---
title: "The Creative Operations Stack"
date: "2026-01-29"
status: "Published"
readTime: "10 min"
abstract: "In a 14-week single-studio pilot, decision latency fell by roughly 68% and no projects went unreviewed. Three integrated frameworks for managing AI-augmented creative work."
codexMode: "System"
codexClass: "Foundations"
heroImage: "/media/codex/creative-operations-stack-v3.png"
keywords:
  - "creative operations"
  - "AI governance"
  - "multi-clock work"
  - "strategic outcomes"
  - "campaign framework"
  - "operations infrastructure"
version: "v1.0"
author: "Andy"
topicTags:
  - "ai-governance"
  - "creative-operations"
  - "workflow-design"
  - "strategic-planning"
difficultyLevel: "Intermediate"
targetRoles:
  - "Studio founder"
  - "Operations leader"
  - "Creative director"
  - "AI-curious executive"
tldr: "AI can do the work. Knowing what to work on, when it's done, and whether it matters — that's the hard part. 14 weeks of measurement, three frameworks, preliminary results."
---

import Figure from '../../components/CodexFigure.astro';
import RelatedArticles from '../../components/CodexRelatedArticles.astro';
import Callout from '../../components/CodexCallout.astro';

Fourteen weeks ago, I started measuring where my time actually went. Not the broad categories — "admin," "creative work," "planning" — but the specific transitions. How long between deciding what to work on and actually working. How many things sat half-finished at any given moment. Which projects aged indefinitely while urgent-but-trivial tasks kept jumping the queue.

The pattern that emerged wasn't what I expected. AI had genuinely accelerated everything that could be automated — first drafts in minutes, research in an afternoon, visual concepts faster than I could evaluate them. But the parts requiring human judgment hadn't sped up at all. If anything, they'd gotten harder, because there was simply more to evaluate, more to prioritise, more to decide whether it was actually done.

The production bottleneck had dissolved. In its place: a governance bottleneck.

---

## The Results

Before building a framework, I measured the baseline. After fourteen weeks running the system, I measured again. These are self-reported metrics from my own tracking; your experience will differ.

**Decision latency: 25 minutes → 8 minutes in my tracking.** That's the time from "what should I work on?" to actually working. Before: reviewing calendars, reconstructing context, weighing priorities. After: check the prioritised list, read the "next action" note, begin.

**Work-in-progress: stable throughout the pilot.** Active items held at 5-6 (high-frequency), 10-12 (low-frequency), 15-18 (dormant). No gradual inflation. No emergency weeks where everything became urgent simultaneously.

**Unreviewed projects: none during the pilot period.** Every aged item surfaced for review as designed. Eight ideas resurfaced from the backlog — work that would likely have been forgotten in a traditional system.

**Throughput: 4-6 completions per week.** Sustainable rhythm, not heroic bursts followed by recovery.

<Callout title="Context and Method">
Single-studio pilot. One founder, one collaborator, AI agents. Specific tools (Claude + Notion). Metrics are self-reported based on task timestamps and weekly reviews. "Completion" means an item moved to Done status with a recorded output. Your context will differ; results will vary.
</Callout>

These numbers came from a system I built for myself. The rest of this article explains how it works — for readers who want the detail behind the results.

---

## The Problem It Solves

When AI compresses delivery from days to minutes, every surrounding system feels the strain. Approval workflows designed for weekly cycles can't keep pace. Planning built around month-long projects doesn't make sense when the work takes an afternoon. Quality checks built for scarcity become overwhelmed by abundance.

The result: organisations that can generate more than they can decide, produce more than they can evaluate, start more than they can finish.

You can't solve this with more AI. It's a structural problem — how work flows through decisions, not how fast it gets produced.

---

## The Framework

Three connected components. Each addresses a specific failure mode; together they form a feedback loop.

<Figure
  src="/media/codex/three-pillars.png"
  alt="Three pillars: Thinking, Doing, Governing — with feedback between them"
  caption="Thinking (purpose) → Doing (execution) → Governing (completion) → back to Thinking."
/>

**Thinking** connects daily tasks to strategic purpose. **Doing** handles scheduling and execution. **Governing** focuses on completion. The key insight: these three need to be connected. Thinking without Doing is strategy that never executes. Doing without Governing is activity that never finishes. Governing without Thinking is process for its own sake.

<Callout title="Pillar 1: Thinking — The Strategic Outcomes Framework">
Purpose works through four connected layers.

**Value Streams** provide portfolio perspective — financial performance, brand, client delivery, innovation, operations, learning. Adapted from the Balanced Scorecard, they help prevent the common problem of over-focusing on whatever's most measurable while neglecting everything else.

**Strategic Outcomes** are specific future states, more concrete than vision statements but more durable than quarterly targets. "Generate £50K recurring revenue" rather than "grow the business." They sit at a 12-24 month horizon — stable enough to guide decisions, specific enough to know when you've achieved them.

**Initiatives** are bounded programs advancing those outcomes — campaigns, client engagements, R&D projects, internal builds. Each links explicitly to the outcomes it serves.

