I left the house for a walk. Proper one. Phone in pocket, no agenda, just some fresh air. About two minutes in, my AI assistant, an assistant I built myself that monitors news and sends me story ideas, pinged me with a list of blog post candidates. Good ones actually. But the way it presented the source URLs was ugly. Long, raw links sitting there unformatted instead of being neatly wrapped in readable text. Small thing. I asked it to fix it. It said it had. I asked it to prove it by running a fresh example. Same ugly URLs. I pointed that out. It looked again, said it found the problem, said it fixed it. Another test. Still broken. Two hours later I was sitting on a park bench, the walk I came for long forgotten, the URLs still just as broken as when I started.
This is not a story about a broken AI assistant. This is a story about what AI actually does to your time, your attention, and your to-do list. And it is more complicated than the pitch.
The Research That Started This Conversation
A few weeks ago Stanford published a study tracking the AI habits of over 200,000 US households. The headline finding was striking. People using generative AI tools were completing everyday digital tasks, things like job hunting, travel planning, and online shopping, between 76% and 176% faster than people who were not using AI at all.
That is not a small efficiency bump. That is transformative by any measure.
But here is the part that got buried under the headline. When researchers looked at what people did with all that saved time, the answer was not what anyone expected. They did not use it to learn new skills. They did not start side projects or get ahead on work. They watched Netflix. They scrolled Instagram. They relaxed.
Which, honestly, is fine. Leisure has value. But it does complicate the story AI companies have been telling us about unlocking human potential and freeing people up for higher-order thinking.
Then There Is the Other Study
At almost the same time, a separate wave of research was painting a completely different picture.
UC Berkeley’s Labor Center followed workers who adopted AI tools through 2025 and found that 67% of them reported working more hours by the end of the year, not fewer. Harvard Business Review research from early 2026 confirmed that AI tool adoption is correlated with increased work intensity, not decreased workload. An Upwork survey found that 77% of workers felt AI tools had actually increased their workload overall.
Same tools. Opposite outcomes. How?
The difference is not the AI. It is the context. The Stanford study was looking at individuals doing personal tasks at home, where nobody is adding to your list when you finish faster. The Harvard and Berkeley research was looking at workplaces, where finishing faster just means your manager sets a higher baseline for next time.
As one Fortune report put it, the reality for a lot of workers is blunt: you used to spend six hours on that. Now it takes forty minutes. But nobody is sending you home early.
Two outcomes of Ai
There Is Actually a Name For This
Economists call it the Jevons Paradox. It was first observed in the 1800s when a British economist noticed that more efficient steam engines did not reduce coal consumption. They increased it. Because when something gets cheaper or faster or easier, you do not do less of it. You do more. Efficiency expands demand.
AI is doing exactly this, just to your calendar instead of a coal mine.
When I started using AI tools seriously, I could write a blog post in a fraction of the time it used to take. So I started writing more posts. I could research a topic in minutes instead of hours. So I started researching more topics. I built an AI assistant that scouts news stories for me. So now I have more story ideas than I can ever publish. I used vibe coding to build and ship projects I never could have attempted before. So now I have more projects running simultaneously than I can comfortably manage.
The finish line did not get closer. It moved.
The New Category of Work Nobody Warned You About
There is something the productivity research does not fully capture, and my park bench afternoon illustrated it perfectly.
AI does not just help you do existing tasks faster. It creates entirely new categories of task that did not exist before you had AI. Before I built my news assistant, I did not have to manage a news assistant. Before I had automated blog workflows, I did not have to debug automated blog workflows. Before I had a self-hosted AI model running on a server, I did not have to maintain a self-hosted AI model running on a server.
Every capability AI gives you comes with a surface area for things to go wrong. And when they go wrong, you are the one who has to sit with them until they are fixed, or until you accept that today is not the day.
I went out for a walk. I came back two hours later having debugged something, fixed nothing, and rested not at all. That is a new kind of busy that did not exist in my life eighteen months ago.
Constant building and breaking with Ai
So Is AI Worth It?
Honestly, yes. For me, unambiguously yes. In the last few months I have built a crypto paper trading bot, shipped a full website for a tailor in Ghana, run a content engine, and managed all of it while working a full-time job. None of that was possible at this pace before AI entered the picture.
But I want to be honest about the trade. AI did not give me time back. It gave me leverage. And leverage is not rest. Leverage is the ability to do more, which mostly means you do more, which means you need more rest, which you then do not take because there is more to do.
The Stanford people spending their AI-saved time on Netflix might actually be onto something.
FAQ
Does AI actually save you time or does it just create more work? Both, depending on how you use it. For personal tasks with a fixed scope, AI genuinely saves time and that time often goes toward rest or leisure. For builders, entrepreneurs, and knowledge workers, AI tends to expand ambition and create new categories of work, so the total time spent often stays the same or increases even as individual tasks get faster.
What is the AI productivity paradox and does it affect everyone? The AI productivity paradox describes the pattern where AI tools make individual tasks faster but do not reduce overall workload, and in many cases increase it. It affects knowledge workers and builders most acutely, particularly in workplace settings where productivity gains are absorbed as higher output expectations rather than freed time.
Why do I feel busier since I started using AI tools? Probably because you are. AI lowers the cost of doing things, which means you attempt more things. Each new capability also creates new maintenance, new decisions, and new rabbit holes. This is the Jevons Paradox at work: efficiency does not reduce consumption, it expands it.
Takeaway
AI is genuinely useful. I am not going back, and I suspect you are not either. But the version of AI that hands you back your evenings and clears your weekends was always a little optimistic. What it actually hands you is more runway. What you do with that runway is still entirely up to you.
AI is genuinely useful. I am not going back, and I suspect you are not either. But the version of AI that hands you back your evenings and clears your weekends was always a little optimistic. What it actually hands you is more runway. What you do with that runway is still entirely up to you.
So I want to know where you land. Are you the Stanford side of this story, AI saved you real time and you are actually using it to breathe? Or are you the Berkeley side, moving faster, doing more, somehow busier than before you started? And how deep are you actually in with these tools right now, dabbling, building, or fully dependent?
If you want the honest, unfiltered version of what building with AI actually looks and feels like, subscribe to my newsletter. I write about the bench moments as much as the wins.