Getting Comprehensive Results from Large Document Sets#

When working with large document sets, AI performs best when you analyze documents in focused batches (by document type, incident, or topic) and then synthesize the findings, rather than asking it to analyze everything at once.

You can use this strategy in both the Search bars and the Report page.

Step 1: Analyze Each Batch#

Use this prompt for each batch. The more specificity the better. Replace anything in brackets with your specific case details (and remove brackets when you're done):

"Cross-reference all [e.g. witness interview transcripts / incident reports / emails / personnel files / policy documents] for evidence that tends to [support / contradict] the allegation of [e.g. mistreatment / harassment / discrimination / safety violations]. What evidence do multiple [e.g. interviews / reports / emails / documents] agree on? Where do the [accounts / statements / records] differ or contradict each other? Are there patterns in what [e.g. witnesses / employees / managers / complainants] report or state? When [accounts / statements / records] contradict, note the specific differences. Document your analysis process."

Step 2: Synthesize Across Batches#

After analyzing each batch, use a synthesis prompt:

"Cross-reference your findings from [e.g. the witness interviews, incident reports, and emails / all document types you've analyzed]. Where do different document types corroborate each other? Where do they contradict? What's the overall picture when all sources are considered together?"

By batching documents into logical groups and then synthesizing, you can ensure more comprehensive coverage.