Read it back
Read your shortlist as if a colleague you respect wrote it. Are you persuaded? If not, you have found the section to tighten. Five questions:
- Does the risk profile name your context specifically, or could it apply to any team?
- Is every tool named precisely? Tool plus account type plus version where relevant.
- Does every rationale connect the tool to your risk profile, or do some describe the tool's features instead?
- Is at least one exclusion explicit, with a reason and review criteria?
- Is the review cadence specific? A real timeframe and an honest trigger, not a vague "review periodically."
Review your shortlist against four dimensions
This is a self-review, and the rubric below is the tool for it. The standard is whether the artefact is something you can stand behind yourself: clear enough to remember, specific enough to act on, and honest enough that you trust your own thinking when you return to it in three months. If you can recruit a colleague, a fifteen-minute exchange of drafts against the same four dimensions is worth doing.
| Criterion | Pass | Merit | Distinction |
|---|---|---|---|
| Tool specificity | Each tool named. | Version or account type named. | Specific feature or configuration that matched the risk profile is named. |
| Risk profile | Relevant factors listed. | Factors connected to the type of AI work in question. | One factor identified that surprised you or shifted a decision. |
| Rationale quality | A rationale for each tool. | Rationales a colleague could understand without explanation. | Precise enough that a junior colleague could swap one tool for an equivalent without changing the logic. |
| Constraints named | At least one explicit exclusion. | Reason for the exclusion named. | Criteria under which the exclusion would be reviewed. |
The Week 6 capstone uses an extended version of this rubric for the same self-review. Knowing the rubric this week is part of the design.