Ask the same question of every document

Individual asks one question across every source in a dataset: one answer per source, directly comparable.

Individual is the simplest workflow. One question. Every document. One answer per document. Use it when you want directly comparable responses across a set: not summaries, not extractions, just the answer to this specific question from each source.

What you get

  • One answer per source.
  • Each answer is grounded in its source, with citations.
  • Answers can be text or structured (depending on how you phrase the question).
  • Downloadable as a per-source folder, a single combined PDF, or saved as notes.

Run it

  1. 1

    Build the dataset (a label, a multi-select, or a Smart Search result).

  2. 2

    Open the workflow runner and pick Individual.

  3. 3

    Type the question.

  4. 4

    Click Run.

What kind of question to ask

Individual works best with questions that have a clean per-source answer:

  • "What's the governing law clause?": clear, document-specific.
  • "Who is the principal investigator?": extractable from each paper.
  • "What's the maximum file size accepted?": answerable from each spec.

Less well-suited:

  • "Which of these is best?": that's a comparison, not a per-source question. Open a Focus instead.
  • "Summarize the document.": that's Summarizer, which is purpose-built for summaries.
  • "Extract these five fields.": that's Data Extractor, with type-safe schema.

If the question is really five questions, ask the most important one with Individual, then run again for the next.

Individual vs. a Focus on a label

This is the comparison that comes up most:

Individual Focus on a label
One question, N answers (one per source) One conversation grounded in all N sources at once
Answers don't reference each other Single composed answer spanning sources
Direct per-source comparison Synthesis across the set

Use Individual when the question is "what does each one say"; use Focus when the question is "what do they collectively say."

What's next

Was this helpful?