Study first
Review the ideas behind the questions
Review the AI-assisted workflow before using generated work. Focus on the task, inputs, review criteria, source checks, risk level, and final human approval.
Define The AI Job Before The Prompt
A useful AI workflow starts with the work decision, not with a blank prompt box.
- AI risk work should start by understanding the purpose, context, users, expected benefits, costs, and risk tolerance.
- Documentation should explain how AI output may be used and overseen by people.
- The business value and context of an AI use should be clearly defined or re-evaluated.
In Practice
Brief First
A good prompt cannot fix an unclear task. Define the audience, source material, output format, reviewer, and decision before generating.
Use Case Limits
A low-risk brainstorming workflow can have a lighter review path than customer-facing claims or personal-data work.
Common mistakes
Starting with a generic prompt before agreeing what the output will be used for.
Define the task, context, use, knowledge limits, and reviewer before generating output.
Q&A
What should a marketer define before using AI for a campaign task?
Define the task, source inputs, audience, risk level, output use, reviewer, and approval path.
When should a team use a stricter AI review path?
Use stricter review when the output affects claims, customers, personal data, brand trust, or a high-impact decision.
Test Generated Work Against Evidence
AI output should be checked against the task and source evidence before it becomes campaign work.
- AI systems should be tested before deployment and regularly while in operation.
- Measurement should use knowledge of the identified risks and inform the decision about how to manage them.
- Advertising claims must be truthful, not deceptive or unfair, and evidence-based.
In Practice
Generated Is Not Reviewed
Treat output as a draft until factual claims, sources, audience fit, and brand risk have been checked.
Claims Need Proof
If generated copy adds a performance, product, or comparative claim, the reviewer needs evidence before it ships.
Common mistakes
Publishing generated copy because it sounds fluent and on-brand.
Check clear sources, factual claims, audience fit, brand risk, and approval status before publishing.
Q&A
What is the safest first review of AI-generated campaign copy?
Check whether every factual or comparative claim has evidence and whether unsupported claims are removed.
Plan For Change And Errors
AI workflows need monitoring because risks, context, and outputs can change over time.
- AI risk management should be continuous, timely, and performed throughout the AI lifecycle.
- AI risk tracking should cover existing, unanticipated, and emergent risks based on intended and actual performance.
- When AI outcomes are inconsistent with intended use, responsibilities should exist to supersede, disengage, or deactivate the system.
In Practice
Freshness Matters
A workflow that worked for one campaign may need a new review when the audience, sources, claim, or risk changes.
Pause Is A Control
If outputs become unreliable, the team should pause or change the workflow rather than lower the review standard.
Common mistakes
Keeping the same AI workflow after repeated factual errors because it saves time.
Pause or change the workflow, document the issue, and review whether the risk now exceeds tolerance.
Q&A
What should happen when an AI workflow repeatedly creates unsupported claims?
Pause the workflow, document the issue, tighten review, and only restart when the risk is controlled.