Self-assessment

01Before AI acts for you, by sending, booking or changing something, do you decide what it may do alone and where you’ll check?

Almost always Often Sometimes Rarely

Principle 1: Human oversight of AI

02When AI answers from your organization’s documents, do you know which document it used?

Almost always Often Sometimes Rarely

Principle 2: A single source of truth for AI

03Before you paste or upload information, do you check that the tool is approved for it?

Almost always Often Sometimes Rarely

Principle 3: Confidentiality of information shared with AI

04When you connect an agent to your email, files or accounts, do you give it only the access the task needs?

Almost always Often Sometimes Rarely

Principle 4: Limited permissions for AI agents

05On questions that matter, do you form your own view before you ask AI?

Almost always Often Sometimes Rarely

Principle 5: Independent judgment and skills

06Do you ask AI for the strongest argument against your idea?

Almost always Often Sometimes Rarely

Principle 6: Challenge and bias awareness

07When AI gives you a figure, quote or source you’ll share, do you check it?

Almost always Often Sometimes Rarely

Principle 7: Verification proportionate to the stakes

08Before you send AI-assisted work, do you cut it down and add your own view?

Almost always Often Sometimes Rarely

Principle 8: Quality of AI-assisted work

09When AI shaped something people would care about, do you tell them?

Almost always Often Sometimes Rarely

Principle 9: Transparency about the use of AI

10When something AI-assisted goes wrong, do you raise it quickly?

Almost always Often Sometimes Rarely

Principle 10: Accountability for AI-assisted results