Repeat-Use Build Filter
Build reusable AI assets and reject costly disposable distractions
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 98%
The Repeat-Use Build Filter adds a reusability test after a proposed AI build has been linked to a desirable outcome. Ask whether the finished artifact or workflow will be used more than once. If repeated use is likely, estimate how often it will run, what recurring work it replaces, and whether those accumulated benefits justify the implementation and maintenance cost. If the result is completely disposable, impose a severe time limit: only proceed when it takes no more than a couple of minutes or when the one-off learning has exceptional value. Otherwise decline the build. The filter counters the temptation created by AI's low apparent creation cost, recognizing that rapid experiments still consume attention, tokens, review time, and maintenance capacity that could be directed toward durable systems.
Origin
Extracted from Marketing Against The Grain as Kipp and Kieran discussed a second filter for distinguishing strategic AI systems from one-off disposable builds.
Core principles
- 01Strategic work creates repeatable actions
- 02Expected reuse increases the return on build time
- 03Disposable work deserves a very low time budget
- 04A promising outcome does not automatically justify implementation
- 05Fast building still carries attention and opportunity costs
How to run it
- 1
Confirm the intended outcome
Identify the useful result the proposed build is expected to create. Discard ideas that cannot pass the outcome test before considering reuse.
Pro tip Apply this after writing an AI-by-outcome strategy sentence.
Watch out Reusability cannot rescue a tool that repeatedly produces an irrelevant result.
- 2
Test for repeated use
Ask whether you or the team will use the artifact more than once. Name the recurring situation that will trigger its use.
Pro tip Estimate a realistic weekly or monthly usage frequency.
Watch out Do not count hypothetical users or imagined future workflows as proven reuse.
- 3
Estimate total cost and return
Compare build, token, review, and maintenance costs with the accumulated value of repeated use.
Pro tip Include the attention required to verify AI-generated work.
Watch out A 30-minute build can still be expensive when dozens of distractions accumulate.
- 4
Apply the disposable threshold
If the artifact is truly one-off, proceed only when it takes no more than a couple of minutes or resolves a high-value uncertainty.
Pro tip Use a strict timer for disposable experiments.
Watch out Do not let a quick curiosity grow into an unplanned project.
- 5
Build or decline
Proceed with reusable, outcome-linked assets and brief high-value experiments. Decline or defer disposable ideas whose costs exceed their one-time benefit.
Pro tip Record rejected ideas rather than interrupting current work.
Watch out Ease of implementation is not sufficient justification.
In the wild
A manager considers building an assistant that retrieves relevant context and drafts responses several times each day. Because it supports a recurring workflow and can reduce response time repeatedly, the accumulated benefit can justify a careful build and ongoing refinement.
→ Build time creates continuing productivity gains rather than a single novelty output.
A marketer notices that AI could create a decorative widget for a page that will be discarded after one presentation. The build would take 30 minutes and has no meaningful learning objective. The filter rejects it because it is disposable and exceeds the permitted couple-minute threshold.
→ The marketer preserves attention for reusable, outcome-linked work.
Common mistakes
Confusing easy with worthwhile
AI can make implementation easy without making the resulting artifact valuable or reusable.
Inventing future reuse
Calling an artifact reusable without naming a recurring trigger overstates its likely return.
Ignoring maintenance
Repeated use can amplify value, but it can also amplify verification and upkeep costs.
Is it for you?
Best for
It is best for knowledge workers and developers who can rapidly create small tools, automations, pages, or agents with AI.
Not ideal for
It is not ideal for intentionally disposable prototypes used to answer a high-value uncertainty quickly.
From the transcript
“And also, like Kieran, strategies also a set of repeatable actions.”
“Am I gonna use it more than once?”
“If this is really like a completely disposable thing and it's gonna take me more than a couple minutes, like I probably shouldn't do it.”
From the episode
They Spent $150,000 on AI Tokens (And Got Nothing)