.com)