AI job replacement: Is the apocalypse coming for IT work?

Dario Amodei, CEO of AI power Anthropic, recently warned–or was he boasting?– that artificial intelligence could eliminate up to 50% of entry-level white-collar jobsOpens a new window in the next five years. This, he predicted, would drive U.S. unemployment to between 10% and 20%, a dramatic increase from current levels.

If he’s right, that would kick the unemployment rate from May 2025’s 4.2%Opens a new window to a minimum of 14.2%. The last time we saw rates that bad was when COVID held sway in April 2020. Before that, you have to go all the way back to the Great Depression.

Ow!

Maybe you should start working on your plumbing license instead of a CompTIA A+ or Cisco CCNA. Is it really going to be that bad?

Pressure on entry-level coders

In his annual shareholders’ letter, Andy Jassey, Amazon’s CEO, wrote that “early AI workloads [are] being deployed to focus on productivity and cost avoidance.” What that means, according to Amazon engineers who spoke to the New York TimesOpens a new window , is “The company had raised output goals and had become less forgiving about deadlinesOpens a new window . … One Amazon engineer said his team was roughly half the size it had been last year, but it was expected to produce roughly the same amount of code by using A.I.” They also feel less like programmers and more like assembly-line workers. That’s not good.

Of course, Sam Altman, OpenAI’s CEO, saysOpens a new window “he’s less interested in replacing coders (100% automation) — and more focused on making them 10x more productive.” But that’s just marketing talk.

Still, AI will make a difference to tech jobs. In a study of developers at MicrosoftOpens a new window , Accenture, and an anonymous Fortune 100 company,  programmers given access to an AI-based coding assistant, GitHub CopilotOpens a new window , completed 26% more tasks on average than those without AI assistanceOpens a new window . Less experienced developers saw even higher productivity gains, with output increases between 27% and 39%, while senior developers saw smaller but still significant improvements (8%–13%). In short, AI may not be able to replace developers, but it can help them.

The same is true for other tech jobs such as system administration. SysAdmins should be used to this now. We’ve come a long way from writing Bash scripts by hand to DevOps tools such as Ansible and Puppet to Infrastructure as Code (IaC) programs like OpenTofu and Terraform. The prospect of, according to ServiceNow WorkflowOpens a new window , seeing 39.7% of their work being augmented by AI isn’t that scary.

AI job replacement will hit more than just tech

Sure, AI is already replacing some entry-level white-collar rolesOpens a new window , particularly in fields like tech, finance, law, and consulting. Recent data shows rising unemployment among recent college graduates, especially in technical fields. As Oxford Economics, a labor research firm, recently notedOpens a new window , many computer science graduates are having fits finding jobs. In particular, they found “There are signs that entry-level positions are being displaced by artificial intelligence at higher rates.”

Mind you, we’ve already seen this coming. Language-learning app company Duolingo has already announced that the “AI-first” company would be letting its freelance workers goOpens a new window . Around the same time, Shopify CEO Tobias Lütke told his executives that they must prove that AI can do particular jobs before asking for more workers.

These AI worker replacement plans don’t always work out. Just ask Klarna, the Swedish retail pricing comparison company, which replaced customer service agents with AI chatbots, only to realize that humans did a better jobOpens a new window .

Still, the experts agree. AI job cuts are coming. But what will the post-AI work world look like? The honest answer is “We don’t know.” As the Brookings Institution observed, “AI’s implications have been stuck at the extremes.Opens a new window On one end, techno-optimists champion a world of abundance and unlimited possibility, of drudgery-busting AI assistants in our pockets, AI-powered scientistsOpens a new window curing cancer, and turbocharged productivity creating prosperity for all. On the other extreme lie sweeping predictions of doom, mass job loss, and the end of human employment-or even existence-as we know it.”

Therefore, according to Brookings, “Though confident prediction is not possible, what is clear is that the design and deployment of generative AI technologies are moving far faster than our collective response to understand and shape them.”

I don’t know about you, but I hate uncertainty, especially when it comes to my job.

Replacing entry-level workers with AI has a cost

Others are sure, however, that AI will replace starter jobs sooner rather than later. Business consulting firm KeyniOpens a new window founder Jared Navarre agrees with Amodei, “The figure, though it seems jarring, appears to be fairly accurate with any of the models we’ve run. It’ll be more of a slow erosion than an overnight tsunami. Any company that has a high percentage of its labor dollars going to those tasks is salivating at the chance to replace those functions and tasks with cheaper AI solutions.”

