Two kinds of AI work

A recent episode of The AI Daily Brief drew a line between what it called Efficiency AI and Opportunity AI. I have been circling that same line for a year without a name for it, so I am borrowing theirs and putting it in my own words: improving a workflow, and creating one.

Improving a workflow is the easy half. I already wrote about it: start with the thing you do forty times.

The harder half

Creating a workflow is harder, because there is no old version to compare against. The test that actually works is not "is this faster." It is: did this exist at all, a year ago?

What passed the test

HALO is my answer. It checks my system, launches the software I use, answers a question, drafts something, all from one page I built for myself. I did not have a slower version of HALO before I built it. There was no HALO-shaped hole in my day that I was filling faster. There was just nothing there, and now there is.

That is the uncomfortable part of this half. Forty-times work hands you a scoreboard: you already know what correct looks like, because you have done it a hundred times by hand. A new workflow hands you nothing to compare against. The only judge is whether you keep using it, which is a slower and less certain kind of proof.

It also asks something different of you. Not "what is annoying me," but "what could I have that nobody has handed me yet." That is closer to an entrepreneur's question than an engineer's, and it is why the genie in yesterday's fable eventually stopped handing over cars. At some point the wish stops being about the thing you already have, faster, and starts being about the thing you do not have at all.

Be Better - Ask what does not exist yet.