If AI is really making us more productive... why does it feel like we are working more, not less...?

*The AI Productivity Paradox*

The promise of Artificial Intelligence (AI) was to optimize our work processes, freeing up time and increasing efficiency. On paper, the tools are delivering on this promise. AI-powered code writers like Github Copilot and Claude can produce effective code, while Large Language Models (LLMs) can summarize meeting minutes and automate tasks like Jira ticket management. However, a closer look at the data reveals a different story.

The Jevons Paradox in Action

The Jevons Paradox states that as a resource becomes more efficient to use, its consumption actually increases, not decreases. This is precisely what's happening with AI. While the tools are making our work processes more efficient, they're also creating a sense of obligation to produce more. We're not using AI to buy back our time; we're using it to increase our "Normal Output" threshold.

For example, if an AI-powered tool can write code faster and more efficiently, the expectation is that we'll write more code. If a tool can summarize meeting minutes quickly, the expectation is that we'll attend more meetings and be responsible for a greater volume of information. This is not about being more productive; it's about being more productive in a specific, quantifiable way.

The Ceiling of Expectations

The data shows that AI has increased our capacity to produce work, but it has not reduced our burden. In fact, the opposite is true: the ceiling of expectations is rising faster than the floor of our workload. We're optimizing the "how" of work to near-perfection, but the "how much" is scaling even faster. This is like running on a treadmill that keeps getting faster – we're not getting any closer to stopping.

The Question Remains

So, is AI making us more productive? The answer is yes, in a narrow sense. We can produce more work, faster and more efficiently. But the real question is: if the ceiling of expectations keeps rising as fast as the tools, do we ever actually get to stop climbing? The answer is unclear, but one thing is certain: we need to re-examine our relationship with AI and the expectations that come with it.