AI is coming into the jobs, not for them

getty

Ask almost any futurist how AI would reshape work, and you’d hear some variation of the same apocalyptic vision: millions of knowledge workers made redundant overnight, HR departments gutted, and CEOs delivering earnings calls beside humanoid assistants. The hype cycle has careened straight into doomsday-prepper territory where the voices are loud, alarmist, and largely missing the plot.

You see, AI coming for the jobs is not what’s happening.

It’s not coming for the jobs as much as it’s coming into the jobs, showing up as a requirement in a shift so pervasive that it’s now redefining what it means to be a skilled worker.

And the data backs it up.

According to DHI Group’s latest numbers from its tech hiring arm, Dice, job postings that list AI agent skills have grown over 2000% since last year. “We’re seeing AI become part of everything, from healthcare to consulting to aerospace,” says Art Zeile, CEO of DHI Group.

“There’s no single ‘AI job’ anymore just like there aren’t many jobs that don’t have an expectation for you to work with AI.”

What we’re seeing isn’t automation-induced obsolescence. This is AIfication of the workforce.

As Zeile puts it: “The most profound change we’re seeing isn’t that jobs are being cut left and right. It’s in how jobs are changing shape.”

Only a few months ago, we were writing obituaries for customer support teams at Klarna after the CEO confidently declared AI had replaced them, only to reverse course. Humans are back because AI didn’t eliminate the need for them, it just changed the texture of the work they do.

If you’re feeling whiplash, not sure what to believe or where the story’s going, let me offer a historical compass that might give you a hand.

When every worker works with AI

As with most things in history, we’ve been here before.

In the 19th century, the global ice trade employed tens of thousands of men harvesting massive blocks of frozen water from northern lakes, insulating them in sawdust, and shipping them to tropical cities where refrigeration hadn’t yet arrived. It was brutal, dangerous work, but it created enormous value with significant portions of local economies invested in the trade. Then came mechanical refrigeration, and within a few years the ice-harvest industry collapsed.

When the machines came to town, everyone knew the jig was up. You could see it in the still icehouses, the idle horses, the silence where there used to be saws. And you wouldn’t have been able to convince anyone otherwise, because the end felt just as obvious as it was.

We can see the exact same shift happening with AI. What most don’t see, yet, is where it all leads.

Here’s what happened with the ice trade. The tasks involved in its original form largely disappeared. What didn’t go away was the demand for cold or people working in service of it. Instead, it exploded. As refrigeration got cheaper and more scalable, entirely new industries emerged from cold chains to everything from your ice cream pint to mRNA stability. The trade itself, the act of ice harvesting, vanished, but the work of keeping things cold multiplied a hundredfold.

That’s exactly what we’re seeing with AI.

Yes, the rote tasks are evaporating and in the mix, many of us will be unemployed. Yes, the job descriptions are mutating, making it harder for people to climb back into employment without retraining themselves, sometimes entirely. But the value center of human work is growing. We’re not witnessing a collapse of work where AI wipes us out, we’re witnessing an absorption event where AI becomes embedded across functions, augmenting the best workers, expanding the scope of what a team can deliver, and creating entirely new categories of economic activity that didn’t exist five years ago.

Our robot overlords aren’t here to kill the job. They are here to redefine the unit of work itself.

This is the shift you’re already seeing the smarter companies rewiring themselves around.

Welcome to the future of work with AI

A few years ago, BCG and Harvard Business School ran a study on how humans interact with AI at work. The researchers coined new archetypes for how workers operate in tandem with intelligent systems: centaurs, who divide tasks between human and machine and cyborgs, who blur the boundary entirely.

Since then, things have moved quickly. We’ve welcomed autonomous agents into our workflows, watched LLMs write everything from strategy decks to source code, and seen a cultural backlash so sharp it gave us new workplace norms, and new archetypes. Now we need to add the Orchestrator, who choreographs multiple AI tools across a system, and the Artisan, who deliberately minimizes AI use to preserve creative edge. Give it a few years and we’ll have proud teetotalers abstaining from the vices of AI entirely, charging a premium for their entirely human outputs.

Two years ago, these categories felt like speculative fiction but they’re becoming standard job templates today.

“We’re seeing the archetypes go from academic concept to operational reality,” says Art Zeile, CEO of DHI Group, parent company of Dice. “In our data, more job descriptions are asking not just for technical skills, but for the ability to direct and collaborate with AI tools, whether that means coding with copilots or managing multi-agent workflows. We see it in the hiring criteria directly.”

While the ice trade took years to wind down, what we’re seeing here is working its way through our economy at lightspeed, and the jagged edge is not stopping at the bleeding edge of tech companies.

“I’m talking to clients across industries, from manufacturing to media, and they’re all reworking job requirements to include AI competency,” says Mahe Bayireddi, CEO of Phenom, an AI-driven HR platform that supports some of the world’s largest employers. “There’s little sense in creating ‘AI roles’ given how AI is becoming part of every role.”

Employers aren’t just hiring AI specialists anymore, just like no one hired digital economy specialists after everything became natively digital. Instead, employers are expecting frontline workers to use AI-enhanced tools, customer service reps to route inquiries with LLMs, and marketing teams to co-create campaigns with generative tools.

That demand for hybrid fluency is pushing hiring managers to rethink what they ask for, as well as how they attract candidates in the first place.

