Why AI-generated reporting still needs a consultant

A missing table, wrong date range or incomplete export can create a convincing but incorrect narrative. Check that the source is complete and that comparable periods really are comparable before asking a model to explain the result.

A useful first pass records the facts without assigning causes. Revenue declined is an observation. Revenue declined because creative is tired is a hypothesis that needs more evidence.

Promotions, stock, margin and sales capacity can explain why a sensible decision looks inefficient in a platform report. The consultant needs to bring that context into the analysis and challenge conclusions that do not fit the wider picture.

Our experience shaping the commentary process has reinforced the value of a person reading the report and articulating what stands out. AI can help structure those observations; it should not replace the act of deciding which observations matter.

Lead with the commercial result. Explain the few movements that change a decision. Separate known causes from possible explanations, then state the proposed action and what evidence will be reviewed next.

When feedback changes an interpretation, revise that section while preserving the source and earlier reasoning. An approval loop should improve the output, not force everyone to restart the process.