Reporting / output usefulness

What users said

• Reporting quality was not great.

• They said their own inputs may have limited the result.

• Still, the output did not produce clear action plans.

Underlying requirement

• Reports need to convert interview data into usable decisions, actions, or next steps, even when raw inputs are imperfect.

Implementation guidance for LLM agents

• Separate reporting into layers:

  • raw findings

  • patterns/themes

  • risks/issues

  • action recommendations

• Detect input sparsity:

  • if evidence is weak, say so explicitly

  • do not overclaim

• Generate action-oriented outputs:

  • what happened

  • why it matters

  • what to do next

• Add report quality checks:

  • did each major finding have evidence?

  • did each theme produce at least one implication or action?

• Ask for missing setup information before starting campaigns if reporting quality depends on it

Acceptance criteria

• Reports contain actionable next steps, not just summaries.

• Reports remain useful even when inputs are somewhat thin.

• Weak evidence is flagged clearly instead of hidden.