Chapter 30. Operating Rhythm
The cadences that make governance a routine rather than an event, and what each body should be looking at when.
30.1 Why cadence matters more than procedure
A governance framework with well-designed mechanisms and no rhythm degrades into reactive review. Work arrives when it arrives, bodies convene when there is a queue, and the periodic activities that keep classification accurate — re-scoring, sampling, calibration — never happen because nothing schedules them.
The framework specifies almost no cadences. What follows is a working rhythm you can adopt and adjust, marked throughout as practice, not requirement.
30.2 A working calendar
Table 77.
| Frequency | Activity | Owner |
|---|---|---|
| Continuous | Control Plane comparison; alert routing | Automated, with named owners per category |
| Daily | Tier 4 alert response; agent boundary divergence response | System Owners, AI Governance Lead |
| Weekly | Risk classification sessions; countersignature clearance; drift triage for Tier 3 | Risk and Compliance Lead |
| Fortnightly | ARB: Tier 3–4 G2 decisions, pattern proposals, dependency conflicts | ARB |
| Fortnightly | AI Governance Body: Tier 3–4 G3 and G4, autonomy grants, residual risk acceptance | AI Governance Body |
| Monthly | Exception register review; expiring exceptions; open G3 conditions | AI Governance Lead |
| Monthly | Aged drift backlog review | Architecture with System Owners |
| Quarterly | Tier 1–2 audit sampling; tier distribution against incident data | Risk and Compliance Lead |
| Quarterly | Pattern library review; recurring exception analysis | ARB |
| Quarterly | Withheld funding concurrence report | AI Governance Body to Enterprise Sponsor |
| Half-yearly | Scoring anchor calibration session | Risk Lead with Architects and Sponsors |
| Annually | Tier 3–4 risk re-score; estate sweep for orphaned systems; principle review | Various |
Two rhythms carry disproportionate value and are the ones most often skipped.
Weekly countersignature clearance. If countersignature waits for a monthly meeting, teams work around it or the classification arrives after architecture is settled. It needs to be a short, frequent, low-ceremony activity.
Half-yearly anchor calibration. Scoring anchors drift in application even when the text does not change. A two-hour session re-scoring a handful of past systems as a group, and discussing disagreements, does more for consistency than any amount of written guidance.
30.3 What each body should see
Architecture Review Board, standing agenda
Table 78.
| Item | Purpose |
|---|---|
| Tier 3–4 G2 decisions | The board’s primary work |
| Pattern proposals and deprecations | Keeps the fast path alive |
| Dependency conflicts referred from solution level | The cross-initiative view nobody else holds |
| TG0 bypass findings | Work that entered the estate outside governance |
| Drift patterns by pattern | The Learn function’s output; pattern quality signal |
AI Governance Body, standing agenda
Table 79.
| Item | Purpose |
|---|---|
| Tier 3–4 G3 and G4 decisions | Primary work |
| Autonomy grants and increases | Always this body, at any tier above 1 |
| Kill-switch condition approvals | Separate approval from general G3 sign-off at Tier 3–4 |
| Residual risk acceptances | Explicit, recorded, owned |
| Funding concurrence for Tier 3–4 initiatives | The bounded exception to portfolio authority |
| Escalated unresolved alerts | Where local resolution failed |
Quarterly governance review, for the Enterprise Sponsor
Six numbers, no narrative required: tier distribution against incident distribution, countersignature adjustment rate, gate lead time by tier, exception count and aged exceptions, aged drift backlog, and automated evidence share. Each is defined in Chapter 29.
30.4 Keeping meetings decision-shaped
Governance bodies drift toward discussion. Three practices keep them producing decisions.
Pre-circulate evidence, and enforce it. A case arriving without its required evidence is deferred, not discussed. This is unpopular for one cycle and then self-correcting.
State the decision required. Every agenda item names the decision and the options. “Discuss the claims triage system” is not an item; “Approve, approve with conditions, or reject G3 for the claims triage system at Tier 3” is.
Record the reasoning, not just the outcome. The Decision record needs enough rationale that someone reconstructing it in two years understands why. This is also what makes drift comparison meaningful later.
[Practice recommendation] Timebox Tier 3 items to fifteen minutes and Tier 4 items to thirty. Items exceeding their box are deferred with a named follow-up, not allowed to consume the agenda. Bodies that routinely overrun stop attracting the senior attendance they depend on.