What busywork would you get rid of first?

My dad, who had an accounting background, used to say a lot of human potential was wasted on tax work. The tax code is complicated enough that companies and people need clever accountants and attorneys to help them pay less. The work is real and the clients benefit. What he thought was a shame was that it had to exist at all: the rules are convoluted enough that navigating them became a career for a lot of smart people.

What stuck with me is that life’s too short to spend on drudgery, especially the kind we impose on ourselves.

There’s another kind of drudgery that nobody designed on purpose: managing our inbound messages. Triaging inbound happens for many of us all day: deciding what needs a response, what needs a decision, what can be ignored. It isn’t irrational. Something important with high stakes might be buried in there. Most checks are low-value on their own, but could reveal something critical. In my corporate jobs this was inescapable and I had a persistent sense that this wasn’t what I should be spending my life on.

Recently, on the a16z podcast, Marc Andreessen said that 30 years from now people will look back and not believe how much time we spent on certain things: “such a waste of human potential… such a waste of human creativity.” That reminded me of my dad’s observation about the tax code and about inbox triage. I think there’s a lot of drudgery that we shouldn’t have to do.

The average knowledge worker gets around 270 messages a day, 117 emails and 153 chats. We get about an hour and twelve minutes of real, uninterrupted focus out of an eight-hour day. More than half the day goes to overhead instead of the work. Multiply that across every knowledge worker, every day, for years, and it adds up to an enormous number of human lifetimes spent sorting noise.

We finally have a choice about what we can do here. Large language models are good at exactly the work it requires: reading text and pulling out meaning and signal. That doesn’t need a breakthrough. I’ve done it reliably with pretty modest models. The low-value but necessary part can be handled for us.

We don’t have to wait 30 years to stop being horrified at it. It’s already sitting in our inboxes, and we can start taking that time back today.

I suspect most of us have our own version of my dad’s tax code observation. What’s yours?