AI Will Steal Developer Jobs (But Not How You Think)
Anthropic’s CEO has said that in three to six months, AI will be writing 90% of the code that software developers are in charge of. Keep in mind, he said this in March.
While the timeline may be fantasy, the reality is that AI is changing what development looks like and it may shift how development divides work as well — merging the frontend and backend, and possibly creating other shifts in how we think about who does what in IT.
For example, recently I asked a frontend-turned-backend developer if he used AI. Absolutely, he said — in fact, he’d recently used it to create in minutes a complete frontend for an application he’d developed. He just picked a design system and told the AI what to build — and poof, it was done. “I didn’t have to worry about libraries,” he added.
He received a lot of compliments on it, apparently.
It’s Real: Developer Anxiety About AI
Lee Robinson talks to a lot of frontend developers as the vice president of developer experience at Vercel, which oversees the framework Next.js and is a web hosting provider.
“There’s a lot of anxiety about AI and software creation in general, not necessarily just frontend or backend, but people are rightfully trying to understand what does this mean for my career,” Robinson told The New Stack. “If the current rate of improvement continues, what will that look like in 1, 2, 3, 4 years? It could be pretty significant. So it has a lot of people stepping back and evaluating what’s important to them in their career, where they want to focus.”
Armando Franco also sees anxiety around AI. Franco is the director of technology modernization at TEKsystems Global Services, which employs more than 3,000 people. It’s part of TEKsystems, a large global IT services management firm that employs 80,000 IT professionals.
Franco theorized that developers who are most concerned, perhaps, aren’t fully aware of where the technology is.
“We like to keep our staff as closely tied to the technology as possible,” he said. “We want them to have a sense of what we call continuous engineering or continuous experimentation so that they feel safe that, if there’s a new technology, a new tool out there, they can bring it in and we will support them to develop it.”
TEKsystems Global Services has used this approach with AI, embracing the tools and technology for coding. As a result, he said, on-staff developers are excited about the potential.
“From my perspective, our developers are really looking at these tools and saying, ‘How can we accelerate the work that we do? How can we become more efficient? How can we ensure better quality? How can we ensure security?’” Franco said. “That’s the lens that we’re looking at [through].”
Separating Fear From Reality
Josh Comeau, a frontend developer and trainer, reflected that even if AI does generate some amount of code, it doesn’t mean that it has replaced developers.
In a recent post, he pointed to a 2024 Forbes headline that touted, “AI Writes Over 25% Of Code At Google—What Does The Future Look Like For Software Engineers?”
“The article’s title makes it sound like AI is doing 25% of the work and human developers are doing the other 75%, but that’s not actually what’s going on here. That title is misleading, in my opinion,” Comeau said. “AI may be generating 25% of the code that gets committed at Google, but it’s not acting independently.”
In fact, to his knowledge, 100% of code at Google is still being created by developers, although AI may be one tool they’re using to do so, he added.
This isn’t the first time in history that people have fretted about new technologies, pointed out Shreeya Deshpande, a senior analyst specializing in data and AI with the Everest Group, a global research firm.
“Fears that AI will replace developers mirror historical anxieties seen during past technology shifts — and, as history shows, these fears are often misplaced,” Deshpande said in a written response to The New Stack. “AI will increasingly automate repetitive development tasks, but the developer’s role will evolve rather than disappear — shifting toward activities like AI orchestration, system-level thinking, and embedding governance and security frameworks into AI-driven workflows.”
“Success will hinge on reimagining developer roles and investing in dynamic upskilling strategies, not on workforce reduction.”
– Shreeya Deshpande, senior analyst, Everest Group
She wasn’t the only one to say this: Over and over, no matter who I asked or what I read, the answer is that developers will still be needed. The real question seems to be more about what role they serve.
I’ve repeatedly heard that developers will still be needed because AI can hallucinate, even in code. For instance, it’s been known to make up API endpoints that don’t exist, one source told me. So “developers will need to check AI code” is an oft-repeated response to the question, “Will AI replace developers?”
But is that a role developers want? Given how much my enterprise software developer spouse complains when fixing code created by someone else, I’m not so sure.
Fortunately, Deshpande and others foresee a more promising world, where developers upskill and find new specializations. She pointed to a similar adaptation that happened with cloud and low-code. We’re already seeing some of that with the vibe coding trend.
“Organizations should recognize that the bottleneck to realizing AI’s full potential will be AI literacy and workforce agility, not a shortage of human developers,” Deshpande stated. “Success will hinge on reimagining developer roles and investing in dynamic upskilling strategies, not on workforce reduction.”
Franco is also optimistic that AI and large language models (LLMs) will replace rote memorization of code, leaving developers to focus on more creative tasks.
“LLMs are not all-knowing. They’re not all-powerful. It’s our responsibility to help them understand what the requirement is and then to refine it,” he said. “It’s still the developer’s responsibility to shape what that solution looks like, and it’s just going to help them grow a little bit faster.”
How Developers Can Prepare for AI
First, don’t panic and leave IT, advised Franco.
“If I go back 20 years, right, there used to be Infrastructure Engineers, database engineers, network engineers; and as the world built or switched towards DevOps and cloud computing, that became DevOps engineers, and now they’re becoming platform engineers,” Franco said. “ So I would say to anyone that’s in technology: No, don’t leave.”
Instead, adopt a continuous learning mindset, he advises. Assume the technology is there for you to use, not to replace you.
Robinson’s guidance is straightforward: Stop focusing on where you’re developing and focus instead on creating products and delivering business value.
“It’s less about can I do CSS, can I build an API, can I do frontend, can I do backend? It more about what is the functionality that I want to add to the product to make it successful, to add revenue to the business, to solve a customer pain point.”
– Lee Robinson, vice president of developer experience, Vercel
“As AI continues to write more and more code for us, and humans shift into an architect and reviewer role, we’ll still write code,” he said. There’ll be more code written than ever, but the relationship changes.
“And in that world, it’s less about can I do CSS, can I build an API, can I do frontend, can I do backend? It’s more about what functionality I want to add to the product to make it successful, to add revenue to the business, to solve a customer pain point. That’s where the value creation shifts.”
It wouldn’t hurt to learn security, either, since AI will cause security to “shift left” — meaning move into development — even more than it has already, Franco said.
Developers will need to train the AI as well, grounding it in best practices and the organization’s mandates, he added. He recommends developers practice working with AI as copilots and in prompting for everyday tasks and, well, just chatting with it.
“I think the number one mistake that people make when prompting with an AI, regardless if it’s for code development or if you need a little bit of help doing some research, it’s how you address the AI,” he said. “A lot of people think that they have to switch the way that they speak and talk to the LLM in some type of structured manner so that it understands. Actually, that’s completely opposite of the way that you should approach things.”
Be specific, but realize that the LLM can train on how you speak, he advised. It will produce better results if it adapts to you rather than you changing how you talk to it, he said.
Above all, engage, he added.
“This is just a new tool in your tool belt,” Franco said. “Attach to it and learn how to use it, and it’s going to make you a better technologist.”
