Resource · Framework

AI tool evaluation framework

Before adopting any AI tool for teaching, learning or administrative use, it helps to answer the same set of questions every time — rather than judging each tool on how impressive its demo looks. This framework is the checklist I use.

All resources
AreaQuestion to ask
Data privacyHow is user or student data handled? Is anything sent to third parties, and where is it stored?
Accuracy & reliabilityWhat's the error rate in practice? Is there a human review step for anything high-stakes?
AccessibilityDoes it work with screen readers and keyboard-only navigation?
Integration effortDoes it fit into existing systems, or does it require retraining staff and changing workflows?
Cost & licensingWhat's the real cost at institutional scale, and are there compliance requirements attached to the licence?
Responsible-use fitDoes it align with the institution's existing AI-use policy, or does a policy need to be written first?

How to use it: Score each tool 1–3 per row (1 = concern, 3 = clear pass) before it reaches a pilot. A tool that scores well on capability but poorly on privacy or accessibility should not move forward just because it's capable — those are the two areas most likely to cause real harm if skipped.