In preparation for a keynote talk tomorrow morning, I again tested the conference website:
- 5 min attended - asking for what I wanted done from agents, and reproducing most interesting problems found
- 25 minutes unattended - working with a paid client while the testing was being done
Every single time I do this, I stop to think about what do we really do with the time when AI runs a job for us.
The time isn't saved. It's used on learning. Helping others. Thinking about next steps. Starting next steps. Deciding on criteria of calling things done when recognizing done is at the center of actions. We do more, and new kinds of things.
The conditions for client(s) actually saving time with AI need a whole lot more changes in the way the work is organized.
