Readiness quiz
Check whether your evidence, permissions and ownership are ready for a pilot.
- Pages
- 20
- Sources cited
- 11
- Last updated
- 26 October 2026
Can we define one business task and its failure consequences?
Evidence to show: task contract and three sanitised examples with expected outcomes.
A desk-designed readiness rubric, not a validated prediction of success. Progress stays in this browser; no answers are sent to the desk.
Your evidence action plan
Answer a question to identify the evidence gaps that belong here.
Read the full research.
What does this quiz tell us?
It identifies whether you can commission a useful, controlled pilot—not whether fine-tuning will work.
How do we answer and calculate the result?
Choose one answer for each of 12 questions: A = 0, B = 1, C = 2 points.
1. Can we define one business task and its failure consequences?
A — 0: We want “better AI” or lower tokens, without an accepted-output definition.
2. Do we have a dated, reasonably optimised baseline?
A — 0: There is only a demo or an undocumented current model.
3. Do we have documented permission for the planned data use? — HARD GATE
A — 0: Rights are unknown or someone assumes owning the database is enough.
4. Are the labels and examples trustworthy?
A — 0: Labels are generated or inherited without checking.
5. Can we protect an independent final test? — HARD GATE
A — 0: The same examples guide training and prove success, or no test exists.
6. Can we judge results consistently?
A — 0: Success is a subjective demo reaction or an unexplained model score.
7. Does the test cover the production cases that matter?
A — 0: Only easy or familiar examples are included.
8. Can we model cost per accepted task?
A — 0: We compare token or GPU prices alone.
9. Do we know the deployment constraints?
A — 0: Serving is assumed to be easy after training.
10. Is a person accountable for acceptance and operation? — HARD GATE
A — 0: There is no named buyer owner after the pilot.
11. Can we leave the supplier or replace the model?
A — 0: “Ownership” is asserted but required assets are unspecified.
12. Have we reviewed the actual use and processing risks? — HARD GATE
A — 0: Data sensitivity, affected people and intended actions are unknown.
What does the total mean?
Apply the hard gates first, then use the score as a preparation guide.
What should the result page or internal note contain?
Write the total, gate status, three highest-priority gaps, owner and next evidence required.
Frequently asked questions
Short answers to the questions buyers ask.
Read →18 / REFERENCEWhat remains unproven
Evidence gaps and limits that still need resolving.
Read →19 / REFERENCEEditorial self-assessment
The research’s own quality scores and limitations.
Read →20 / REFERENCESources
Trace the section’s claims to the original sources.
Read →Useful evidence.
Visible limits.
Practical, transparent instrument with auditable scoring and non-compensable blockers. Its validity is procedural rather than predictive, and that limitation is explicit.