Map the information flow
Describe approved inputs, the AI-assisted step, the output and its next user. Identify where private project information would enter and which environment is authorised to process it. This makes the review boundary concrete before documents or datasets are introduced.
Assign checks to the consequence
Routine extraction, exploratory synthesis and a client-facing recommendation do not need identical review. Define which outputs require source checking, numerical verification or a domain expert. Name the person responsible for the final technical conclusion and the conditions that require escalation.
Retain an evidence trail
Keep the source material, output version, material corrections and sign-off decision together. If the source changes, identify the conclusions that depend on it. A traceable workflow helps reviewers understand both what was checked and what remains outside the agreed scope.
Review the workflow after change
A new model, instruction set, data source or use case can change performance. Use selected evaluation cases and observed errors to decide what needs reassessment. Scientific advisory can help define the domain checks and interpretation; software implementation and system ownership require their own agreed roles.
Sources & context
Background reading checked 9 October 2026. The review questions and structures above are BGF editorial guidance; the references are not BGF project evidence or endorsements. Apply them to the relevant site, data and agreed scope.