AI proposes. People approve. Knowledge keeps itself honest.
AI proposes a change. A person approves it. Every approved change becomes part of a readable, auditable history. Rules then check what's complete, what's stale, and what still holds up.
- 01Source inputchat, doc, ticket, call
- 02Proposed changeAI drafts memory diff
- 03Human reviewapprove, edit, reject
- 04Canonical memoryhuman-readable, versioned
- 05Indexes & answerscited, permission-aware
Corrections and new sources re-enter as proposed changes. The loop is the product.
Canonical memory is readable, rule-checked, and fully auditable.
Every approved change is a signed, versioned update. You can read it, compare it, roll it back, or export the whole thing — nothing is locked inside WiseWare. AI never writes to canonical memory on its own; a human approves first, and the rule checks then run against what was approved.
For the technical reader: canonical memory is plain markdown files in a regular git repository. Rules written in CEL check those records for things like stale evidence, missing obligations, and overdue reviews.
- Source A transcript, chat, ticket, doc, or note is accepted as input.
- Proposal WiseWare drafts a change against existing memory objects.
- Review A human approves, edits, or rejects — with the source side-by-side.
- Rules Approved records are evaluated for gaps, stale state, conflicts, and required review.
- Canonical Approved changes land as a new version. Indexes, links, and health checks rebuild.
- Answer Questions get cited answers from current memory — or a flagged caveat.
@Quarterly access review · CTL-12
owner: Security
cadence: Quarterly
−last reviewed: 2026-01-15
+last reviewed: 2026-04-03
+reviewer: CTO
+evidence: Q2-Access-Review.pdf
state: current
Sources
+— Jira AC-128 (signed off 2026-04-03)
+— Slack #security · thread 2026-04-03
A proposed change from a closed ticket. One reviewer click turns it into canonical memory.
WiseWare does not replace the systems where work happens.
Your CRM, ticket tracker, and GRC tools are built for getting work done. They're not built to explain it. WiseWare sits alongside them and keeps a clean record of what's true, what was decided, and what evidence backs it up — across all of those systems.
AI can't act reliably on knowledge buried inside workflows and screens. It needs something it can actually read.