Drift Detection

How policies, systems, and workflows quietly rot

Training 008 · Core Practices
Time: 15–25 minutes

Core stance

Most organizational decay is not caused by bad decisions.
It’s caused by unchallenged persistence.

Drift happens when reality changes and artifacts do not.

Why this lesson exists

Drift is dangerous because:

• It feels harmless
• It accumulates quietly
• It’s invisible to insiders
• It’s expensive to correct later

Organizations rarely fail at once.
They fail by letting “almost correct” become “official truth.”

What drift is (and isn’t)

Drift is

• Gradual divergence between intent and reality
• Unnoticed mismatch between artifacts and practice
• Silent assumption decay

Drift is not

• Deliberate change
• Experimentation
• Adaptation with awareness

Healthy change is explicit.
Drift is change without acknowledgement.

Common drift vectors

Vector 1 — Assumption drift

• Constraints lift
• Context shifts
• Risk profiles change
• But decisions remain frozen

Vector 2 — Process drift

• Shortcuts become normal
• Exceptions become default
• “Temporary” becomes permanent

Vector 3 — Documentation drift

• Docs lag behind practice
• People learn to ignore them
• Trust evaporates

Vector 4 — System drift

• Configurations evolve
• Integrations multiply
• No one revisits original boundaries

Why drift is hard to see

Drift:

• Happens slowly
• Rewards speed
• Avoids confrontation
• Doesn’t break immediately

By the time drift is visible, continuity damage is already done.

Drift vs evolution

A simple distinction:

• Evolution: “We changed this, and here’s why.”
• Drift: “It ended up this way.”

Only one preserves continuity.

Installing drift detection (lightweight)

You don’t need audits. You need signals.

Effective drift signals include:

• “Last verified” markers
• Explicit owners
• Revisit triggers tied to conditions
• Periodic “still true?” checks
• Discomfort with outdated language

Drift detection is noticing misalignment, not policing behavior.

Drift detection and AI

AI systems accelerate drift by:

• Reusing outdated artifacts
• Scaling old assumptions
• Masking decay with performance

Without drift detection:

• AI institutionalizes mistakes
• Governance lags further behind

Exercises

Drill 1 — Drift Scan

Pick one:

• Policy
• Workflow
• System
• AI usage guideline

Ask:

• Is this how things actually work?
• If not, what changed?
• Was that change intentional?

Drill 2 — “Still True?” Marker

Add a simple marker to one artifact:

“Last verified: ___
Revisit if: ___”

That alone reduces drift.

Drill 3 — Drift Conversation

Have a 10-minute team conversation:

“What are we doing differently now than we were a year ago?”

Write down three items.
That’s drift becoming visible.

FAQ

Is drift always bad?
No. Unacknowledged drift is bad. Acknowledged change is evolution.

Who owns drift detection?
Everyone notices drift. Continuity ensures it’s addressed.

How often should we check for drift?
When assumptions change—not on arbitrary schedules.

Suggested next step

Pick one artifact people quietly work around.
Name the drift. Decide whether to formalize or retire it.

That’s how decay becomes evolution.

Next: Training 009 — Key-Person Risk Reduction Without Politics
How to diffuse knowledge safely without threatening expertise.