AI can only be as useful as the knowledge it has to work with. Yet most organisations have no editorial system for deciding what knowledge can be trusted, what has changed and what should be discarded. The Editorial Operating System is the missing governance layer between organisational knowledge and AI.
Potentially, every organisation developing an AI strategy has a governance problem.
It may already have policies governing privacy, security, ethics, risk and the use of AI. It may have an AI committee, responsible-AI principles and rules governing which employees can use which tools.
But there is a more fundamental question:
Does your own organisation actually know what of its knowledge can be trusted?
Most organisations have thousands of documents, policies, procedures, databases, presentations and webpages. But which version is authoritative? Which is current? Who is responsible for it? What evidence supports it? And, what should an AI system do when two parts of the organisation appear to say different things?
You can spend heavily on AI and still fail to make your organisation’s own knowledge work for you. Worse, you can undertake an AI-readiness project and still do it badly if you treat knowledge as a technical data problem rather than something requiring editorial governance.
This is because AI can simply retrieve, combine and interpret this knowledge without necessarily knowing whether it is authoritative, current or internally consistent.
And that is the missing layer in many AI strategies: the governance of organisational knowledge itself.
And that changes the strategic question.
It is no longer simply:
How should we govern our use of AI?
It is also:
How should we govern the knowledge that AI uses?
That is editorial governance. And, because this is a new, emerging challenge, this is where AI’s greatest potential operational risk lies.
Editorial governance is not AI governance
AI governance asks questions such as:
- What AI should the organisation use?
- Who may use it?
- What risks are acceptable?
- How should privacy and security be protected?
- How should AI decisions be monitored?
Editorial governance, however, asks different questions:
- What does the organisation actually know?
- Which version of that knowledge is authoritative?
- Who is its custodian?
- Who has the right to change it?
- What evidence supports it?
- When was it last reviewed?
- What should an AI system regard as authoritative, and how will it know this?
The distinction is fundamental:
AI governance governs the machine. Editorial governance governs the knowledge that gives the machine something meaningful to say.
Many organisations already have editorial systems – they just don't call them that
Publishers understand the professional disciplines of:
- Commissioning
- Authorship
- Editing
- Fact-checking
- Approval
- Publication
- Versioning
- Corrections
- Attribution
- Archiving
Most businesses perform equivalents of these activities. The difference is that they are usually fragmented across departments, systems and workflows rather than operating as one coherent, unbroken, end-to-end editorial process.
A policy document, sales presentation, website page, product specification, internal procedure and customer-service answer may each contain different versions of what the organisation believes to be true.
The problem is therefore not that organisations lack knowledge.
It is that they lack an editorial system for governing it as a single, consistent living body of knowledge.
I’ve spent around 20 years working within professional publishing systems. What is clear to me now is that organisations are facing a publishing problem without necessarily recognising it as one.
But enterprise knowledge presents organisations – publishers, even – with a problem they rarely encounter at the same scale.
There is no final edition.
The organisation’s knowledge is permanently in motion.
New projects generate new knowledge while existing knowledge is revised, challenged, superseded and reused elsewhere. Knowledge is being created, edited, questioned, replaced, reused, archived – and increasingly consumed by machines.
Organisations therefore need an editorial operating system for a body of knowledge that is permanently changing.
Because the challenge is longitudinal, this requires what we may describe as a “knowledge clock.”
A knowledge clock is not simply a record of when something was published. It tracks when knowledge was formed and became valid, how long it remains legitimate, when it needs to be reviewed, and what happens when the knowledge it describes changes.
The editorial control plane sits between people creating knowledge and machines consuming it.
That is the role editorial governance needs to play.
Why AI makes the editorial governance problem urgent
Poorly governed organisational knowledge has always created problems.
Employees find the wrong document. Customers receive inconsistent answers. Staff rely on outdated procedures. Duplicated information accumulates.
AI changes – and amplifies – the scale of the problem
Once AI can retrieve, summarise, combine and generate answers from diverging streams of organisational knowledge, inconsistencies can become potentially operational.
