Regulating the use of generative AI in the workplace
Specify permitted inputs, prohibited uses, reviewers and incident handling. Distinguish internal assistance from public release and provide concrete examples.
Putting ai workflows into practice
Generated output may omit information or introduce unsupported statements. Separate model output, human approval and execution; a recommendation does not grant release authority.
Checking the result
Test both sufficient and incomplete inputs and compare accuracy, omissions, rework and stopping behaviour. Verify sources and commitments before public release.
