The Promise of a Digital Co-worker
Tools like GitHub Copilot were introduced with a clear value proposition: to accelerate development and reduce the frustrating, repetitive parts of coding. By handling boilerplate code, suggesting solutions, and even generating tests, these AI assistants
were meant to conserve developers' mental energy, help them stay focused, and ultimately make their jobs more satisfying. Early studies showed significant gains, with some developers completing tasks up to 55% faster. The vision was simple: let the machine handle the mechanical work so humans can focus on the complex, architectural decisions that drive innovation.
The Productivity Paradox in Action
The problem isn't that the tools don't work; it's how their output is being managed. Instead of banking the time saved, many organisations are simply raising the bar. A common pattern has emerged: management sees a tool that promises a 50% speed boost and translates that directly into a 50% increase in expected output. What was once a productivity gain becomes the new baseline. Studies have found that while executives expect AI to boost productivity, a significant majority of employees report that AI has actually added to their workload. This creates a cycle where any efficiency gain is immediately consumed by higher demands, leaving developers feeling more pressured than before.
More Code, Not Necessarily Better Code
A critical flaw in this push for higher output is the focus on quantity over quality. AI assistants are incredibly proficient at generating vast amounts of code, but this code isn't always efficient, secure, or maintainable. This shifts the developer's role from a creator to a reviewer, tasked with the mentally taxing job of verifying and debugging AI-generated output. This is often called the "verification overhead." One study found that while developers using AI felt 20% faster, they actually took 19% longer to complete tasks correctly because of the time spent correcting the AI's work. Furthermore, the flood of new code creates bottlenecks in later stages like code review and testing, which still rely heavily on human judgment.
The Human Cost of AI-Driven Quotas
The relentless pressure to do more is leading to a new form of cognitive exhaustion, sometimes called "AI brain fry" or "AI fatigue". Developers report feeling a unique kind of burnout, one born not just from long hours but from the high intensity of constantly supervising, correcting, and managing AI agents. This is compounded by a loss of ownership and professional pride; when a machine generates the code, the developer's sense of authorship can diminish. Research shows that frequent AI users can report significantly higher burnout rates than their peers, feeling overwhelmed by the expectation to produce more without adequate training or support. The very tool intended to reduce frustration is, in many cases, amplifying it.
Shifting Focus from Output to Outcomes
Experts argue that the solution lies in redefining productivity. Instead of measuring crude metrics like lines of code, organisations should focus on outcomes that matter: software quality, system reliability, and developer well-being. This involves a cultural shift where management trusts developers to use tools to work smarter, not just faster. Success with AI requires more than just installing a tool; it requires a strategy that supports developers with training, sets realistic expectations, and values deep, thoughtful work over sheer volume. Without this, companies risk burning out their most valuable assets in a misguided quest for AI-fueled efficiency.
