Use this template to measure AI-assisted platform builds honestly. Fill it in as you go — don't try to reconstruct scores after the fact.
See methodology.md for scoring definitions.
- Project: your project name
- Date: build date
- AI Tool: tool name and model
- Platform: target platform (e.g., EKS 1.34, GKE, AKS)
| Component | Toil Reduced (1-10) | Correction Cycles | AI Time | Est. Manual Time | Toil Shifted? | Notes |
|---|---|---|---|---|---|---|
| Component 1 | ||||||
| Component 2 | ||||||
| Component 3 |
- Total AI-assisted build time: ___ hours
- Estimated manual build time: ___ hours
- Net toil reduction: ___%
- Total correction cycles: ___
- Average toil reduced score: ___/10
- Components where AI genuinely reduced toil: /
- Components where AI shifted toil (Partial): /
- Components where AI fully shifted toil (Yes): /
- Components where starting from scratch would have been faster: /
List the top 3-5 components where AI provided the most value.
List the components where AI required the most correction cycles and why.
What caused most corrections? Stale knowledge? Wrong API versions? Misunderstood requirements?
Based on your experience, what practical advice would you give teams considering AI-assisted platform builds?
Toil Reduced (1-10):
- 10 = Would have taken hours manually, AI did it in minutes with no corrections
- 5 = AI produced a starting point, but required significant human correction
- 1 = AI output was wrong enough that starting from scratch would have been faster
Correction Cycles: Number of times AI output needed human intervention.
Toil Shifted?:
- No = AI genuinely reduced total effort
- Partial = Some reduction, some shift
- Yes = AI converted "writing config" toil into "reviewing and debugging AI config" toil