Teaching Agents That Where You Sit Shapes How You Think
From the lab to pilot plants and commercialization, the same project needs different criteria at each stage. Could agents learn to recognize those shifts in decision-making too?
Something clicked for me today—
When the decision-making framework changes, the conditions and criteria for judging the same thing need to change with it.
I thought about it and realized it would take a bit of explaining, so I’m putting down this note for now. I’ll find time to write about it properly another day.
But I found an amusing analogy that comes close enough to help me remember it: the Chinese saying “屁股决定脑袋”—roughly, where you sit shapes how you think. Ha~
It happens across people, departments, situations, stages of a process, and all sorts of other contexts. If this is an unspoken “rule” in the human world, it should apply to agents too.
Then I thought of a particularly good example. Back when we took a technology out of the lab and toward industrial use, it would generally pass through several stages—
The laboratory (lab + pilot scale)
--> Pilot plant
--> Demonstration project (sometimes included in the pilot-plant stage)
--> Industrialization / commercialization (commercial plant, large-scale production)
From start to finish, it is still the same undertaking or project. But the inputs, outputs, and conditions used to make decisions differ greatly at each stage: from verifying scientific feasibility, to engineering feasibility, to commercial feasibility. Each step has a different “decision-making framework.”
Thinking about it this way, I find the hypothesis—or perhaps the idea—quite interesting: let the Agentforce go through the best practices that human ingenuity has worked out too. And we are probably only at the starting point.
Notes taken during the transition from Workforce to Agentforce — and we're in between.