The Productivity Placebo Effect
On the surface, AI coding assistants are a runaway success. Developers report feeling significantly more productive and fulfilled. Surveys from GitHub show users feel 55% faster on tasks and are able to focus on more satisfying work. They report feeling less
frustrated and better able to maintain their 'flow state'. However, a gap is emerging between this feeling of speed and the actual time it takes to complete a project. Some rigorous studies have found that while developers believe they are working much faster, their actual task completion times can sometimes be slower. One 2025 study by METR, a model evaluation group, found that experienced developers using AI tools were surprisingly 19% slower on average, even though they believed they were about 20% faster. This gap between perception and reality is being called the 'AI productivity placebo'.
More Code, More Problems
So if AI is generating code faster, where does the time go? The answer often lies downstream. While AI accelerates the initial writing phase, it doesn't eliminate the other essential parts of software development: reviewing, testing, debugging, and deploying. In fact, it can create new bottlenecks. A study by Faros AI found that teams with high AI adoption merge nearly 98% more pull requests, but the size of these requests has grown by 154%. This has led to a 91% increase in the time spent on code review. As one developer noted, one person's 10x output can become another person's 10x review burden. The time saved in typing is often shifted to reading, evaluating, and sometimes untangling complex or incorrect AI-generated code. Developers report spending more time debugging AI-generated code, which eats into the initial time savings.
Parkinson's Law in the Age of AI
There's a classic principle in productivity known as Parkinson's Law, which states that "work expands to fill the time available for its completion." This 1955 observation is proving highly relevant in the AI era. When AI tools create efficiency and free up time, the natural tendency isn't always to log off early. Instead, that newfound time is often filled with more work. This could mean tackling more features, aiming for a higher level of polish, or endlessly refactoring for 'future-proofing'. The scope of a project can creep ever larger simply because the capacity to do more exists. Without clear priorities and the discipline to focus on outcomes over hours logged, AI's efficiency gains are simply absorbed into doing more, rather than finishing sooner. Organizations may see more activity, but not necessarily faster delivery of what truly matters.
A Shift in the Developer's Role
Ultimately, AI is changing the very nature of a developer's job. The most valuable skill is shifting from writing code to evaluating it. The core work is moving from implementation to strategy and oversight. Time previously spent on boilerplate code or searching for syntax is now reinvested in higher-level activities. An MIT study found that developers with AI tools increased their time on core coding activities by over 12% while decreasing time spent on project management tasks by nearly 25%. They are able to do more of what they enjoy—the complex problem-solving—and less of the administrative drudgery. This shift is largely positive for job satisfaction, with studies showing that up to 90% of developers feel more fulfilled when using AI assistants. The 'saved' time isn't vanishing; it's being reinvested into creating better, more robust software, even if it doesn't shorten the workday.
