What an organization actually knows about itself, where AI fits, and how to tell the difference.
Most organizations don't have an independent, current artifact that states how they actually work: what the rules are, what terms mean across teams, who's accountable for what. That knowledge exists, in code, in people's heads, in institutional memory that contradicts itself, but nowhere anyone can simply point to. AI adoption is making that absence harder to ignore, though it isn't a new problem and AI didn't cause it.
There is another problem I think is being overlooked: understanding what an organization actually is at its essential level, the network of actors, acts, and commitments that would persist even if every piece of software supporting it were replaced. Redesigning for or with AI without that understanding tends to produce systems that automate a step faithfully while missing what the organization needed in the first place.
This site is a personal effort to work through that gap: how a foundation of organizational knowledge could actually be made explicit, current, and verifiable, for environments where correctness, auditability, and accountability matter.
The synthesis draws on two traditions that rarely talk to each other: philosophy of meaning, and formal data, conceptual, and enterprise modeling. See my key influences.
The same term can mean incompatible things across teams, or two different terms can quietly mean the same thing, and a dashboard's semantic layer solves neither. It's a different problem than the one that breaks real-time operations.
Under the closed world assumption, an unrecorded fact is false, not merely unknown. When money, rights, or compliance are at stake, "I don't know" isn't a good enough answer.
Authority and accountability attach to an organizational role, not to whoever performs the act. Swap a person for an AI system and the role's obligations don't move, a distinction most "AI accountability" writing misses entirely.
A rule written down and never resynchronized isn't a smaller version of the truth. It's a historical record that keeps looking authoritative long after it's stopped being one, which is worse than having no documentation at all.
These essays explore organizational design, AI governance, semantic modeling, and the practical challenge of making enterprise data clear enough to support accountable systems. They build loosely on each other, from semantic foundations, to the automation boundary, to why organizational analysis has to come first, to the layered nature of meaning in a database, but each stands on its own.
A working Object-Role Modeling tool with AI-assisted model creation and validation. It is an applied demonstration of using formal conceptual structures to ground AI query and agent systems in explicit data and policy semantics.
The names below come from two areas that are usually discussed separately: philosophy of meaning and formal approaches to data, conceptual, and enterprise modeling. I am interested in what happens when ideas from both are applied to enterprise AI and data systems.
A formal model does not settle meaning once and for all. It records the distinctions and constraints a community has agreed to work with, making them available for examination and testing. The model remains useful only while those commitments continue to match practice, and it has to change when practice changes.
If you are working through similar questions, especially around data platform design, semantic modeling, AI governance, or accountable automation, feel free to send me a note. This site is a personal side project, so I may not always reply quickly, but I am glad to hear from people exploring related ideas.
If any of the influences above are ones you keep coming back to as well, mention that in your note. I like hearing from people converging on similar territory from different directions.
Email: gsawatzky@knowledge-foundation.ai
I also keep a private Collaboration Zone with notes and essays shared only with people I'm actively working with. It's invite-only: sign in there with Google to request access, or just email me directly and I'll set you up.