Knowledge Base manages four complementary bodies of content under one governance model. The glossary holds business terms and their authoritative definitions, organised by domain so that finance, credit risk, accounting, standards, tax, supply chain, manufacturing, sales, human resources and technology each own a coherent vocabulary. Manuals carry long-form procedural guidance, decomposed into an ordered hierarchy of sections and subsections that mirror how a real handbook is structured. FAQs capture question and answer pairs grouped into categories such as getting started, accounting and control, financial services, business relations, catalogue management, supply chain and system administration. How-to articles carry step-by-step task guidance for recurring operations.

Every item is authored in a rich editor, categorised, tagged, versioned and searchable through full-text and faceted search that spans all content types at once. Authors work through a controlled lifecycle from draft to review to published to retired, with each transition attributed and time-stamped. Content can be exported in full, cross-referenced between items, and linked outward to the records and screens it explains. The result is a curated corpus rather than a document dump: current, attributable and fit to be quoted verbatim to a customer or an auditor.

The model is best understood as a small number of content shapes bound together by category, version and ownership. A knowledge item is the common unit: a defined term, a manual section, a frequently asked question or a how-to article. Each item carries a title, a body, a state in its lifecycle and an owner accountable for its accuracy. Items never float free. Every one is filed under a category, and categories form the navigable taxonomy that lets a reader move from a broad domain down to the specific answer they need.

Manuals introduce structure of their own. A manual is a parent that gathers ordered sections, and sections may themselves contain sections, so a handbook reads as a coherent hierarchy from cover to fine detail rather than a flat list of pages. Cutting across all of this is version history: each published change preserves the prior wording, who altered it, when and why, so the meaning of a term or the wording of a procedure can be reconstructed as it stood at any past date. Because relationships are explicit, a single glossary definition can be referenced from many manuals and answers, and updating it once corrects every place that leans on it. The concepts stay deliberately few, which is what keeps the corpus easy to browse, to govern and to trust.

Authoring follows a governed path. A contributor drafts an item, assigns its category and tags, and submits it for review. A steward with authority over that domain checks accuracy, completeness and tone, then publishes it or returns it with comment. Publication makes the item live to every consumer at once, and retiring an item withdraws it without destroying its history. Because each step is attributed, the module answers not only what the current guidance is but who approved it and when.

Consumption is the higher-volume workflow and is deliberately frictionless. A reader searches across the whole corpus or narrows by category, opens the matching term, section or answer, and follows cross-references to related material. Service agents pull answers straight into a ticket response, portal users resolve their own questions without raising a case, and analysts export a category for offline review or distribution. A third workflow is maintenance: scheduled review cycles flag items that have gone stale, owners reconcile definitions that have drifted apart, and gaps revealed by unmatched searches or repeated questions are turned into new content, so the corpus stays current by design rather than by accident.

The value of a knowledge base lies in governance and precision, and this module treats both as first-class. Every mutation, whether a new definition, an edited section or a retired answer, is written to a complete audit trail recording the action, the affected item, the author and the detail of the change, so the corpus is defensible under review and its evolution is fully reconstructable. Access is enforced by fine-grained permissions, so authoring, reviewing and publishing rights are held only by those entitled to them, while read access can be opened as widely as the organisation wishes.

Search is engineered for recall and ranking across mixed content, returning a glossary term, a manual passage and an FAQ in one weighted result set and letting the reader disambiguate by category. Versioning is genuine rather than cosmetic: prior revisions are retained in full and can be compared, so the exact wording in force at a point in time can be produced on demand, which matters wherever definitions carry regulatory or contractual weight. Categorisation is consistent across all content types, giving a single taxonomy that the service desk, the portal and the AI Workers all navigate the same way. Content is validated on entry and parsed safely, and errors are logged, so the corpus that consumers depend on stays clean, well-formed and reliable.

Knowledge Base is infrastructure for the rest of the suite rather than an island. The Service Desk is its most direct consumer: agents resolve incidents and requests against published answers and manual sections, and unresolved questions feed back as candidate content. The Customer Portal exposes a curated, permission-filtered slice of the corpus for self-service, deflecting routine enquiries before they become cases. The glossary underpins consistency across the operational modules, giving finance, accounting, credit risk, catalogue management and supply chain a shared vocabulary so that a term means the same thing in a report as it does in a procedure.

Manuals document how those modules are operated, from accounting close routines to financial services handling and catalogue governance, keeping procedure and system in step. Administration and identity determine who may author, review and read, so knowledge governance aligns with the organisation's wider access model. Because categorisation and ownership are shared conventions, the Knowledge Base can annotate and explain records held elsewhere in the suite without duplicating them, turning scattered institutional memory into a single dependable layer that every other module can quote.

AI Workers are first-class users of the Knowledge Base and treat it as their primary ground truth. On the query side, a Worker answers a natural-language question by retrieving and citing the exact glossary term, manual section or FAQ that applies, so its response carries a traceable source rather than an unattributed assertion. This grounding runs throughout the suite: when a Worker explains a figure, drafts a portal reply or advises an agent, it quotes governed content and links back to it.

Workers also act on the corpus. They execute authoring actions under the same permissions and audit trail as any human, drafting new answers from resolved tickets, proposing definitions for terms that appear across records without one, and extracting structured content from uploaded documents to seed manuals and FAQs. They watch for exceptions: contradictory definitions, sections that have gone stale against a changed policy, high-frequency searches that return nothing, and they raise these as alerts to the responsible steward. In decision support they surface the relevant procedure at the moment a task is performed elsewhere in the suite. And they participate directly in governance, standing as a review or approval node in the authoring workflow, checking a draft for consistency with existing content and either endorsing it or returning it with specific comment before a human steward gives final sign-off.