Platform
One governed home for what your organization knows.
Knowledge is captured from where work happens, approved by a named person, stored as a versioned record, and served wherever it is needed.
The Cognatum Knowledge Loop. Eight steps on a closed loop, in four phases. Intake: 01 Capture, 02 Structure. Refine: 03 Clean. Elevate: 04 Enrich, 05 Improve. Govern & activate: 06 Approve, 07 Integrate/Deploy, 08 Reuse. Step 08 returns to step 01, around Knowledge base.
Knowledge base
Governed, for people and AI
01 · Intake
Capture
Ingest from every source.
02 · Intake
Structure
Normalize the format.
03 · Refine
Clean
Update, deduplicate, reconcile, archive.
04 · Elevate
Enrich
Add context and tags.
05 · Elevate
Improve
Refine clarity, usability.
06 · Govern & activate
Approve
Human sign-off, audited.
A person handles this step
07 · Govern & activate
Integrate/
Push live everywhere.
08 · Govern & activate
Reuse
Serve people and AI.
→ back to 01
Intake
- 01 Capture Ingest from every source.
- 02 Structure Normalize the format.
Refine
- 03 Clean Update, deduplicate, reconcile, archive.
Elevate
- 04 Enrich Add context and tags.
- 05 Improve Refine clarity, usability.
Govern & activate
- 06 Approve Human sign-off, audited. A person handles this step
- 07 Integrate/Deploy Push live everywhere.
- 08 Reuse Serve people and AI.
The Cognatum Loop is Intelligence Management in an AI powered world. Leveraging ISO 30401 standards and knowledge management best practices over the last 30+ years, Cognatum was built to support the most effective way to manage company knowledge across systems and teams.
Ask
Ask, and get the approved answer.
A question in plain words. An answer that names its source, its version, and the person who approved it.
Every answer is drawn from entries a named person has signed off on, and it says so on the answer itself: the source document and its version, who approved it, and how long ago. One approved source is cited, not a list of files that might be relevant.
Use disclosure D-214 rev 6. It supersedes rev 5 for all retail communications and carries the current market-risk language. SOP-4471 v3
Knowledge
Every entry, with its state.
The library is a list of records, not a pile of files. Each entry carries a type, a state, an expiry and a reach.
Answers, decisions, procedures and notes sit in one place under the same six columns. The state says whether an entry is approved, in review, stale, draft or archived. The expiry says when it is due another look. The reach says how many answers cite it today. The whole list exports when a reviewer asks for it.
| Select | Title | Type | Taxonomy | Status | Expires | Reach |
|---|---|---|---|---|---|---|
| Which retention period applies to client trade communications? | Q&A | Records | Approved | in 214 days | 38 | |
| Standardize retention schedules on the records policy | Decision | Compliance | Approved | in 149 days | 31 | |
| Applying a legal hold to a retention schedule | Procedure | Legal | Approved | in 24 days | 27 | |
| How long are supervisory review records kept? | Q&A | Supervision | In review | in 96 days | 18 | |
| Which regions are approved for data residency? | Q&A | Security | Stale | Expired 6 days ago | 12 | |
| Retention questions raised in the Q3 examination | Note | Compliance | Draft | No expiry | 4 |
Approvals
A person signs. It is recorded.
Nothing goes live on its own. The queue routes each entry to someone with the authority to approve it.
An answer drafted from a source, a procedure whose new version supersedes the old, an entry whose source changed and needs a second look: each one waits for a named person. On approval the entry is sealed to a version, and the record shows who approved it and when.
Published as the approved answer.
Approver: you · just now · sha256 c4d1…b90
Health
What is wearing out, and why.
One score for the whole knowledge base, and a table of the entries pulling it down.
Every entry sits in a band: healthy, at risk, critical or stale. The table shows its score, the direction it moved since the last check, how many answers rely on it, and the reason it sits where it does: a source changed, no owner confirmed, awaiting approval, one open conflict. Recompute whenever you want a current picture.
