What AI is Doing to Work

I see a lot of people talking about how AI tools will make work more efficient, or expect that using them will make work more efficient. The gist seems to be that AI tools do more in less time than a human can and that’s an efficiency gain. We can crank out more widgets or features or completed tickets in the backlog or whatever in the same amount of time, we’re so much more efficient!

AI, as a probabilistic system, doesn’t work like that.

That sort of linear efficiency logic makes sense when tasks are simple. Chopping wood, assuming certain parameters like having a sharp axe and idle weather conditions, is a simple task. Place the log in front of you; swing the axe (blade down) into the top of the log; repeat until the log is split into smaller pieces. There’s really no creative or critical thinking necessary there. It’s a “solved problem.” In cases like this or, to give another example, on a Fordist assembly line doing more of the same in a given time span equals increased efficiency.

The way AI is getting used in organizations, however, is far from that sort of situation. Instead, AI doesn’t always repeat the same pattern (and that’s on purpose!) and the people employed in the organization generally aren’t performing well-scripted tasks. They face complex problems that require thinking. Hence why post-Fordist, information-processing industries are sometimes called “knowledge economies.” In these sorts of workplaces it isn’t about doing the same thing over and over but faster; it’s about doing different things in different ways at different tempos and bringing them together at the right moment in order to (hopefully) get nonlinear productive effects. It’s about producing avalanches. Or if you prefer, phase shifts.

So to figure out whether you or your organization are getting those benefits requires a different mental model of how work happens and the way that AI is deployed within that work context. That mental model is going to be specific to you or your organization. And the specific benefits or harms of AI are actually non-fungible. They don’t transfer precisely because the context matters and the context is different in each situation. What to look for are those phase shifts. You can recognize when they happen because you’ll experience them as qualitative changes, not quantitative ones.

So asking if AI is making you more efficient is the wrong question. Ask if it’s making you better or worse in ways you care about and assess how so from there. Oh, and don’t necessarily expect that to translate into dollars. Money is a pure quantitative measure, so qualitative changes like this won’t really register there. So have fun with that, all you AI investors.