Yaroslav Kushta/Moment via Getty ImagesAs AI model capabilities ramp up exponentially, fears of economic disruption are growing, despite little evidence AI is causing mass job reduction just yet. The median takeaway across several predictions and studies is AI is getting better at certain work-related tasks, but the rate of meaningful adoption means you have more time to adjust than you might think.
Also: Why replacing staff with AI backfires – and 5 ways smart leaders generate real value instead
A recent survey from Resume Now shows AI isn’t competing in the workplace. For one, 97% of workers surveyed said they rely on their own judgment or another human’s when making decisions, and 72% said they would trust a coworker’s judgment over an AI suggestion if the two disagreed. Hopefully, findings like these mean the so-called “meat proxy” coworkers are in the minority, at least for now.
Expertise still matters
I recently moderated a panel discussion in New York on AI in higher education. After, I spoke with several attendees about their thoughts on AI’s current capabilities. An investment banker with a data science background remarked to me he could replace a significant portion of his team with Claude if he were allowed.
For most others, AI remains in a support role, leaving higher-level work like strategy to human employees. According to Resume Now, workers are still most comfortable using AI to speed up routine information-gathering tasks but pull back when a task requires “strategic thinking or judgment.”
Productivity gains have been a core selling point for AI at work – but by some metrics, AI isn’t meaningfully making teams more efficient, instead creating time-consuming “workslop.” Resume Now’s survey found pressure isn’t coercing most people into turning to AI for a quick fix: 93% of respondents said they “rely on themselves, coworkers, or managers when decisions are high-stakes and time-sensitive,” the company’s report said.
Almost a quarter of workers reported using AI for research, while 14% said they use it for writing emails and other messages, and 13% use it to summarize information. Numbers drop to 9% for those who tap AI for data analysis and just 3% for those looking for strategic or planning advice from AI.
Also: I gave Claude Cowork 7 non-coding jobs, and it earned a spot in my toolbox
That last stat especially helps make the case that human experts are still a more reliable default than even top AI models. Professional services firm Pearl runs an expertise-based leaderboard that ranks frontier model performance against human professionals; by its measure, an AI model has yet to stack up.
“Public benchmarks report 80–95% parity with human experts — but those datasets are embedded in training pipelines,” Pearl’s website states. “We scored leading models against a private, expert-authored dataset they’ve never seen. The gap between benchmark marketing and real-world expert alignment is significant, and it doesn’t close with newer models.”
Also: These companies are actually upskilling their workers for AI – here’s how they do it
As of September, Anthropic’s Claude Fable 5 holds the leaderboard’s top spot across business, tech, law, health, and more disciplines with an 80% score — what Pearl would call a B- student. Opus 5, Fable 5.1, and Perplexity’s Sonar Deep Research follow closely behind at just under 80%.
While there are plenty of studies assessing AI’s ability to complete economically relevant tasks, there is too much variance in definitions of success to create a broad standard for AI to compare with. Still, MIT predicted in April AI would be “minimally sufficient” on some of those tasks by 2029 — given that, 80% is keeping pace decently well.
Usage-based trust could turn the tide
A new Gallup poll published today found AI use is up from last year, but trust in the technology is trending downward, from 32% last year to 28% now. If that stat sounds familiar, that’s because it is: Most surveys about AI use have been revealing the same trend periodically for over a year. Data from June 2025 and June 2026 give very similar impressions — as AI use spreads, trust does not.
That said, usage creates more trust than not using AI at all, according to Gallup.
“AI use is the clearest differentiator of trust: 43% of Americans who use AI sometimes or more trust it, more than three times the rate among those who use it rarely or never (12%),” the report said.
Also: US workers are the world’s biggest AI skeptics – and it’s not just about job loss
Gallup saw an increase in adults who reported using AI “sometimes, often, or always at work, at school, or in their personal lives” from 39% in 2025 to 51% this year. The report noted that though this study was not specific to AI use at work, “the gradient and directionality” of this increase is similar to its workforce-specific findings from Q1 of this year, which found half of American adults reported using AI in their roles — up 10 points from Q2 of 2025.
Resume Now’s findings confirmed usage engenders trust: 64% of respondents said their trust in using AI at work had either stayed the same or increased in the last year. Read alongside Gallup’s findings, if adoption rates hold as AI gets more capable, that could mean AI gains more credibility in the workplace over time.
