Skip to main content

Reviewing AI Output

AI-generated content requires your professional review before it becomes part of any audit deliverable. Here's how to review effectively.

General principles​

  1. Read everything the AI produces — Don't accept blindly
  2. Check citations — Click through to verify claims match the source
  3. Validate completeness — Did the AI miss anything important?
  4. Check for accuracy — Are facts, figures, and references correct?
  5. Apply judgment — Does the conclusion make sense given what you know?

Reviewing risk analysis output​

When reviewing AI-extracted risk items:

  • Relevance — Is this risk relevant to the audit scope?
  • Categorization — Is the risk category correct?
  • Risk level — Does the assigned level (High/Moderate/Low) match your assessment?
  • Rationale — Does the explanation make sense? Check the cited sources.
  • Completeness — Are there risks the AI missed that you know about?

Reviewing generated reports​

When reviewing AI-drafted report sections:

  • Factual accuracy — Do the stated findings match your actual test results?
  • Tone and language — Is it appropriate for your audience?
  • Citations — Are claims properly supported by source documents?
  • Completeness — Are all significant findings covered?
  • Consistency — Does it align with your other report sections?

Reviewing audit programs​

When reviewing AI-generated objectives and procedures:

  • Coverage — Do the procedures adequately address the identified risks?
  • Practicality — Are the procedures feasible to execute?
  • Sample sizes — Are suggested sample sizes reasonable?
  • Attributes — Do the testing attributes capture the right checks?

Editing AI output​

You can always:

  • Edit inline — Modify any text directly
  • Regenerate — Get a fresh draft for any section
  • Discard — Remove AI content and write your own
  • Supplement — Add your own content alongside AI text

AI limitations​

Be aware that AI may:

  • Miss context that isn't in the uploaded documents
  • Use slightly different terminology than your organization's standards
  • Generate content in a style that doesn't match your voice
  • Occasionally misinterpret complex regulatory requirements

Your expertise fills these gaps.