This is not
a white paper.
We are publishing our observations before our conclusions.
An ongoing investigation into what actually changes when AI enters organizational work.
We expected AI adoption to be a technology problem.
We kept finding something else.
Different
organizations.
Over several weeks, conversations across an AI community and practitioners involved in implementation repeatedly returned to similar problems.
Different work.
Similar tensions.
AI capability
is improving.
Output ↑ Review ↑
Trust ? Ownership ?
Capability can scale faster than an organization’s ability to decide what should happen next.
The technology questions
The work questions
The bottleneck may already have moved.
As execution becomes cheaper, choosing, evaluating, trusting and intervening may become more consequential.
Organizations may not struggle because AI is unavailable.
They may struggle because the judgments surrounding AI remain largely invisible.
Working hypothesis — to be tested, not assumed.
Judgment is work.
But where are these decisions represented in the systems we are trying to automate?
Bruno Latour entered the laboratory before theorizing how scientific facts were produced.
Before explaining science,
he observed science.
Before automating judgment,
should we first observe judgment?
The unit is not the person.
It is the decision episode.
Observation is not the endpoint.
Can something that matters to expert judgment become something a system can see, evaluate or act on?
This is an ongoing investigation.
I’m looking for organizations already deploying AI and willing to examine what is actually happening inside the work.
Research collaboration · field inquiry · organizational experiments
Ryan Son
Seoul, Korea