The Messy Middle: Data on AI and Your Job

44% of workers now say AI does more harm than good. 52% say they don't feel any need to develop new skills because of it. Those aren't fringe opinions from a niche survey. They're the headline findings from Jobs for the Future's 2026 survey of workers and learners, and together they describe a workforce that's souring on AI at the exact moment it's becoming harder to opt out of, let alone ignore.

This tension is palpable. There's a nuanced debate somewhere between "AI will take your job" panic and the "just learn AI, it's easy" cheerleading. The honest, unglamorous middle, where the data is genuinely mixed, is where the conversation should live. The winners and losers are being sorted right now, and where you land depends less on the technology and more on what you decide to do about it in the next twelve months.

Side-by-side stat callout: 44% of workers say AI does more harm than good, 52% don't feel they need to develop new skills for it, sourced to JFF's 2026 Survey of Workers and Learners

The Sentiment Flip Nobody Saw Coming

In 2024, more workers thought AI would do more good than harm. That flipped. Now it's 44% harm versus 38% good, according to JFF. Something happened in the space in just two years that caused many to turn their cautious optimism into open skepticism, and it isn't a mystery: workers were told to adapt, then largely left to figure out how on their own.

This contention comes directly from the data. 56% of workers say they've never been consulted about how AI gets used at their own job. Preparedness to use AI, self-reported by workers, even declined since 2024. And the primary way people are learning these tools isn't through employer training or formal education. It's YouTube. It's social media. It's trial and error on their own time (and potentially even their own dime).

Employers asked for adaptation and didn't show up with the training budget to make it possible. That gap is where at least some of this wave of AI resentment is coming from, and it's a real, structural failure, not a workforce that's simply afraid of change.

Some Companies Are Growing With AI. Most Aren't.

Here's where the debate gets genuinely messy, and where a single headline stat can mislead you if you don't read past it.

TechCrunch reported on new research from Ramp and Revelio Labs showing that companies that spend heavily on AI, what the researchers call "high-intensity adopters," grew headcount 10.2%. Entry-level hiring at those same companies also rose 12%, directly contradicting the popular narrative that AI is gutting junior roles first.

That'd be great news if true, right? Well, set that next to Goldman Sachs' separate finding that AI erased roughly 16,000 net jobs a month over the past year, concentrated hardest at the entry level.

Both of these data points are true. They're just true about different companies.

The Ramp/Revelio sample skews toward tech-forward, VC-backed firms that were probably already growing. It's genuinely unclear whether AI caused the growth or just showed up at companies that were going to grow anyway. The researchers themselves are careful about this: their paper "does not show that AI universally creates jobs," but it does push back on the idea that AI broadly destroys them.

What's actually happening is a bifurcation. Companies with capital, technical talent, and the management bandwidth to actually implement AI well are leaning into it as either an R&D play or a pursuit of growth and expansion. Companies that bought a few subscriptions, ran a pilot, and called it a day are seeing nothing. No growth, no efficiency gains, nothing to show for the spend.

That's not a story about AI. That's a story about execution. It always has been.

Split callout: Ramp/Revelio Labs data showing +10.2% headcount growth and +12% entry-level hiring at high-intensity AI adopters, next to Goldman Sachs data showing -16,000 net jobs lost per month concentrated at entry level, captioned "Both are true. They're just true about different companies."

The Generation Caught in the Middle

To see where this bifurcation is landing hardest, look at the age gaps. That's where there's a real divide. For example, JFF found that 44% of workers aged 16 to 34 are actively reconsidering their career path because of AI. For workers 55 and older, that number is 4%. To simply call that a "gap" feels like an undersell. This data tracks with what Pew Research found this year too: adults under 30 are more likely than older adults to say AI will have a negative effect on society and on them personally.

Pew Research chart titled "Younger adults are more wary of AI's potential impact on society and on them than older groups," showing % who say AI's impact over the next 20 years will be negative, broken out by age group 18-29, 30-49, and 50+, for both society and personal impact

This isn't irrational. Entry-level roles are exactly the roles most exposed to the kind of automation AI is good at right now: drafting, summarizing, first-pass research, junior coding tasks, data entry, and the like. The generation entering the workforce is watching the bottom rung get automated out from under them while hearing, simultaneously, that they need to "just learn AI" to compete. It's a lot to ask of anyone starting out.

And it isn't happening in a vacuum. The unease is showing up in how younger workers view AI's infrastructure, not just its output and societal implications. Even OpenAI's own April 2026 policy blueprint, Industrial Policy for the Intelligence Age, proposed robot taxes and a public wealth fund modeled on Alaska's oil dividend. On the surface, this acknowledgment from inside the industry, an explicit understanding that their business goals have global ramifications, seems like a positive. If the current path of AI is leading toward a concentration of gains in a way that forgets many, then of course we need a fix. OpenAI surely has an incentive to take a stand in this direction; they have a public perception they need to ensure remains favorable enough for new subscribers to join their platform as paying users.

