Bad idea: “Artificial Intelligence” automatically improves productivity.
AI improves productivity by enabling automation of what could not be previously automated.
AI boosts productivity in the same way all automation does: by reducing the cost of effort. What’s new is the ability to automate more creative tasks that could not be previously automated.
We can always be more productive, if the outputs don’t have to work.
AI, specifically Large Language Models (LLMs), is (still) prone to hallucinations. Even worse, these hallucinations can appear convincingly accurate to novices. Generating more output that doesn’t function correctly is not the same as being more productive. To truly assess productivity, one must include the additional effort required to fix any resulting issues.
AI can lead to overconfidence; overconfidence reduces productivity.
From the Accelerate State of DevOps Report 2024:
“…since AI allows respondents to produce a much greater amount of code in the same amount of time, it is possible, even likely, that changelists are growing in size. DORA has consistently shown that larger changes are slower and more prone to creating instability.”
Essentially, AI can lead to overconfidence and a tendency to revert to the inefficient practice of large batch processing. The issues arising from large batches can negate any productivity gains achieved through AI.
Short-term productivity might not indicate long-term productivity.
If junior employees rely too heavily on LLMs to complete tasks without truly learning and understanding the work, they will never develop expertise. As current experts retire and there are no skilled individuals to replace them, we can expect productivity to decline in the long run.
