AI Governance
Paper Gravity: When Governance Becomes the New Drift
Bureaucracy already knows how to drown itself in paper.
AI gives it a faster printer.
That is the simplest way to understand one of the coming governance risks around agentic systems. The problem is not that documentation is bad. Documentation is necessary. Audit trails matter. Evidence matters. Handoffs matter. Review records matter.
The problem begins when documentation becomes the center of gravity.
An agent can produce a report. Then a summary of the report. Then a compliance note. Then an evidence pack. Then a handoff. Then a review artifact. Then a risk register update. Then a meeting brief explaining the artifact chain.
All of this can look mature.
Some of it may be useful.
But if the paper layer starts expanding faster than the working layer can verify it, the organization has not become safer. It has become heavier.
This is paper gravity.
It happens when governance artifacts begin to pull attention away from the live system they were supposed to clarify. People start managing documents about the work instead of the work. Agents start preserving the shape of compliance instead of the truth of the situation. Handoffs become authority objects. Reports become shields. Evidence packs become substitutes for understanding.
In a human-only bureaucracy, this is already damaging.
In an agentic environment, it can scale frighteningly fast.
An agent does not get tired of writing. It does not feel the social cost of asking a team to read another twenty-page document. It can fill gaps with plausible structure. It can make stale material look orderly. It can produce documentation with a confidence that exceeds the reliability of the underlying work.
That is why governance design has to distinguish between evidence and paper.
Evidence connects back to a live claim.
Paper often only connects back to another paper.
Signalane’s concern is not that organizations will document too much in a simple numerical sense. The concern is that they will document in the wrong direction.
Good documentation should answer:
- what changed
- why it matters
- what evidence supports it
- what remains uncertain
- who can decide
- what should happen next
Bad documentation answers:
- where the template is
- which checkbox was completed
- how the status can remain green
- which phrase avoids blame
- how the artifact can be handed to the next lane
The AI Act may increase documentation pressure. That is not automatically bad. But if organizations respond by building paper factories instead of working judgment systems, they will create a new drift layer.
The agent will not need to rebel.
It will only need to keep producing plausible paper.
And the human system, already trained to respect documents, may mistake that for governance.