“Gen Z, especially, isn’t interested in jobs where they feel like a cog,” Bayireddi says. “They’ve grown up with AI in their hands and they want to use it in meaningful ways. They expect their tools to be smart, and their workplaces to be smarter.”

And boy are they adapting fast.

“Honestly? Gen Z is lucky,” he adds. “They’ve been interfacing with ChatGPT for years. That’s the world they know. Adoption isn’t the problem for them, it’s the old infrastructure of work that’s creaking under the weight here.”

Which brings us to a simple but crucial mental shift underlying the trend.

For all the headlines and hand-wringing, we’ve spent too much time mythologizing AI and treating it like either the apocalypse or the second coming. That framing does little to help us navigate what’s actually happening on the ground.

Instead of bracing for an abstract future, we’d be better off grounding our thinking in something much more practical: AI is a tool. An incredibly powerful one, yes, but still a tool. And if we can understand how it functions across four core “tool” dimensions.

Thinking of AI as a tool

Part of the confusion comes from how we talk about AI. But AI, like every transformative technology before it, is ultimately just a tool. And tools do four things: they help us do more work, better work, they can do the work themselves, or they open up venues for entirely new kinds of work for us.

AI happens to do all four, which is like giving every worker a jetpack instead of a power drill.

“This is why job descriptions are evolving so rapidly,” Zeile notes. “ What’s clear is that companies aren’t posting for one kind of AI job, they’re embedding AI tools into roles that already exist and these jobs are transforming across a number of dimensions.”

Of course, jetpacks come with new expectations. Employees are being asked to use these tools wisely, effectively, and, most urgently, with originality.

Which brings us back to the humans.

The limits of our AI nativity

“AI has raised the minimum bar on acceptable quality,” says Max Spero, co-founder of Pangram, an AI detection company working with universities, publishers, and hiring platforms. “If you’re not using AI, you’re probably not competing. But if you’re only using AI, your output looks like everyone else’s.”

Pangram builds detection systems that can tell when a piece of writing was generated by an LLM. Their clients, like Quora and top universities, don’t want to ban AI. They want to identify how it’s used, and whether it’s adding value. More than that, they want originality and creativity at the center of the humans who use AI.

“AI has this tendency to collapse into the most likely answer,” Spero explains. “We call that mode collapse. You lose the weirdness. You lose the originality, and you lose the human who made it all come together.”

The implication is clear as day. When AI can write, code, and even reason better than any of us, the real differentiator is no longer whether you can perform a task, but whether you can perform it in a way that matters.

“The strongest students, the ones who want to learn, they’re learning five times as much using AI,” Spero says. “The problem isn’t that students are using AI, it never has been. The problem is that our systems aren’t ready for what AI-empowered learners can do, and they’re not being coached into using them effectively.”

In other words, it’s not enough to build centaurs and cyborgs, nor can we stop at building up orchestrators or artisans either. We need everyone in the workforce to use their tools with precision as well as originality, and the urgency isn’t limited to Gen Z.

“We can’t fall into the trap of thinking the solution is banning AI entirely,” Spero warns. “We need to make AI work for humans as much as with them.”

That means building guardrails not just for AI itself, but for the ways we deploy it at work. A recent study by Microsoft and OpenAI offers a clear warning: workers who relied heavily on AI for tasks involving judgment or strategy were more likely to accept incorrect answers and showed reduced critical thinking over time. In some cases, performance degraded despite having access to better tools. We can call this automation complacency as well as just pure optimization, where users defer to AI even when they shouldn’t.

Get ready to tinker and learn

AI can raise the floor for everyone, but if we’re not careful, it might lower the ceiling too.

That ceiling won’t come down because the tools aren’t capable, mind you, it’ll come down because we aren’t ready to use them.

Kai Roemmelt, CEO of Udacity, doesn’t mince words on the scale of the gap at play here.

“We’ve reached the point where 61% of professionals believe AI could replace their role in the next three to five years,” he says. “And yet only 5% of organizations are providing AI training at scale. That’s more of a chasm than a gap.”

Roemmelt’s point isn’t that universities or corporate learning teams are asleep at the wheel. It’s that the traditional cadence of curriculum design, semester-long delivery, and accreditation was built for a slower age. “By the time a university gets an AI course approved, the technology it’s teaching may already be two product cycles out of date,” he says. “We need training that keeps pace with the tools, and that means practical, hands-on, and grounded in real workflows.”

Udacity is doing its part with its new Agentic AI and AI for Business Leaders Nanodegree programs designed less like lectures and more like field exercises. That emphasis on “learning by shipping” mirrors the way AI itself is evolving—iterative, experimental, and constantly in production.

“It’s coming into my job too,” he says. “If I’m not playing with the tools, if I’m not seeing what they can actually do in context, I can’t lead a team that’s supposed to be using them.” Roemmelt’s conviction is that no amount of slide decks or theory will prepare a workforce for agentic AI, the only way through is to get your hands dirty.

The irony is that AI’s spread into every job is creating a situation where even non-technical roles need a level of technical fluency. Roemmelt sees this as the next great corporate literacy. “Many companies are still focusing on upskilling just their tech teams. That’s a mistake,” he says.

“If AI is embedded in every role, then everyone, from marketing to operations, needs to be able to orchestrate it.” And if they can’t, well we’ll just see how fast that ceiling comes down.