The organisation is effectively turning its knowledge base into an “executable layer.”
Consider an organisation with five documents describing the same process, produced at different times by different departments.
A search system may return five results.
A human may be able to work out which one should be trusted.
An AI system, however, may synthesise those conflicting sources into one apparently coherent answer.
Yet, while the language may be coherent, the underlying knowledge-logic remains unresolved.
That is the problem editorial governance must address.
Because, when it runs the risk of derailing the organisation, that makes editorial quality a strategic, productivity concern rather than merely one that is administrative.
An organisation may not know which of its own knowledge is trustworthy, which version is authoritative, where a claim originated, who has the authority to change it, or how competing versions should be reconciled.
This is why simply machine-readable knowledge without clear and traceable provenance can become “machine-readable uncertainty.” Worse, it can emerge as “machine-readable liability.”
AI readiness therefore requires organisations to reconcile their knowledge before expecting machines to reason across it.
How to build an editorial governance system for AI
There are five primary steps.
1. Assign knowledge ownership and authority
Every important body of organisational knowledge needs an owner.
This may not be the person who created it. It is the person accountable for its accuracy and continued validity.
For each significant knowledge domain, the organisation should establish:
- The knowledge domain
- The accountable owner
- The contributing experts
- The approval authority and process
- The review frequency and process
Implicit in this, it should also establish a hierarchy of sources.
- Who was the primary author or source?
- Who approved its interpretation?
- Who provided operational guidance?
- Who created or evolved derivative content?
- Who decides what happens when sources or meanings conflict?
If nobody owns a piece of knowledge, nobody can guarantee that it remains trustworthy.
2. Define editorial standards
The publishing analogy becomes concrete when organisations establish standards for:
- Terminology
- Attribution of factual claims
- Dates
- Sources
- Definitions
- Metadata
- Language
- Structure
- Links between related concepts
You cannot make knowledge meaningfully machine-readable until you have established what good knowledge looks like – and put in place the checking systems to uphold it.
3. Create knowledge lifecycles
Every piece of knowledge on which AI depends needs a lifecycle.
The flow of production within that lifecycle may look something like:
Draft → Review → Approval → Publication → Monitoring → Revision → Succession or replacement → Archiving
It should also accommodate corrections and exceptions.
The organisation should be able to answer specifically:
- Who created this?
- Who approved it?
- When?
- On what evidence?
- What did it replace?
The important point is that organisational knowledge should not be treated as a static document.
It has a state, a history and a future.
4. Make organisational knowledge machine-readable
Machine-readability does not simply mean putting documents into a vector database, or connecting a chatbot to SharePoint.
For organisational knowledge to be technically meaningful to AI systems, each significant knowledge object needs enough structure, context and provenance for a machine to understand not just the information itself, but its place within the organisation’s knowledge environment.
Ideally, each significant knowledge object should carry information about:
- What it is and what it means
- Where it came from, including its source and evidence
- Who created it and who owns or stands behind it
- When it was created or updated
- Its current status
- Which version it represents
- What it relates to and how it connects to other knowledge
- When it should next be reviewed
This allows an AI system to distinguish between information that is merely present and knowledge that is current, attributable, contextualised and trustworthy.
It also provides the foundation for maintaining organisational knowledge as a living system, rather than treating documents as static repositories of information.
5. Measure the quality of organisational knowledge
Organisations should be able to measure the quality of the knowledge available to their AI systems.
If an organisation cannot measure the quality of the knowledge available to its AI systems, it cannot properly understand its AI-readiness.
Conclusion – AI strategy is becoming knowledge strategy
The organisations that gain the most from AI will not necessarily be those with access to the most powerful models.
They will be those that have done the harder work of making their own knowledge trustworthy, structured and intelligible.
AI may change how organisations think.
But, as AI must first become a reader, before it can make any greater strategic contribution to any organisation, editorial governance determines what they have to think with.
Contact
Graham Lauren
graham@investigativeai.com.au
+61 416 171 724
#InvestigativeAI #editorialgovernance #editorialoperatingsystem #knowledgeclock