Institutional Memory
Document type: Arqua Public Architecture Concept
Status: Current public explanatory reference
Canonical parent: Enterprise Intelligence Architecture
Canonical vocabulary source: Enterprise Intelligence Architecture — Canonical Definitions
Canonical definition
Institutional Memory is the governed continuity of accepted institutional understanding, evidence, decisions, relationships, states and outcomes preserved across time, organisational change and technology evolution.
The defining characteristic of Institutional Memory is continuity, not storage.
Purpose
Institutional Memory exists to prevent the institution from repeatedly losing, fragmenting or silently rewriting what it knows.
It preserves accepted institutional understanding together with the evidence, decisions, relationships, states, outcomes and revisions needed to make that understanding durable and reconstructable over time.
Institutional Memory is not merely a place where information is stored. Its architectural role is continuity: preserving what the institution has accepted, how that understanding changed, and the evidence and outcomes associated with those changes.
Standing boundary
Accepted Representation, Architecture of Record and Institutional Memory are related but not equivalent.
- Accepted Representation is a representation that has passed the relevant institutional processes of interpretation, evidence assessment, semantic governance, authority and acceptance for a declared scope and purpose.
- Architecture of Record records the standing, applicability, implementation obligations and consequence-relevant boundaries of accepted architecture and representations.
- Institutional Memory preserves accepted institutional understanding and its supporting evidence, decisions, states, outcomes and revisions across time.
Governed Evaluation / Acceptance
↓
Accepted Representation
├────→ Architecture of Record
│ standing + applicability
│ obligations + consequence boundaries
│
└────→ Institutional Memory
continuity + evidence
decisions + outcomes + revisionArchitecture of Record is therefore not Institutional Memory. Institutional Memory is also not an acceptance engine and does not by itself create institutional authority.
Inputs
Institutional Memory may preserve:
- accepted institutional representations;
- provenance and supporting evidence;
- decisions and decision context;
- relevant relationships and states;
- outcomes;
- revisions and supersession history.
Not every enterprise signal, document, model output or transient observation should silently acquire the standing of institutional memory. Material preserved as Institutional Memory must carry the standing required by the applicable governance model.
A governed record may preserve the fact that an assertion, objection or hypothesis was made without accepting its contents as institutional understanding. Its source, status and relationship to accepted understanding must remain explicit. The standing of the preserved record is not a blanket endorsement of every claim it contains, the correctness of a recorded decision or permission to use that decision again.
Outputs
Institutional Memory provides durable and reconstructable institutional understanding that can support later interpretation, governance, publication and runtime-context formation.
It preserves continuity without implying that every preserved item is current, true for all purposes or automatically suitable for runtime reliance.
Historical reconstruction
Reconstruction distinguishes the representation and evidence preserved for an earlier context from later corrections and interpretations. Evidence available to a process is not necessarily evidence actually used by a decision-maker; claims about use require their own supporting records. A later explanation must be identified as a later interpretation, not substituted for the original rationale.
When the preserved evidence is incomplete, that limitation must be explicit. Historical reconstruction is not a claim of complete replay, an instruction to re-execute a past action or renewal of an earlier permission. Access to retained information and reliance for a new purpose remain governed by the applicable current conditions.
Relationship to Enterprise Representation Intelligence
Enterprise Representation Intelligence forms and evolves governed enterprise representations and supports their evaluation and acceptance.
Institutional Memory preserves representations that have acquired the required standing, together with the evidence, decisions, outcomes and revisions needed for continuity.
Outcomes can return as evidence. Evidence can challenge existing representation. Governance determines what is revised and reaccepted. Institutional Memory preserves that evolution rather than overwriting prior standing without trace.
Relationship to Runtime Context Assembly
Runtime Context Assembly is not Institutional Memory.
Institutional Memory is persistent continuity across time. Runtime Context Assembly forms and qualifies purpose-specific operational understanding for a particular Operational Intent. Relevant memory projections, accepted representations, current evidence and governed capabilities participate through Published Enterprise Products, subject to scope, authority, policy, semantics and permitted use.
This view shows one memory-consumption path. Not every Published Enterprise Product originates in Institutional Memory: other accepted representations and governed capabilities may acquire publication standing separately. Preservation alone does not grant publication standing or suitability for the current intent.
Memory provides continuity; RCA provides situation-specific formation and qualification. Runtime context does not silently become accepted Institutional Memory.
Relationship to Enterprise Product Publication
Accepted institutional representation may be published as a governed Enterprise Product according to its standing, publication authority and permitted use.
Publication is a governed participation mechanism. Institutional Memory itself is not equivalent to a Published Enterprise Product. A bounded projection of Institutional Memory may be exposed as an Enterprise Memory Product when publication standing and purpose are explicit.
Relationship to AI
AI may consume projections of Institutional Memory, current evidence and other governed context, but Institutional Memory is not an AI memory feature, vector store or model context window.
AI does not independently create institutional truth or acceptance merely by generating, retrieving or retaining information. Institutional standing remains governed and accountable to the enterprise.
What Institutional Memory is not
Institutional Memory is not, in isolation:
- a knowledge base;
- document repository;
- graph database;
- vector store;
- data lake;
- archive;
- model context window;
- Architecture of Record;
- Runtime Context Assembly;
- a substitute for institutional governance or acceptance.
Technology may implement or support Institutional Memory, but technology does not define the architectural responsibility.
Continuity principle
Institutional Memory preserves what the institution has accepted and how that understanding has evolved. It does not convert storage into authority.
The preservation of prior states, revisions, evidence and outcomes allows institutional understanding to remain reconstructable as organisations, systems and technologies change.
Related architecture
- Enterprise Intelligence Architecture
- Enterprise Representation Intelligence
- Architecture of Record (AoR)
- Runtime Context Assembly
- Enterprise Intelligence Architecture — Canonical Definitions
Standing
This page is the current dedicated public explanatory reference for Institutional Memory. It does not replace the canonical vocabulary entry or a future formal architecture paper. Material changes to the canonical definition, scope, parentage or relationships require controlled architecture reconciliation.
Publication provenance — website reconciliation history
These notes document website-level clarification. The canonical definition and authoritative source standing are unchanged.
Web reconciliation: 11 September 2026 — P0-2I-B v1.0. Runtime-consumption depiction aligned with governed publication and RCA formation plus qualification; the canonical memory definition is unchanged.
Further web clarification: 17 September 2026 — Record standing and embedded claims, historical reconstruction, evidence available versus actually used, and incomplete-history limits clarified. An at-a-glance explanation is added. The canonical definition, current public explanatory standing and preceding reconciliation history are retained.