An ontology a reasoning model can actually ingest
An agentic-platform readiness assessment over a large corpus of reported phishing, and the ontology it depended on — entities and relations modelled from…
The decision
Can an agentic platform reason over this corpus, and what has to be true first?
The problem
Years of submissions from many contributors, each with its own schema, its own confidence conventions and its own idea of what counts as a report — with heavy duplication, because one campaign gets reported hundreds of times within an hour. Answering only what a model can do today produces a demo; answering only what the data needs produces a project with no destination.
Outcome
A documented ontology covering the entities and relations the corpus supports, an extraction and entity-resolution path onto it, and a staged readiness assessment that named the unanswerable questions rather than deferring them.