Last week in this newsletter, I summarized some interesting results from a study that analyzed the behavior of 164,000 knowledge workers. It found that introducing AI tools increased administrative tasks by more than 90% while reducing deep work effort by almost 10%.
The problem, I concluded, was that digital productivity tools sometimes speed up the wrong tasks, which might feel efficient in the moment, but lead us to accomplish less over time. As I emphasized, AI is not the only technology to produce this paradoxical side effect —we saw something similar with email, mobile computing, and online meeting software as well.
So, what’s the solution to avoid these traps?
In today’s episode of my podcast, I suggested three ideas that might help. I want to summarize them here as well:
Idea #1: Use a Better Scoreboard
Make sure you measure what actually matters in your job. If you’re a professor at a research institution, for example, this might be the number of papers you publish per year. If you’re a team manager, it might be the number of priority projects completed per month.
When you introduce new digital productivity tools into your workflow, don’t focus too much on their impact on individual tasks (e.g., “Wow! That email was much faster to send than a fax,” or “AI just finished a task in 20 minutes that would have taken me 3 hours!”). Pay attention instead to your scoreboard. If you’re not producing more valuable output than before, the tool isn’t really making you more productive.
Idea #2: Focus on the Right Bottlenecks
If you look closer at many knowledge work projects, you’ll identify a key bottleneck that determines how fast they can be accomplished. If you want to become more productive, you should look for ways to deploy tools that improve this specific step.
When working on Deep Work, for example, I spoke with a prominent Wharton professor who told me that one of the keys to publishing journal papers in his field was access to interesting data sets. He published more papers per year than most of his peers, largely because he spent more time building relationships with companies and institutions in search of good data. This was the bottleneck for his work.
Accordingly, any tool that could help him cultivate more such relationships and gather better data from the relationships he had already formed would directly improve his productivity. Compare this, for example, to using Claude Code to speed up the process of producing plots for his papers. This might, in limited windows of time, make his job more convenient, but not necessarily increase the number of papers he publishes per year.
Idea #3: Separate Deep from Shallow Work
My final idea is the simplest: on your daily calendar, clearly separate time for focused effort that directly produces value from administrative, logistical, and collaborative tasks. In this way, if a digital productivity tool ends up accidentally increasing the volume of shallow work you face each day, you’ll limit the damage to your ability to make progress on important projects.
This makes it easier to experiment with different tools without worrying that you might end up — like many of the subjects in the study cited above — suddenly overwhelmed by the ultra-fast processing of minutiae while the big things slowly languish.
