A major new workforce study has uncovered something that should give every business leader pause before they hand their team another AI tool and expect productivity to soar.

ADP Research’s People at Work 2026 report, drawn from more than 39,000 workers across 36 markets, found that half the global workforce now uses AI at least multiple times a week. That number signals real, ground-level adoption — not just pilot programs and executive announcements. But buried inside the same data is a finding that complicates the story entirely.

Daily AI users are four times more likely than non-users to say they feel less productive than they could be.

Read that again. The people using AI the most, every single day, are the ones most likely to feel like they are underperforming.

What the Data Actually Shows

The paradox gets stranger when you look at the full picture. Daily AI users were measurably better off on almost every dimension the survey measured:

• Engagement: 30% of daily AI users were fully engaged at work, compared to just 14% of those who never use AI
• Stress: Only 11% of daily users reported negative work stress, versus 23% of non-users
• Team relationships: Daily AI users felt better about their colleagues and working environment

By every indicator of workplace health, AI power users are doing better. They are more switched on, less burnt out, and more connected to their teams.

And yet they feel less productive than they could be.

Why This Is Happening

There are a few credible explanations, and none of them suggest AI is failing.

The first is the nature of what remains. When AI handles the routine checklist work — data entry, first-draft emails, scheduling, research summaries — what is left for the human is the harder, harder-to-measure stuff. Strategic thinking. Judgment calls. Creative problem-solving. These tasks do not produce the same visible throughput as knocking off a to-do list. Workers used to measuring themselves by tasks completed find themselves working on problems where “done” is ambiguous.

The second is the attribution effect. If an AI drafts your report, summarizes your meeting notes, and answers your routine inbox, how much of your day’s output do you actually credit to yourself? There is a real psychological gap between productivity achieved and productivity felt when a large chunk of the visible work is being done by a tool.

The third is expectation inflation. The promise of AI often gets framed as doing more, faster. When workers adopt AI but still feel constrained by meetings, approvals, strategy bottlenecks, and human dependencies, the gap between what the tool enables and what the system allows can feel frustrating rather than freeing.

What This Means for Your Business

If you are rolling out AI tools across your team and expecting a clean line between adoption and output, this research should recalibrate that expectation.

The productivity gains from AI are real, but they are not always immediate, not always visible in the metrics you are used to measuring, and not always felt by the people doing the work. That matters because perception drives behavior. A team that feels less productive despite using AI is a team at risk of abandoning the tools, reverting to old workflows, or simply disengaging from the whole initiative.

There are three things businesses can do with this:

First, change what you measure. If AI is handling the volume work, the human contribution shifts to quality and strategy. Measuring output in units per hour will miss the real value being created. You need metrics that capture decision quality, strategic output, and work that was simply not possible before.

Second, invest in training, not just tools. The engagement gap between AI users and non-users is massive. But the productivity paradox suggests that giving people access to AI is not the same as helping them integrate it well. Workers who understand how to genuinely hand off routine work — and how to redirect that freed-up cognitive capacity into higher-leverage tasks — will get more from the tools. That requires training, not just licenses.

Third, address the attribution problem directly. Teams need frameworks for recognizing human-plus-AI output as legitimate achievement. If the culture still rewards people for doing things the hard way, AI adoption will always feel like a compromise.

What This Means for Business Leaders

The half of the global workforce now using AI weekly is an extraordinary development that would have seemed implausible five years ago. The productivity paradox does not undo that progress. It contextualizes it.

AI adoption is creating a workforce that is more engaged, less stressed, and more capable — but also one that is recalibrating how it thinks about work, output, and contribution. That is a leadership challenge as much as a technology one.

The businesses that get the most from AI will be the ones that take the human side of the transition as seriously as the technical side. That means acknowledging the adjustment, updating how performance is measured, and giving people the context and skills to feel genuinely capable with the tools they are using — not just technically proficient with them.

The data says AI is working. The feeling says the integration is still a work in progress. Both things are true, and managing that gap is where real competitive advantage lives right now.

---

If you want the playbook other teams are using with Claude and Codex right now, grab the free Working With Claude field guide. Download it here.