They shouldn’t get too pleased about this prospect, though. Justina Raskauskiene, Human Resource Team Lead at OmnisendOpens a new window , commented, “I think that as AI takes over more routine tasks, the traditional idea of ‘entry-level’ will start shifting, and the ones at risk will actually be the middle layer – that is, the people who understand how things work but aren’t making the big decisions.”

Wyatt Mayham, founder and head of AI consulting at Northwest AIOpens a new window , thinks we will see 50% cuts in white collar jobs, “but not in the apocalyptic way that sounds. What’s really happening is that the shape of entry-level work is changing. Tasks that used to justify a full headcount, like data cleanup, writing first drafts, summarizing calls, building reports, etc, are now being handled by AI.”

Mayham continued, the entry job “role doesn’t disappear, but the value prop of a new hire certainly changes. You’re not going to hire someone to sift through spreadsheets; you’re ultimately hiring them to know what questions to ask and how to act on what AI surfaces.” This means that companies that treat AI as augmentation and not an outright replacement will be the companies that win.” Others, not so much.

As David Brudenell, Executive Director at Decidr.aiOpens a new window , an Australian AI company, warns, “AI won’t wipe out half of white-collar jobs. But bad leadership might. Let’s stop blaming the tools. Most layoffs come from a copycat strategy, not cutting-edge tech. AI becomes the scapegoat when leaders don’t have a better plan. Yes, AI will take over repetitive tasks. That’s not a crisis, it’s a cue. The smart move is to redesign work, not reduce headcount. Shift people to the things machines can’t touch: judgment, trust, creativity, relationships.”

For instance, one problem, which seems to have escaped some executives is “If they don’t hire entry-level tech-savvy people how will they ever learn enough to be mid-level staffers?In a New York Times story, a tech executive said “his company had stopped hiring anything below an L5 software engineerOpens a new window — a midlevel title typically given to programmers with three to seven years of experience.”

I hate to tell him, but mid-level developers don’t spring forth fresh out of college ready to do tough jobs.

As Jeff Tilley, MunclyOpens a new window , a Salesforce consulting company CEO, told me, “The real challenge is skill development. The actual problem won’t be the lack of entry-level positions, but rather that these positions won’t provide employees with proper skill development. This is what will make it harder for beginning clerks and white-collar workers to gain experience.”

Besides, as Raskauskiene observed, ‘entry-level’ doesn’t mean ‘low value.’ These roles are about building foundational skills, not just executing tasks. Even with AI, someone still needs to understand context and make relevant decisions – and that comes with experience, which has to start somewhere.”

There’s still time to pivot

What will all this mean for the job market? On LinkedIn, Jarrod Anderson, Chief Artificial Intelligence Officer at AI business consulting firm SYRV.AIOpens a new window , predicted, “In the immediate term, the labor market is likely to experience turbulenceOpens a new window rather than collapse. Over the next few years, we can expect heightened churn in certain white-collar jobs, but also rapid growth in demand for AI-related expertise.” Old jobs disappear, new jobs appear; it’s always been that way with technology.

Eric Vaughan, CEO of the AI development company IgniteTechOpens a new window , told me, “Amodei focuses on entry-level job elimination, but he’s looking at this backwards. Research shows AI gives the biggest productivity boost to lower-skilled workers. It’s the great equalizer. At IgniteTech, we didn’t eliminate entry-level positions. We elevated them. Our newest hires come in as ‘AI Innovation Specialists’ who can accomplish things that would have required senior-level expertise just two years ago.”

Vaughan continued, “When we implemented AI support systems, our entry-level technical support people jumped to L2-quality work almost overnight. They didn’t lose their jobs—they became more valuable. The AI handled routine tickets while they tackled complex problems they never could before. That’s the story Amodei isn’t telling. Yes, some roles will disappear. But entirely new categories of work emerge when humans and AI collaborate effectively.”

So, should you be worried? No, but you should be concerned and take steps to future-proof your career. Rather than let anxiety ruin your life, work out where you can use your talents in the 2030s job market. That means using AI to your advantage. I’ve reinvented my career several times over the decades. I started in mainframes, moved on to Unix, discovered I was better at translating tech jargon into English than programming IBM 360 Assembler and C, and here I am today using Perplexity instead of Google for story researchOpens a new window .  You can do it too.