AI Assisted Work
Define acceptable use of AI tools for drafting, analysis, automation, and sensitive data.
Purpose
AI-assisted work needs boundaries for data entry, output verification, disclosure, and accountability. The person using the tool remains responsible for the decision and final artifact.
Policy shape
| Rule | State the expected behavior in direct language. |
|---|---|
| Reason | Explain the operating risk or obligation behind the rule. |
| Owner | Name who maintains the policy and decides exceptions. |
| Evidence | Define what proves the policy is being followed. |
| Review | Set a review cadence and update trigger. |
When to use
- When the work repeats often enough that memory is no longer reliable.
- When more than one person may request, perform, review, or inherit the work.
- When risk, approval, evidence, or handoff needs to be visible later.
Common failure modes
- Sensitive details are entered into unsuitable tools.
- Generated output is accepted without verification.
- Automation hides who approved the final result.
Review guidance
Review this page after a material incident, after a role or system change, and on a normal cadence appropriate to its risk. During review, check whether the owner is still correct, whether inputs are still complete, whether the output is still useful, and whether related pages need updates.
External guidance
These resources are references for terminology, control thinking, or review design. DailyWF adapts the ideas into lightweight operating pages rather than reproducing full standards.
- NIST AI Risk Management FrameworkAI risk governance structure for mapping, measuring, managing, and governing AI risk.
- CISA Secure by DesignSecure-by-design principles for software and technology evaluation.