Getting good answers

Asking Good Questions#

Better questions yield better answers. Aim for:

• Specific target: “In risk_register.xlsx, list mitigations for data loss”.
• Clear action: “Summarize”, “Compare”, “Extract”, “List key risks”, “Outline steps”.
• Narrow scope: Reference a file name or a concept instead of the whole corpus.

Avoid questions like “Give me all the text of every document”. Ask for summaries, specific sections or the passages that matter instead.

With models that support agent mode, AITEZA can list the files that mention a term, with the number of matches per file. With other models, answers draw on the most relevant passages only. See Choosing a model.

Targeting Specific Files#

Mention the file name to pull from it:

• “In contract.docx, summarize termination clause.”
• “Compare proposal_v1.docx and proposal_v2.docx on pricing.” If names are similar, add context (“the PDF version” or a folder hint). Use backticks for clarity.

Hybrid Search (Why It Matters)#

AITEZA combines meaning (semantic) and exact word matching (keyword). This means:

• You can find where a term is discussed (“Where is ‘data residency’ mentioned?”)
• Misspellings or wording differences are handled intelligently
• Direct keyword queries still work if the term is unique

In agent mode, AITEZA also searches for exact phrases across all files and then reads the relevant files in parts.

Tips for Better Results#

• Clean digital documents give the best results. For scans without a text layer, reprocess the file with OCR text recognition; for complex layouts and tables use Visual document analysis. See Processing Pipelines.
• Use follow-ups: “Expand on the risks you listed” instead of a new broad question
• Combine semantic + keyword: “In strategy.docx, what actions support ‘market expansion’?”
• Break complex requests into steps

FAQ#

Q: Can I ask for the full content of a file?
A: Request summaries or specific sections instead. That keeps answers concise and relevant.

Q: How do I make sure it looks at a specific file?
A: Include the exact file name (backticks help). Add context if names are similar.

Q: Why didn’t it list all documents with a word I searched?
A: Lists are complete only with models that support agent mode: AITEZA then searches every file for the term. With other models, only the most relevant passages are used. Choose a model with agent mode (see Choosing a model).

Q: Can I use file attachments while connected to a dataroom?
A: Yes. You can attach files and images to a message while datarooms are connected. Web search cannot be switched on then, but AITEZA can ask you before each single web search (see Questions from AITEZA).

Q: Can I connect multiple datarooms at once?
A: Yes. You can connect one or more datarooms to any chat and query across them simultaneously.

Q: My PDF is a scan. Why do I get poor answers?
A: Reprocess it with OCR text recognition. If tables or the layout still come out wrong, use Visual document analysis (VLM).

Q: How current is web content?
A: AITEZA uses the pages of the websites added to the dataroom, as of their last scan. Rescan a website by hand, or set up automatic updates (see Websites and images).