**Tasks** are daily work, each connected to an initiative.

The value is in the linkage. Point to any task, trace two hops up, and you should be able to explain why it matters in a single sentence: "This LinkedIn post builds authority in our methodology, which drives consulting inquiries — a Q1 priority."

I call this the golden thread. When it's intact, every piece of work connects to purpose. When it breaks — when you're doing things for reasons you can't articulate — the framework helps make the break visible before it accumulates into drift.

[Full article →](/codex/strategic-outcomes/)
</Callout>

<Callout title="Pillar 2: Doing — Multi-Clock Work">
Scheduling starts from a simple observation: different types of work have different natural rhythms, and forcing them onto a single calendar tends to create problems.

**Starvation** happens when important-but-not-urgent work never gets scheduled. Research projects age indefinitely. Strategic thinking happens "when there's time," which functionally means never. The urgent crowds out the important until the important becomes urgent — at which point it's usually too late to do it well.

**Thrashing** happens when you switch contexts too frequently. You start the client deck, check email, remember the invoice, open the R&D document, and accomplish nothing meaningful on any of them. Each switch carries overhead that compounds across the day.

The solution is three frequency bands with different rules:

**High-Frequency** covers work requiring attention multiple times per week — client deliverables, active production, anything with external deadlines. A limit of 3-6 active items helps ensure you can actually complete things rather than just touching them.

**Low-Frequency** covers work touched every 2-6 weeks — strategic planning, skill development, relationship nurturing, research. Limit of 8-12 items. These surface weekly for explicit refresh-or-defer decisions.

**Dormant** holds parked concepts with periodic review. Not abandoned — deliberately set aside. Limit of 15-20 items. An aging mechanism helps surface neglected items; they accumulate points over time until they cross a threshold and appear for review.

Beyond scheduling, the Doing pillar includes **process definitions** that encode how your studio actually works (not generic workflows), **tool connections** that link AI to your systems with clear boundaries on what needs human approval, and **improvement loops** where completed tasks feed learning back into the system.

[Full article →](/codex/multiclockv2/)
</Callout>

<Callout title="Pillar 3: Governing — Completion Protocols">
Three mechanisms help work actually finish.

**Work-in-progress limits** encourage finishing. There's a well-known relationship in operations: the more things you have in progress simultaneously, the longer each one takes to complete. The framework sets limits at task, brief, and campaign levels. When you're at capacity, new work waits until something completes. This feels constraining at first, but tends to result in more things actually getting done.

**Aging mechanisms** help prevent starvation. Every day an item isn't touched, it accumulates points. Cross a threshold and it surfaces for review — advance it, take action, or explicitly extend its dormancy with a reason. This makes it harder for the backlog to become a graveyard of forgotten ideas.

**Completion protocols** define what "done" actually means. Not "I worked on it" but "here's the output, here's verification it meets the spec, here's what we learned." Three checkpoints: Specification (define done before starting), Verification (confirm completion with evidence), Learning (extract insights for next time).

Together, these create structure for accountability. The rules are explicit, the thresholds are visible, the logic can be reviewed. When something goes wrong, you can trace backward through the system and find where it broke. This is what I mean by governance embedded in workflow design rather than enforced through inspection after the fact.

</Callout>

---

## Beyond Creative Studios

The framework started in a creative context — photography, content, brand development. But the underlying problem may appear wherever AI has accelerated production while human judgment remains the constraint.

Consider contexts like insurance operations, where AI might draft decisions quickly but approvals still take time. Or financial services, where quotes can be instant but exceptions queue up. Or marketing teams, where AI generates many variants but deciding which one ships remains difficult.

I don't know yet how well this transfers. That's partly why I'm writing about it.

---

## What I'm Looking For

I'm looking for a creative studio to pilot the full system.

**You:**
- 1-10 person team
- Multiple parallel workstreams
- Already using AI tools
- More output than clarity, more starts than finishes
- Willing to share before/after metrics

**What you get:**
- Implementation support
- Custom process definitions for your workflows
- 6-week pilot with weekly check-ins
- Measurement of decision latency, WIP, throughput

**What I ask:**
- Permission to use anonymised metrics as case study (subject to your written approval before any publication)
- Candid feedback on what works and what doesn't
- Introduction to 2-3 peers if successful

Pilot is offered at no cost for qualified studios, in exchange for feedback and case study participation. We'll only publish aggregated or anonymised metrics with your explicit agreement. Please don't share any third-party confidential information you're not authorised to disclose.

This is research, not a product pitch. I have something that works for me. I want to know if it works for others.

Interested? [Get in touch via the product page](https://nullproof.studio/products/mynullproof/) or connect on [LinkedIn](https://www.linkedin.com/company/nullproof-studio/).

---

<RelatedArticles
  items={[
    {
      title: "Multi-Clock Work Framework",
      href: "/codex/multiclockv2/",
      description: "The three-band system, prioritisation scoring, and aging mechanics."
    },
    {
      title: "Strategic Outcomes Framework",
      href: "/codex/strategic-outcomes/",
      description: "The four-layer hierarchy connecting tasks to strategic purpose."
    },
    {
      title: "Campaign Framework",
      href: "/codex/campaign-framework/",
      description: "Three-tier system for content execution across platforms."
    }
  ]}
/>