Sign in or create a free account to save this Card to your Deck.

Principle Card · A rule to test

Document permitted uses of AI and company information

DATA PERMISSION IS PART OF AI COMPETENCE.

Employees need clear rules for which AI tools may be used, what company or customer information may be entered, what outputs require checking, and where uncertainty must be escalated. A policy is useful when staff can apply it during ordinary work, not only when a compliance document is opened. The rules should reflect the company's contracts, privacy obligations, security controls, intellectual-property requirements, employment responsibilities, and approved technology environment. Governance should make safe work easier to recognize rather than leaving every employee to invent a personal risk standard.

Expanded teaching

Expanded teaching

Define permitted tools and purposes

State which AI services are approved for company work, which use cases are permitted, and whether different rules apply to public models, enterprise accounts, internal systems, or connected agents.

Classify information people handle

Give practical examples of information that is permitted, restricted, or prohibited for each approved environment. Include customer records, employee information, financial data, credentials, contracts, proprietary material, and other sensitive business content as applicable.

Define output review

Specify when AI-generated analysis, customer communication, code, financial information, personnel content, or other outputs require checking by a responsible person before use. Review requirements should match consequence, not just content type.

Create an escalation path

Employees should know what to do when they are unsure whether information may be used, when an AI system produces a concerning output, or when restricted information may have been exposed. Silence and improvisation increase risk.

Where AI can help

AI can help draft examples, answer policy questions from an approved policy source, and route uncertain cases. The governing policy and exceptions must be approved by the responsible business, security, privacy, legal, and HR authorities as appropriate.

Verify understanding

A policy is not implemented because it was published. Test whether employees can correctly classify realistic scenarios, identify approved tools, recognize required review, and find the escalation path.

Diagnostic inquiry

Questions to ask, not conclusions

Can an employee facing a normal AI task determine which tool is approved, whether the information may be used, what output review is required, and where to escalate uncertainty?

Questions that distinguish possible explanations

  • Which common employee scenario is most likely to produce an accidental data or policy violation?
  • Do your current rules distinguish tool approval from information approval?
  • How do you know employees understand the policy rather than simply having received it?

Practical application

Practical application

Create a one-page AI Permitted Use Guide covering approved tools, permitted and restricted information, required human review, prohibited actions, and escalation contacts. Test it with five realistic employee scenarios and revise any rule people interpret inconsistently.

Ask Atria

Talk through this Card

Talk it through with us

If this Card describes your week, a free conversation is the fastest way to see what it means for your company and what to do next.

Book a free conversation