Why AI Use at Work Feels Like Growth and Threat at Once

I was on a call last Tuesday with someone who builds AI features for a living, and she said something out loud that a lot of people are thinking quietly. She had spent the morning shipping faster than she ever had, with the coding agent doing the scaffolding and her doing the judgment. By lunch she felt sharp. By four in the afternoon she said it like a confession: “I cannot tell if I was promoted today or quietly erased.”

That is the feeling the research just named.

What the research shows

A new paper in Group & Organization Management, published online April 15 by Kanimozhi Narayanan and colleagues, ran a ten-day daily-diary study with 173 working adults in the United States across data analysts, data scientists, creative directors, operations heads, and HR managers in industries from finance to healthcare. Everyone used AI tools every workday, anywhere from thirty minutes to eight hours. Three check-ins per day. Two work-weeks.

The headline: on days with more AI-supported work, the same person appraised that work as a challenge AND as a hindrance, in parallel. The two appraisals were not endpoints on a spectrum. They were happening at the same time.

The challenge side fed task crafting and innovative behavior, the small acts of reshaping how a job gets done. The hindrance side fed job-replacement anxiety and a quieter kind of cynicism about the organization. The researchers also tested whether a generally positive attitude toward AI fixes this. It only mutes the hindrance side. The challenge appraisal happens regardless of how someone feels about AI in the abstract.

A separate perspective piece in Frontiers in Psychology one day later, by Xinyi Qi and colleagues, names the deeper version of this and calls it identity dissonance: publicly embracing AI as the language of progress while privately worrying that dependence on it weakens originality, authorship, or autonomy. Two different research traditions, the same shape from two angles.

This sits next to the AI productivity multiplier conversation, where the headline gains and the felt experience of the workday refuse to line up. It also sits next to recent work showing how ai reliance erodes self-belief, where heavier daily use slowly thins the felt sense of being capable. The diary study takes that shape and pins it to a clock.

What it doesn’t tell us yet

This is one diary study, 173 workers, one country, two weeks. Every measure was self-report. The researchers tested four outcomes: task crafting, innovative work behavior, job-replacement anxiety, and organizational cynicism. Job-replacement anxiety is the closest construct in the model to the felt sense of being a professional, and even that is identity-adjacent rather than identity proper. There is no follow-up at month three, no second sample in another country yet, no measure of whether the dual appraisals settle into one feeling or stay split over a longer span. Take it as the first careful look at a real pattern, not the settled story of what AI work does to the working self.

One thing to notice in your work today

Look at the last full AI-heavy workday. Today probably, or yesterday. Was there a moment of feeling sharper than usual, picked up by what the tool helped get done? Probably yes. Was there a moment in the same day when something small twisted in the chest at the thought that the work was no longer quite ours, or that the version of us who used to do this work was being scoped down? Probably also yes.

The research says those two feelings are not contradictory. They are co-occurring. The shape of your workday is still being figured out, and no one needs to pick one feeling over the other. The small steady move is to notice that both showed up, name them gently, and keep going.

That noticing is the practice for the week.