Half of US Workers Call AI Productivity Gains Only 'Somewhat Positive'

What happened

A new Gallup survey finds that around half of US employees describe the productivity gains they have experienced from workplace artificial intelligence tools as only "somewhat positive." According to Gallup's findings, reported by No Jitter, the strongest productivity ratings were concentrated among more technical or task-specific uses of AI, such as coding or building presentations, rather than broader, open-ended applications.

The data suggests a split in how workers experience AI on the job: clear, bounded tasks with a defined output appear to generate more confident positive assessments than general-purpose or ambiguous use cases. Gallup's framing points to a workforce that is engaged with AI tools but still tentative about their overall value.

Why it matters

For organisations investing in AI to improve service delivery, this is a signal worth taking seriously: adoption is not the same as conviction. Employees are the frontline interpreters of any AI-enabled process, and if their own experience of these tools is lukewarm, that ambivalence is likely to surface in how confidently — or cautiously — they use AI when serving customers.

From a behavioural-economics standpoint, the pattern also reflects a familiar dynamic: people trust and value tools more when the task and the payoff are legible. Coding and presentation-building have a visible before-and-after; the AI either produced the artefact or it didn't. Broader, fuzzier applications leave more room for doubt about whether AI actually helped, which dampens perceived value even where genuine time savings exist.

The Renascence take

The headline finding — that only half of workers see AI gains as more than "somewhat positive" — will likely be read as a story about AI underdelivering. We think that's the wrong lesson.

What this data really exposes is a scoping problem, not a technology problem. Employees rate AI highest where the task is narrow, the output is checkable and the win is obvious — exactly the conditions that make any tool, human or digital, easy to trust. Service and CX leaders rolling out AI internally should resist the temptation to launch broad, general-purpose "AI assistants" and instead sequence deployment around discrete, verifiable tasks first, letting employees build confidence through small, legible wins before extending AI into judgement-heavy or ambiguous parts of the workflow. Trust in AI, like trust in any new process, is earned task by task — not declared by rollout.