The knowledge retention glossary

Knowledge retention is the practice of capturing and preserving the expertise, context, and decisions employees build, so an organisation keeps that knowledge when people leave or change roles.

Corporate amnesia is what happens when an organisation repeatedly loses and relearns knowledge because the reasoning behind past work leaves with the people who did it.

Institutional knowledge is the collective, often undocumented know-how a company accumulates over time: how things are done, why decisions were made, and who knows what.

A knowledge digital twin is an AI model of a role or person, trained on their real work, that teammates can question after that person leaves or changes roles.

Knowledge management is the discipline of creating, organising, sharing, and maintaining an organisation's knowledge so the right information reaches the right people.

Enterprise AI readiness is how prepared an organisation's knowledge and data are for AI: well-governed, permissioned, and accessible enough for AI tools to give accurate, safe answers.

Retrieval-augmented generation (RAG) is a technique where an AI model fetches relevant information from a knowledge source at query time and uses it to ground its answer.

The Model Context Protocol (MCP) is an open standard that lets AI assistants and agents connect to external tools and data sources through a common interface.

Knowledge coverage is the measured state of what an organisation actually knows about its critical work: what has been verified, what needs re-verification, what criteria were never recorded, what is contradictory, what has no owner, and what only one person knows.

A knowledge gap is a piece of knowledge that critical work depends on but that the organisation has never recorded, verified, or assigned an owner to.

A verification loop is the governed path that turns an identified knowledge gap into an approved, source-traceable answer: identify the gap, name the verifier, ask one short contextual question, record the human decision.

A governed AI agent is an AI agent that uses an organisation-approved model and operates only with permitted data, tools, identities, and approval policies, where high-risk steps still require human approval.

Operating memory is the governed layer of organisational context that both people and approved AI agents may draw on: decisions, rationale, exceptions, and approved source material, each permission-aware and traceable to its source.

Tribal knowledge is know-how that exists only inside a team's shared experience rather than in any document: the workarounds, warnings, and reasons behind how work actually gets done.

Tacit knowledge is expertise a person applies without being able to fully explain it: judgement, pattern recognition, and instinct built from experience rather than instruction.

Knowledge transfer is the deliberate handing over of expertise, context, and responsibility from one person or team to another, usually when someone leaves, changes role, or a project moves hands.

A knowledge transfer plan is a written schedule of what expertise must move from one person to another before a set date, naming each item, who receives it, and how the transfer is confirmed.

Employee offboarding is the process of closing out someone's employment: revoking access, recovering equipment, settling obligations, and transferring their knowledge and responsibilities before their last day.

A knowledge audit is a structured review of what an organisation knows, where that knowledge lives, who holds it, and where it is at risk of being lost.

A single source of truth is one designated place where a given fact is maintained, so that every person and system reads the same authoritative value rather than a copy of it.

Enterprise search is search across an organisation's internal systems, returning results from documents, tickets, messages, and applications while respecting each user's existing permissions.

A knowledge base is a structured, searchable collection of an organisation's documented information, used internally by staff or externally by customers.

Permission-aware retrieval is retrieval that applies each user's existing access rights at query time, so an AI answer can only be built from sources that user was already allowed to open.

An AI hallucination is a response that is fluent and confident but not supported by the model's sources or by fact, produced because the system is predicting plausible text rather than looking up a verified answer.

Human in the loop is a design in which an automated or AI-driven process pauses for a person to review, approve, or correct an action before it takes effect.

AI governance is the set of policies, controls, and accountabilities that determine how an organisation builds, buys, and uses AI: which systems are approved, what data they may use, who is answerable for outcomes, and how use is evidenced.

An AI governance framework is a structured model for managing AI risk and accountability, setting out the principles, roles, risk tiers, and controls an organisation applies across the life of an AI system.

AI agent governance is the control of AI agents that take actions rather than only produce text: which identity an agent acts under, which tools and data it may use, which steps require human approval, and what record it leaves behind.

Agentic AI describes AI systems that pursue a goal over multiple steps, choosing actions and calling tools along the way, rather than returning a single response to a single prompt.