Three-Layer Quality Stack
Set quality standards for time, project selection, and individual outputs.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 94%
The Three-Layer Quality Stack treats AI-work quality as a hierarchy rather than an artifact-level editing problem. At the top, leaders determine how people should spend their expanded capacity. The middle layer selects the small number of projects that matter from the much larger set AI makes possible. The bottom layer defines what excellent work looks like for each chosen deliverable and codifies those standards. Managing only the bottom layer can create polished but useless applications; managing only priorities can still produce weak execution. Leaders therefore move up and down the stack, aligning time allocation, project choice, and output quality with strategy and customer value. The result is focused leverage rather than faster production of low-value work.
Origin
Extracted from Marketing Against The Grain, where Hillary Gridley described quality management as a stack extending from team activity to project selection and individual work.
Core principles
- 01Quality must be managed at multiple levels.
- 02Productive activity is not the same as valuable activity.
- 03Project selection precedes execution quality.
- 04Every deliverable needs an explicit definition of good.
- 05Leaders must align the layers rather than optimize one in isolation.
How to run it
- 1
Set the strategic direction
Clarify the customers, value, priorities, and outcomes that should govern the team's work. Use these as the foundation for every lower-level decision.
- 2
Govern time allocation
Decide what kinds of activity deserve the team's attention and what should stop. Account for the much broader scope of action AI gives each employee.
- 3
Select high-value projects
Compare the many possible projects and choose the few most likely to create customer or business value. Explicitly reject attractive distractions.
- 4
Define deliverable quality
For each selected project, state what a good output must achieve. Include audience, purpose, constraints, evidence, and acceptance criteria.
- 5
Codify recurring standards
Turn stable criteria into briefs, rubrics, checklists, or focused AI tools. Make the standards available at the moment work is produced.
- 6
Audit the complete stack
Review whether time, project selection, and artifact quality all support the same strategic outcome. Correct the highest broken layer before polishing lower ones.
In the wild
A marketing team has ideas for one hundred AI applications. Leadership first defines the customer outcome that matters, then limits the team to three projects with credible impact. Each selected project receives explicit acceptance criteria for usefulness, quality, and adoption. The remaining ideas are paused rather than celebrated as evidence of transformation.
→ The team directs its new capacity toward a few valuable products instead of shipping numerous unused experiments.
Common mistakes
Polishing the wrong project
Excellent execution does not create value when the underlying project has no strategic purpose or audience.
Rewarding AI activity
Counting tools, prompts, or generated artifacts encourages visible experimentation rather than meaningful outcomes.
Managing only one layer
Focusing solely on time, priorities, or artifact quality leaves the other layers free to undermine results.
Is it for you?
Best for
It is best for leaders managing teams with more AI-enabled capacity than organizational focus.
Not ideal for
It is not ideal as a substitute for a missing company strategy or unresolved customer priorities.
From the transcript
“And then on the level below that, it's like for any one project, right? What is the work that is coming out of it and…”
“Like you can't you can't just focus on one piece of it. Uh, you really do have to be managing at all layers of that.”
From the episode
If Your Team Is Producing AI Slop, Here's How To Fix it