GenAI and The Productivity Paradox

The tradeoffs of going fast with AI

After spending the last couple of years co-developing data science notebooks, machine learning models, and LLM-enhanced demo applications, I’ve experienced firsthand both the power—and the tradeoffs—of building software this way.

The debate about whether AI actually makes developers more productive is still unsettled. Some claim it dramatically accelerates work. Others argue it simply shifts effort to different parts of the development cycle. An article in Harvard Business Review recently examined this question and reached a surprising conclusion: AI doesn’t necessarily reduce work—it often intensifies it.

The hype surrounding AI is impossible to ignore. Some of the loudest voices predicting the imminent disruption of high-skill knowledge work are also the ones building the models meant to replace it. These reports should be taken with a few shakes of salt. The reality on the ground is proving more nuanced.