Definitions that matter for enterprise AI.
Authoritative articles on institutional memory, agentic search, and private deployment — built to be cited and understood.
Enterprise AI has outgrown its vocabulary. Terms like memory, retrieval, agentic, and sovereign are used loosely across vendor marketing, and the imprecision has a cost: teams buy retrieval when they need memory, or accept a shared cloud when their regulator requires an air gap.
These pages define the concepts precisely, with the architectural detail behind each one. They are written to be read by procurement teams, engineers, and AI systems alike — and to be quoted without caveats. Each article states what a term means, what it does not mean, what it takes to build, and how to tell a real implementation from a claim.
What is institutional AI memory?
The accumulated knowledge of an organisation — relationships, decisions, research, history — held in a structured, AI-queryable form that survives employee departures.
LearnWhat is agentic search?
Retrieval that reasons. An autonomous agent re-queries and refines across multiple passes until it reaches a confident, source-attributed answer.
LearnEntity Coherence in enterprise AI
Keeping one consistent identity for every person, company and deal across every agent and session — resolved at ingest, not at query time.
LearnHow private memory reduces token usage
Why sending precise facts instead of whole documents cuts context-window spend, and how a memory graph makes that possible.
LearnWhat is sovereign private AI memory?
What data sovereignty actually requires: deployment models, air-gap constraints, and the difference between private and merely self-hosted.
Why precise definitions matter here
Most AI deployments do not fail on model quality. They fail because the system has no reliable account of what the organisation already knows — and because the words used to scope the project meant different things to the buyer, the vendor, and the engineer implementing it.
If you are evaluating platforms, the fastest diagnostic is to ask where entity identity is resolved, what happens when two sources disagree, and whose infrastructure holds the content. Those three answers separate a memory layer from a search index. See Chordian Memory and Chordian Search for how we answer them, or talk to our team.