· nervico-team · liderazgo-tecnico · 6 min read
Firing Developers for AI: Why It's Going Wrong and What to Do Instead
Companies that replaced developers with AI are accumulating technical debt and losing quality. Junior hiring is down 50%, creating a future crisis. The answer is not fewer engineers but better tools in better hands.
Klarna laid off 700 customer service staff and replaced them with OpenAI-powered AI. Months later, CEO Sebastian Siemiatkowski publicly admitted: “We went too far.” Service quality dropped, customer complaints rose, satisfaction ratings plummeted. Klarna started rehiring humans in a hybrid model.
This is not an isolated case. Builder AI, a startup valued at $1.5 billion, collapsed when court filings revealed that 700 human engineers in India were manually performing tasks the company marketed as “fully autonomous AI.” According to Challenger, Gray and Christmas, approximately 55,000 jobs were eliminated in 2025 directly attributed to AI. And 55% of companies that rushed to replace humans with AI now regret that decision.
The narrative of “AI replaces developers” sounds great in an investor presentation. In reality, companies are discovering it is more complicated than it looks.
The Broken Promise
The promise was attractive: AI can write code, so we need fewer programmers. Engineers at Anthropic and OpenAI claim that 100% of their code is now written by AI. Claude Code creator Boris Cherny declared: “Coding is practically solved.”
But there is a difference between “AI can write code” and “AI can maintain a production system.” A CAST Software study analyzing 10 billion lines of code detected a 4x increase in “code cloning” caused by AI. The code works, but nobody understands why. And when it fails, nobody knows how to fix it.
The Veracode report confirms that 45% of AI-generated code contains security vulnerabilities. In Java, the rate exceeds 72%. And most concerning: this figure does not improve with larger or newer models. This is not a temporary problem. It is structural.
A Fortune survey of 6,000 executives revealed that nearly 90% of companies had seen no impact on employment or productivity from AI over the past three years. Apollo’s chief economist summarized: “AI is everywhere except in the incoming macroeconomic data.” Nobel Laureate Daron Acemoglu (MIT) described the gains as “disappointing relative to industry promises.”
The Junior Death Spiral
This is perhaps the most alarming development of all.
Junior hiring has dropped nearly 50% between 2023 and 2025. A Stanford study (Brynjolfsson, Chandar, Chen) confirms that developers aged 22 to 25 have lost nearly 20% of their jobs since late 2022. Developers over 26 have maintained stable or growing employment.
Computer science enrollment at the University of California dropped 6% in 2025, the first decline since the dot-com bust. New graduate hiring at the 15 largest U.S. tech companies has fallen 55% since 2019. Unemployment among CS graduates reached 6.1% in 2025.
Economists call this the “junior death spiral” and the logic is simple: if you do not hire juniors today, you will not have seniors in five years. AI can automate basic programming tasks, but it cannot create experienced engineers. That requires years of real work, mentoring, mistakes, and learning in production systems.
The paradox is perfect: companies need juniors who will become tomorrow’s seniors, but AI has eliminated the entry ramp. The same talent that is “too junior” today is exactly the talent that will be “impossible to find” in 2030.
What Brad Smith (Microsoft) Says
Microsoft’s president was direct: “We’re not talking about using AI to replace software engineers, but we are talking about using AI to change the art of software engineering, to uplevel the people who are in this extraordinary and extraordinarily important profession.”
Smith added: “Technology should help people get smarter. Usually, when that happens, you find you want more people who can do this work. You’re even willing to pay them more than in the past.”
This is a relevant message coming from a company that has integrated Copilot across its entire product suite. Microsoft is not eliminating developers. It is investing in tools that make them more productive.
The U.S. Bureau of Labor Statistics projects 17% growth in software development jobs from 2023 to 2033, adding 327,900 new positions. Significantly above average for all occupations. Not exactly the profile of a dying profession.
The Model That Works: Senior Teams + AI as Multiplier
The EY data is clear: only 17% of organizations are reducing headcount because of AI. The remaining 83% are doing something different: using AI so their existing teams produce more.
This is the model we see working in practice:
AI automates the repetitive. Boilerplate generation, basic testing, documentation, refactoring known patterns. All of this can and should be automated.
Seniors focus on what matters. Architecture, design decisions, security, critical code review, business domain knowledge. These are the tasks AI cannot do well and that determine whether a system works or becomes a ticking time bomb.
Juniors learn in a supervised context. Instead of eliminating juniors, smart companies are training them to work with AI: reviewing model output, understanding limitations, identifying vulnerabilities. It is a new skill that did not exist two years ago, but will be critical in five.
The result is a team that produces more, faster, with fewer errors. Not a smaller team that produces the same with more risk.
How to Prepare Your Company
If your company is considering how to navigate this transition, there are five questions you should ask:
Have you measured the real impact? Not what was promised in the demo, but what was measured in production. How many additional bugs, how much new technical debt, how much review time.
What about security? If 45% of AI code has vulnerabilities, who is detecting them in your organization? Do you have a review process that accounts for this?
Do you have a talent pipeline? If you stop hiring juniors, who will maintain your systems in 2030? Senior talent does not appear spontaneously.
Are you multiplying or substituting? Companies using AI to multiply their engineers are winning. Those using AI to substitute them are accumulating debt.
What is your competitive advantage? If your advantage depends on the quality of your engineering, cutting engineers is the worst possible investment.
At NERVICO, we work with senior teams that use AI as a multiplier. Not as an excuse to reduce quality. Not as a shortcut to avoid difficult decisions. As a productivity tool in the hands of people who know how to use it.
If you want to evaluate how to integrate AI into your development team without the risks we are seeing in the market, our free technical audit can help you define a realistic plan.
