CIRK Scoring Guidance — How to Score Tasks

Scoring guidance

Score pragmatically and consistently.

The goal is not theoretical precision. The goal is consistent execution governance.

Scoring flow

For each task, ask:

C How much of the system must be understood?

I How likely is it that multiple cycles are needed?

R How much human validation is required?

K How risky is rollout or integration?

Then assign a value from 1 to 3 for each dimension.

Practical advice

Three rules for better scoring.

Prefer relative consistency over perfect precision

If one team scores slightly differently from another, that is acceptable. What matters is that the scoring is stable enough to drive predictable policy.

Score the execution reality, not the political preference

Do not lower a score just because the team wants faster approval. Do not raise a score just to block work. CIRK should describe the task honestly.

Use the highest meaningful signal

If review is truly critical, score R3 even if the rest of the task feels simple. If rollout is sensitive, score K3 even if implementation is small.

Contrast

Same speed, different governance.

These tasks may both be quick to draft with AI, but their execution conditions are radically different.

Update copy on marketing page

C1 I1 R1 K1

Trivial

Modify billing webhook behavior

C2 I2 R3 K3

Complex

Execution window

Deriving time from the vector.

For teams that need a time-adjacent signal — particularly during transition from traditional planning — CIRK vectors can be mapped to an execution window.

Vector profile	Execution window	Why
All ≤ 1	Minutes	Autonomous, no gates
Low R, low K	Hours	Fast execution, light review
High I (I3)	Hours to 1 day	Iteration cycles add elapsed time
High R (R3)	+1 day	Human review is calendar-bound
High K (K3)	+1–2 days	Deploy coordination, runbooks, staging
Score ≥ 10	Days	Multiple approval gates compound
C3 I3 R3 K3	Blocked	Must decompose before estimating

This is not a CIRK output. It is a derived heuristic that teams can calibrate to their own cadence.

High R and high K add calendar time — not because the work is large, but because humans and processes are in the critical path.

Scoring FAQ

Common scoring questions.

"My PM needs a deadline. What do I tell them?" +

Give them an Execution Window derived from the vector. Explain that the window is governed by review and rollout constraints, not code complexity. This is more accurate than story points ever were — because it models the actual bottleneck.

"Can I track velocity with CIRK?" +

Not in the story-point sense. But you can track throughput by execution mode: how many autonomous tasks shipped this week vs. how many supervised tasks completed. That tells you more about team capacity than a velocity chart.

"What if my team is not using AI agents?" +

CIRK still works. Even in fully human workflows, the dimensions hold: context depth, iteration likelihood, review burden, and integration risk are real constraints regardless of who writes the code. The execution policies apply the same way.