Either way, the Overton window shifting toward a discussion of public wealth funds and automation taxes as necessary correctives tells you the anxiety doesn't just exist on the fringes.

Interestingly, the data doesn't split cleanly along "young people hate this" lines either. A Redfin/Ipsos survey found Gen Z (48%) and Millennials (50%) are actually more supportive of AI data centers in their own neighborhoods than Gen X (38%) or Boomers (22%), largely because they see the job creation angle. Whether or not the job creation idea becomes reality, the perception across generations isn't "young people are against AI." It's closer to: young people are more exposed to AI's downside at work, more skeptical of who's capturing the upside, and still pragmatic enough to want the jobs that come with the infrastructure. That's not a contradiction, that's the nuance headlines rarely hit.

Bar chart titled "Who Supports an AI Data Center in Their Own Neighborhood?" showing % support by generation: Gen Z 48%, Millennials 50%, Gen X 38%, Boomers 22%, sourced to Redfin/Ipsos survey fielded Nov. 2025

"Just Learn AI" Has a "Just Learn to Code" Problem

I'll say the uncomfortable part directly: I don't think "just learn AI" is good advice on its own, and I'm suspicious of anyone selling it as a clean fix.

George Carlin had a bit about how the people at the top will always find a way to move the finish line right as you're about to reach it. Every generation gets handed a new version of the same instruction: adapt or fall behind. It's phrased like sound advice when it's really just the newest required cost of staying employed.

"Just learn to code" was that instruction for the last decade, and it aged badly, not because coding wasn't valuable, but because it treated a complex, unevenly distributed opportunity as a simple individual choice. It ignored access to training, time, money, and the fact that not every job or person benefits from the same intervention. AI adoption has the exact same trap built into it. It doesn't make equal sense for every role. A plumber doesn't need to learn prompt engineering to do good work, but a bookkeeper probably should. The nuance matters.

But here's where I land anyway: for the people and businesses where AI adoption does apply, there has never been a lower-cost, lower-barrier moment to actually do it. Setting aside maybe the 2017 through 2021 zero-interest-rate window, when capital was so cheap that starting almost anything felt easier than it should have, this might be the best time in a generation to build something or level up your own capability. A $20/month subscription and a stack of free YouTube videos will teach you more, faster, than almost any formal program could have five years ago. The floor for getting started has never been lower.

That's not to sugarcoat the 44% or dismiss the 52%. It's suggesting that the 52% who don't feel the need to upskill are making a mistake, given where the incentives are pointed. Companies, especially the VC-backed and shareholder-driven ones, are under constant pressure to do more with fewer people. That rings true for the 10 or so companies propping up what feels like the entirety of the US stock market. That pressure isn't going away, and betting your career on it not affecting you is a bet against the direction every incentive in the system is currently pointing.

What This Looks Like When You Actually Do It

Different people come to this line of thinking and AI adoption in different ways.

My Laveur co-founder was formally educated on how to apply AI to business transformation as part of his graduate work. My path was more self-taught, letting curiosity lead my learning path. I'd always been poking at these tools because I wanted to understand what they could actually do and how they could make my own work, as well as the work of those around me, easier. Both paths led to the same place. Neither required the other's starting point.

We've also both worked inside teams where this wasn't a mandate so much as a nudge. Companies quietly (or sometimes loudly) signaled that AI-native workflows were becoming table stakes without ever fully spelling out how to get there. That's most companies right now. The direction is clear. The onboarding and execution isn't.

We've already proven our abilities within enterprise-level organizations, launching and scaling products vital to daily operations and in pursuit of strategic business goals. When we set out to prove that we could build with these tools outside of the comfort of a full-time job, and do it well, not just quickly, we built three products for ourselves: USA CPG, Reel In, and Local Archives. Different problems, different audiences, same underlying playbook on how to execute. You can read more about how that actually worked in Three Products, One Playbook. The point of building them wasn't necessarily the products themselves. It was proving, to ourselves first and to the public more generally, that the tools and the approach actually hold up outside a tutorial.

Where This Leaves You

The data isn't going to resolve itself into a clean, reassuring story. It shouldn't. Some companies are growing because of AI. Most aren't seeing much of anything yet. Younger workers are right to be uneasy about who's capturing the gains, and right to keep showing up anyway. The people getting left behind aren't getting left behind because AI is inherently a threat. They're getting left behind because nobody handed them a way in, and they didn't go looking for one themselves.

That second part is the piece you actually control. It's not too late. It's not even close to too late. But "not too late" isn't the same as "no urgency." The window to be early is still open. It won't stay that way forever.

Let's talk