OPEN RESEARCH METHOD · 2026
AI Product Readiness Benchmark.
A transparent research framework for measuring where fast-built AI and SaaS products fail after the happy path. Results will not be published until the sample is large enough to support them.
WHAT WILL BE MEASURED
Five layers, fifteen observable checks.
- Flow. Primary task clarity, hierarchy and decision sequence.
- States. Loading, empty, error, permission, partial completion and recovery.
- System. Pattern reuse, component consistency and design/code alignment.
- AI trust. Scope, progress, evidence, uncertainty and human control.
- Recovery. Undo, retry, correction, escalation and handoff.
PUBLICATION RULES
No synthetic statistics.
We will not publish percentages from generated examples, convenience guesses or tiny samples. The first public benchmark requires at least 100 independently reviewed product flows and will include sample size, selection method, scoring rubric, exclusions and observed limitations.
Company-identifying details will not be published without permission. The benchmark is intended to produce useful aggregate product-design evidence, not a leaderboard of individual products.
The current scorecard is a diagnostic framework, not benchmark data.
CURRENT STATUS