> If it's a multiplier you need to either increase the requested work to keep the same humans or reduce the humans needed if you keep the same workload. It's not straightforward which way each business will follow.

I guess what you’re saying is technically true while being somewhat misleading.

“Increase the requested work” is one way of saying “reduce the amount of scutwork that needs to be done”. Personally, I’m ok having less scutwork. I’m also ok letting AI do optional scutwork that falls into the “nice to have” category (e.g., creating archival information).

On a personal level, I have used AI to automate a lot of required scutwork while freeing up my time to engage in higher added-value tasks. In terms of time, the biggest areas have been preliminary research, summaries, and writing drafts.

Additionally, in one specific use case that I can talk about, I have helped a medical billing office prioritize and organize their work based on estimated hourly value of items processed as well as difficulty (difficult stuff was prioritized for certain times of the day). This work had been eye-balled in the past, and could be done with a decent degree of accuracy with quite a bit of time, but AI did it with higher accuracy and almost no dedicated human time. The whole office appreciated the outcomes.

There are many wins like this that are available with AI, and I think many folks just haven’t found them yet.