Where should the boundary exist between AI governance and human governance in enterprise Confluence?

[AGASTYA ANOOP SHARMA]

As Rovo becomes capable of generating content, restructuring documentation, and recommending edits, an important governance question emerges.

Which responsibilities should remain exclusively human?

For example:

approving architectural decisions,

resolving conflicting documentation,

defining canonical knowledge,

or validating regulatory content.

Should Confluence eventually support governance policies that explicitly define where AI assistance ends and human authority begins?

[Katerina Bletsch]

I governance is not only page approval; it is authority over canonical knowledge and terms. 

I would draw the boundary at authority, not at content creation.

AI can draft, summarize, restructure, and find inconsistencies. But humans should own anything that becomes authoritative knowledge.

In Confluence, I would keep these human-owned:

• approving architecture or regulatory decisions
• defining canonical terms
• resolving conflicting definitions
• deciding which version is the approved source
• validating compliance or customer-facing content

Risks

The important risk is not only “AI may write something wrong”. The bigger risk is that AI-generated wording becomes treated as the official source without a clear owner,
approval, or change history.

For example, in regulated documentation, terms like “incident”, “control owner”, “approved document”, “risk acceptance”, or “high-risk AI system” should not be casually
rewritten by AI. They need an approved definition, owner, and review trail.

So I would use this pattern:

• AI may suggest wording
• a human owner approves the canonical version
• Confluence stores the approved source
• critical terms and regulatory content require review before becoming official
• old or conflicting wording is periodically found and cleaned up

This is where many AI governance workflows stop too early: they govern the page, but not the meaning inside the page.

Disclosure: I build Compliance Glossary for Confluence, so I work specifically on governed terminology and approved definitions. I see this as one part of the human-governance
layer, not something AI should own by itself.

I’m also open to a short call if it would be useful to compare notes and understand the specific governance pains you are seeing in practice.

[Viswanathan Ramachandran]

Hi @AGASTYA ANOOP SHARMA 

Good post

The boundary should sit at ownership and accountability.

AI handles scaling and consistency, while humans own decisions, quality, and risk.

The Core principle is

• AI governance rules about how it can generate, edit, or suggest content (templates, styles, tools)
• Human governance is who approves, owns, and takes responsibility for that content (page owners, reviewers, compliance)

In simple and in practical: 
AI’s role is governed by policy and Humans role governed by processes. 

[Rob Hean]

This is one that could easily vary between organizations and or environments.  For example an engineering firm that manages bridges would likely have different rules than one that builds websites.

For me AI is a tool, and I wouldn't trust Outlook, or my shoes, or a hammer to make a decision about something. I'm personally ok with AI making the mechanical choices (font size, formatting, etc), but anything that involves authorization, approval, review should still stay with humans 

This isn't to say AI can't help, or recommend paths forward, just that humans are better equipped for that decision making. (Also what's the recourse if AI makes a mistake?)