AI Was Supposed to Save Us Time. Why Are We Still So Busy?

I use AI because it genuinely saves me time. It helps me organize information, work through ideas, research unfamiliar topics, and get through repetitive parts of my work much faster than I could before. The more I use it, the more I’ve started to notice something like an AI productivity paradox: the tools get faster, but the free time never quite shows up.

In theory, that should leave me with more free time. What usually happens instead is that I use the time I saved to do something else.

A task that once took an hour might now take twenty minutes, but the remaining forty minutes rarely turn into a longer lunch or an earlier finish. I usually move on to another task, start something that I previously would not have had time for, or decide that I can fit one more thing into the day.

The more I use AI, the more I have started to wonder whether technology actually gives us time back, or whether it simply increases the amount of work we believe we can reasonably do.

AI really does make some work faster

The time savings are not imaginary.

BCG’s fourth annual AI at Work survey found that 42% of frontline employees who use AI regularly save at least a full workday each week.

An extra day in a working week sounds significant, especially when the same tools are being used across writing, research, coding, analysis, meetings, and administrative work.

The interesting part is what happens after that time has been saved.

Most discussions about AI productivity stop at the moment a task becomes faster. If something that used to take sixty minutes now takes twenty, we count forty minutes as a productivity gain.

That makes sense mathematically, but it does not necessarily describe what the working day feels like.

The forty minutes still exist, and somebody usually finds something to put in them.

A faster task does not necessarily mean less work

One of the things I notice when using AI is that it often speeds up one part of a process rather than removing the whole process.

Writing is an obvious example.

AI can help produce a first draft very quickly, but I still have to decide whether the argument makes sense, whether the information is accurate, whether the tone sounds natural, and what needs to be removed or rewritten.

The first version arrives faster, which is useful, but the judgment around it still belongs to me.

The same pattern appears in other kinds of work. Research can be summarized more quickly, but somebody still needs to decide what matters. Code can be generated faster, but it still needs to work. Ideas can be produced almost instantly, but having more options creates more decisions.

AI reduces the cost of producing something, but it does not always reduce the cost of deciding whether that thing is good.

We can create more options than we can comfortably evaluate

This is one of the changes I find most noticeable.

Before generative AI, asking for ten versions of an idea could be impractical because producing them took time. Now ten alternatives can appear almost immediately.

That sounds like an advantage, and often it is. More options can help when I am stuck or when I want to explore several directions before committing to one.

The problem is that every additional option also needs to be evaluated.

If I ask AI for ten headlines, I now have ten headlines to compare. If I generate five approaches to a problem, I still have to decide which approach makes the most sense. If I ask for several versions of a piece of writing, I may spend more time comparing and combining them than I expected.

Producing alternatives has become extremely cheap. Human attention has not.

That means the bottleneck can move from creation to judgment.

AI makes more work feel possible

There is another effect that I think matters just as much.

When something becomes easier to produce, we tend to produce more of it.

If a report takes half a day, there is a natural limit to how many reports anyone is going to request. If AI cuts that process down to an hour, suddenly more reports become realistic.

The same applies to content, analysis, personalization, research, testing, documentation, and almost anything else AI can accelerate.

This can create real value. There are projects I can attempt now that would have been difficult to justify before because they required too much time.

But the ability to do more can quietly turn into the expectation that I should do more.

Once a certain level of output becomes possible, it does not take long before it starts to feel normal.

Working for yourself makes this especially easy

When I work for myself, nobody tells me what the correct amount of output should be.

If AI makes research faster, I can research more. If it makes writing faster, I can publish more. If it helps me build something I previously would have needed outside help for, I can add another project to my list.

That freedom is one of the things I like about AI. It gives one person access to capabilities that would once have required considerably more time or resources.

It also makes it very easy to keep expanding the amount of work I expect from myself.

The capacity arrives first, and the expectations tend to follow.

After a while, the extra capacity no longer feels like extra capacity. It simply becomes the new baseline.

That is probably one reason the productivity gains from AI do not always feel as dramatic as the numbers suggest. We quickly adjust our expectations to what is newly possible.

Using AI creates work of its own

AI also introduces a new layer of work that did not exist before.

I have to decide which tool is appropriate, explain the context, give instructions, review the result, correct mistakes, and decide how much of the output I actually want to use.

Sometimes this process is still dramatically faster than doing everything manually. Other times I realize that I have spent so long refining instructions and correcting the result that the shortcut was smaller than I expected.

There is also the constant stream of new tools, features, models, and integrations to evaluate.

Even when I decide not to use most of them, figuring out what is worth paying attention to takes some attention of its own.

This does not make AI less useful. It simply changes the nature of the work.

I spend less time creating certain things from scratch and more time directing, reviewing, selecting, and deciding.

The real question behind the AI productivity paradox

I used to think about AI mainly in terms of efficiency.

Could this make the task faster?

That is still useful, but I have started asking another question more often: does this task need to be done in the first place?

There is not much value in making unnecessary work extremely efficient.

AI makes it tempting to expand everything. More research, more versions, more ideas, more experiments, more content, more projects. When generating something becomes easy, it can be difficult to notice when the better decision would have been not to generate it at all.

Sometimes the best use of AI is not doing more work. It is reducing the time spent on work that deserves less attention.

Saving time and keeping time are different things

I still want AI to make my work faster, and I expect that it will become much better at doing that.

What I am less convinced of is that faster work will automatically lead to shorter working days.

Technology can reduce the time required for a task, but it does not decide what happens to the time that remains. A company can fill it with more output, and someone working independently can fill it with another project, another idea, or another responsibility.

AI may already be saving us a significant amount of time.

The harder part is deciding whether we are going to keep any of it.