Gen3 Workflows Document Review
- Create a dataset for the documents you want to review together.
- Upload your documents to that dataset (Uploading and Importing Documents).
- Open GT Chat and select an agent configured for document Q&A or RAG over datasets.
- Select the paperclip in the chat message bar at the bottom of the screen.
- Select the dataset you want the agent to reference for this conversation.
- Send a prompt and ask questions about the documents in that dataset.
Document review is one of the most common tenant workflows: operators need repeatable retrieval over a bounded corpus, not one-off pasted text. Datasets hold the source material; agents define how answers are formed; chat attaches the right dataset scope per conversation so retrieval stays predictable.
- Confirm documents finished processing in the dataset (pending ingestion blocks the next chat message).
- If answers miss content, check Managing Dataset Content and the agent’s default dataset attachments in Building Agents.
Use an agent intended for document Q&A—one with dataset/RAG instructions and retrieval tuning. Generic chat agents may work but often produce weaker citations and less predictable retrieval over uploaded files.
The paperclip opens Add Datasets to Conversation. You can attach multiple datasets when the review spans collections; keep scope minimal so citations and retrieval stay focused.
- Ask for summaries, comparisons, or gap analysis across uploaded files.
- Reference document types explicitly (“compare the SOP PDF to the checklist DOCX”).
- Use Reviewing and Exporting Conversations when you need an audit trail of the review thread.
When GT API is enabled for your account, automation pipelines (scheduled ingestion, external tools) should use GT API dataset upload keys and published agent aliases rather than manual chat uploads.