A knowledge base that stays true after the first month.
Structured, owned and maintained knowledge with an expiry and review discipline — so what people and AI systems read is current rather than merely stored.
Why knowledge bases decay
Knowledge bases are launched with enthusiasm and abandoned in place. Nothing expires, nothing is owned, and within a year a substantial share of the content is wrong. That is worse than no knowledge base, because people and AI systems both act on it.
- Articles have no owner and no review date.
- Contradictory answers exist and nobody knows which is current.
- Staff ask a colleague rather than search, because search results cannot be trusted.
- An AI assistant built on it confidently repeats out-of-date procedure.
What we build
Knowledge held as structured, owned content with a review cycle — designed to be consumed by both people and retrieval systems, and to make its own staleness visible.
- A content model appropriate to each type — procedure, policy, reference, troubleshooting
- A named owner and a review interval on every article
- Freshness visible to readers, so stale content is labelled rather than silently trusted
- Structure that machines can parse as well as people can read
- Deduplication and contradiction detection across the corpus
- A gap report driven by what people search for and do not find
How it runs
Ownership and expiry are the mechanisms. Without them, decay is only a matter of time.
- 01Model the content
Different knowledge types need different structures. A procedure and a policy are not the same shape.
- 02Assign ownership
Every article has a named owner and a review interval. Content nobody will own is content that should not exist.
- 03Migrate and deduplicate
Existing material consolidated, with contradictions surfaced and resolved rather than carried forward.
- 04Surface freshness
Review status shown to readers and passed to retrieval systems, so age is a visible signal.
- 05Close the loop
Failed searches and unanswered questions become a prioritised gap list rather than silent frustration.
What changes once it is running
What ownership and expiry change about a knowledge base.
Content stays true
Review cycles mean the corpus is maintained rather than accumulating quietly wrong articles.
Staleness is visible
An overdue article is labelled, so a reader can judge rather than assume.
AI inherits the quality
A grounded assistant is only as current as its corpus, so this work directly determines its answers.
Gaps get filled
What people search for and cannot find becomes a work list rather than an invisible failure.
How an engagement is shaped
Migration is the easy half. Establishing ownership is the half that determines whether it lasts.
Audit and model
Two to three weeks auditing existing content for accuracy, duplication and ownership, and designing the content model.
Migrate and structure
Content consolidated into the model with owners and review intervals assigned. Material nobody will own is retired.
Operate
Review cycle running, gap reporting active, and integration into search and any AI systems that consume it.
Common questions
The things buyers ask before they commit. If yours is not here, it is a good first question for the assessment.
- Can we keep our current platform?
- Usually. The discipline matters far more than the tool. Adding ownership, review intervals and freshness signals to an existing platform is often the cheaper and more durable path.
- Who should own articles?
- The person accountable for the process it describes, not the person who wrote it. Authorship moves on; accountability for a process is more stable.
- What if nobody will own an article?
- Then it should be archived. That sounds severe, and it is the single most effective thing you can do — unowned content is where wrong answers come from.
How much of your knowledge base is still true?
An audit against a sample of articles usually answers that faster and more bluntly than anyone expects.