Decision Provenance

The smallest unit of continuity

Training 006 · Core Practices
Time: 15–25 minutes

Core stance

Most continuity failures begin with a lost decision.

Systems don’t usually break first.
The understanding of why they exist does.

Decision provenance is the practice of preserving just enough information about a decision so it remains intelligible, defensible, and revisitable over time.

Why this lesson exists

Organizations often preserve:

• Code
• Policies
• Configurations
• Outputs

But they lose:

• Why a decision was made
• What alternatives were rejected
• What assumptions were true at the time
• What would justify changing course

When that happens:

• Reversal feels risky
• Change slows down
• AI trains on outcomes without intent
• Accountability becomes fuzzy

Decision provenance fixes this at the smallest possible scale.

What decision provenance is (and is not)

Decision provenance is

• Lightweight
• Context-preserving
• Attached to real work
• Designed for future readers
• Explicit about tradeoffs

Decision provenance is not

• A meeting summary
• A justification memo
• A compliance artifact
• A post-hoc explanation
• A performance review

Provenance exists to preserve meaning, not to defend egos.

Why decisions decay faster than systems

Systems enforce decisions automatically.
Humans remember decisions selectively.

Over time:

• Context disappears
• Assumptions change
• Constraints lift
• People leave

Without provenance, the system keeps enforcing yesterday’s intent—whether or not it still applies.

The minimum viable decision record

A continuity-safe decision record answers four questions:

1. What did we decide?
   (Be concrete.)

2. Why did we decide it?
   (What problem were we solving?)

3. What tradeoffs did we accept?
   (What did we knowingly give up?)

4. What would trigger a revisit?
   (Conditions, not dates.)

That’s it.

If you capture only these four things, continuity improves immediately.

Example (good)

Decision: We will centralize customer audit logs in System X.
Why: Current logs are fragmented and fail audits.
Tradeoffs: Increased vendor dependency; slower access for engineering.
Revisit if: Audit scope changes or vendor pricing exceeds threshold Y.

Readable in 18 months. Reversible in principle.

Example (bad)

“We decided to move logs to X after the Q3 meeting.”

No why. No tradeoffs. No revisit logic.
Continuity failure baked in.

Where decision records should live

Decision provenance works best when attached to the work, not stored in a separate system.

Good locations:

• Architecture decision records (ADRs)
• Tickets or epics
• Policy sections
• Config repositories
• AI system manifests

Bad locations:

• Personal notes
• Slide decks
• Chat threads
• “Someone’s memory”

Decision provenance and AI

AI systems:

• Act on historical decisions
• Generalize past intent
• Scale impact rapidly

Without decision provenance:

• AI amplifies outdated assumptions
• Consent boundaries blur
• Accountability dissolves

With provenance:

• AI mandates are explicit
• Review thresholds are clear
• Failures are explainable

Decision provenance becomes AI governance without bureaucracy.

When to require decision provenance

Not every decision needs a record.

Require provenance when a decision is:

• High-impact
• Hard to reverse
• Automated
• Long-lived
• Compliance-relevant
• Delegated to a system or AI

Low-stakes, reversible decisions don’t need overhead.

Common resistance (and how to address it)

“This will slow us down.”
→ A 5–10 minute record prevents weeks of relearning later.

“We’ll remember why.”
→ You won’t. And future you definitely won’t.

“This is just more documentation.”
→ No. This is intent preservation, not instruction writing.

Exercises

Drill 1 — One Real Decision

Pick a decision from the last 30 days that will still matter in 6 months.

Write a four-line decision record using:

• What
• Why
• Tradeoffs
• Revisit triggers

Stop at four lines.

Drill 2 — Find a Decision Without Provenance

Identify one system, policy, or automation where:

• The decision exists
• The rationale does not

Capture provenance retroactively—briefly and honestly.

Drill 3 — Provenance Gate

Choose one workflow (e.g., production changes, AI deployment).

Define:

“Decisions of type X require provenance.”

That single rule creates continuity at scale.

FAQ

Is this the same as ADRs?
ADRs are one implementation. Decision provenance is broader and applies beyond architecture.

Who owns decision records?
The decision owner—not documentation teams.

Can provenance be wrong?
Yes. That’s why revisit triggers matter more than certainty.

Suggested next step

Pick one decision-making area (product, security, AI, compliance).
Introduce a four-question decision record.

You’ve just installed the smallest, most powerful continuity primitive.

Preview: Training 007 — Survivable Workflows
How to make processes reconstructable without over-documenting.