BW-301c · Module 3
AI-Generated Summaries
6 min read
AI-generated executive summaries are structurally correct and argumentatively empty. This is a precise criticism. The AI will produce a summary with a problem statement, a solution statement, a supporting evidence section, and a call to action. The format will be correct. The language will be professional. The argument will be generic in exactly the ways that matter — it will not advance the claim the document actually makes, because the AI does not know what claim the document should make. It knows what an executive summary looks like. It does not know what your executive summary should argue.
This is not a reason to avoid AI in summary writing. It is a precise description of where the human writer's work begins.
- Understand What AI Is Summarizing AI summarization models extract the most frequently occurring claims and the most explicit conclusion statements from the source text. They are excellent at this. What they are not doing: evaluating which claims are most important to the decision at hand, recognizing the implications of claims that are underrepresented in the source text, or choosing which evidence is most persuasive for a specific reader. If your document's most important finding is buried in Section 5, the AI summary will underweight it. If the most important implication is unstated in the body, the AI summary will not state it.
- Use AI for the Draft, Not the Argument The practical workflow: generate an AI summary of your complete document to establish the structure and identify the claims the AI found most prominent. Read the AI summary alongside your four diagnostic sentences. Where they agree, the AI has done useful work — the structure is sound, the main claims are captured. Where they diverge, the human writer's work is to understand why. Is the divergence because the AI missed something (which you add)? Or because the AI identified a structural problem in the document (which you fix)?
- The Specificity Audit After generating an AI draft, run the specificity audit: underline every claim in the summary that could appear in any document of this type. "The analysis reveals significant opportunities for improvement in operational efficiency." Could appear in any report, for any company, about any process. These sentences are not wrong — they are informationless. Replace each underlined sentence with the specific version: which opportunities, in which processes, by how much, supported by which evidence. The specificity audit is where the human writer's irreplaceable contribution lives.
Do This
- Use AI to generate a structural first draft and identify the most prominent claims
- Run the specificity audit and replace every generic sentence with the document-specific version
- Check the AI summary against your four diagnostic sentences and resolve every divergence
- Ensure the AI summary's ask is specific: the exact decision, the exact amount, the exact date
Avoid This
- Submit an AI-generated summary without the specificity audit — you are submitting the most generic possible version of your argument
- "The following executive summary was generated with AI assistance" as a disclosure that substitutes for quality review
- Trust the AI to identify the most decision-relevant finding — it identifies the most prominent finding, which may not be the same thing
- Use AI output length as a quality signal — a longer AI summary is not a more thorough one