Corpus health
71
| Item | Type | Health | Reach | Band | Why |
|---|---|---|---|---|---|
| Applying a legal hold to a retention schedule | Procedure | 74 ↗ was 67 | 38 answers | Healthy | cited by 38 answers · reviewed this quarter |
| Which regions are approved for data residency? | Answer | 52 ↘ was 61 | 24 answers | At risk | source changed · no owner confirmed |
| Standardize retention schedules on the records policy | Decision | 68 → was 68 | 17 answers | Healthy | stable since approval · owner active |
| How long are supervisory review records kept? | Answer | 59 ↗ was 51 | 12 answers | At risk | awaiting approval · one open conflict |
| Retention questions raised in the Q3 examination | Note | 31 ↘ was 44 | 6 answers | Critical | no approved source · superseded twice |
Gaps
The questions nobody has answered.
Every question asked with no approved record behind it is counted and ranked.
The list shows what people and assistants keep asking that the library cannot yet answer, and how often each was asked. It is the work queue for the next entries: the gaps that matter most sit at the top, and closing one is a drafting task with a named approver at the end of it.
A question nobody has an approved answer to is not a miss that disappears into a log. It becomes a recorded gap with an owner, routed back to Capture, which is how the library learns what it does not yet cover.
Approved records
318
In review
41
Flagged stale
27
Open gaps
37
- Which regions are approved for data residency? 24
- Who approves a retention exception over seven years? 19
- What evidence do we keep for a supervisory review? 17
- Which model versions are approved for regulated answers? 14
- How is a superseded procedure withdrawn from every channel? 11
Find SMEs
Who may review, who may approve.
Authority is a fact about a person, and the directory shows it beside their name.
Each expert is listed with a department, a primary area, the hours they give the library each week, how many items are in their queue, and whether they can review, approve, or neither. Routing an entry to the right person is a lookup, not a guess.
| SME | Department | Primary Expertise | Availability | In queue | Authority |
|---|---|---|---|---|---|
| D. Whitfield Head of Records Management | Records Management | Records retention and legal hold | Same Business Day 6h / wk | 3 | Can review Can approve |
| M. Alvarez Compliance Director | Compliance | Supervisory procedures and examinations | Within 2 Days 4h / wk | 0 | Can review Can approve |
| J. Park Senior Counsel | Legal | Data residency and cross-border transfer | Same Week 3h / wk | 1 | Can review |
| R. Okonjo Quality Systems Lead | Quality | Change control and deviation handling | Within 2 Days 5h / wk | 0 | Can review |
| T. Lindqvist Platform Operations Engineer | Operations | Connector configuration and access roles | Same Week 2h / wk | 0 | No review authority |
Dreaming
Links the graph proposes.
Between approvals, the system looks for entries that belong together and says why it thinks so.
A proposed link names the two entries, the relation, a confidence, and the signals behind it: how often the two are cited together, how close their wording sits. A person promotes the link or rejects it. Nothing links itself.
walk_hits=2, co_cited=3, embedding_cos=0.87
When nothing matches
It says when nothing matches.
Related approved knowledge was found, but no entry was a confident match. So the answer says that, and shows its working.
The banner tells the reader to check the cited sources before relying on the answer. A retrieval strength sits beside it, with the ten source scores it is the mean of, and a caveat that similarity is not correctness. A governed answer is worth having only if the system can also decline to answer.
Based on similarity of matched sources, not a correctness guarantee
Mean of 10 retrieved source scores
- 75%
- 60%
- 61%
- 56%
- 52%
- 35%
- 32%
- 41%
- 35%
- 24%
Go deeper
Five ways in.
Each page below is part of one system: a store that holds the approved version itself, rather than pointing at a copy of it somewhere else. Start wherever your question is.
Start here
How it works
The whole argument end to end, from the question a person asks to the record that answers it.
Explore →The mechanism
The loop
The eight steps that run continuously: AI captures and prepares knowledge, a person approves it, and it is served everywhere.
Explore →The architecture
System of record
Why storing the approved entry is different from pointing at content wherever it happens to live.
Explore →Connections
Integrations
How knowledge gets in from the systems where work happens, and out to the tools that ask for it.
Explore →Evidence
Standards
The frameworks a regulated organization answers to, what each asks of you, and what Cognatum supplies toward it.
Explore →Knowledge governed. Intelligence everywhere.
See it on your own content, in your own environment.