Question or research?
A quick answer with links, or a document checked by a second agent: attachments, delivery, export, time and cost.
A space offers two ways to get an answer from documents. “Ask” — one AI agent reads the documents it needs and answers in a conversation with links. Deep research — two roles: the analyst writes a document, the checker verifies every claim against the sources and sends it back for rework until everything matches.
How to choose
| Ask | Deep research | |
|---|---|---|
| What you get | an answer in a conversation with links to documents and knowledge base entries | a document with links and the discrepancies between sources named |
| Who does the work | one AI agent | an analyst and a checker: every claim is verified against the sources by a second agent |
| When it fits | one question about one or two documents; “does A match B” | bringing several documents together, finding all the discrepancies, preparing something that will be forwarded and kept |
| Attachments | yes, up to 10 MB per file | no — attach the file to a question in “Ask” or put it into a connected source |
| Clarifications | continue in the same conversation; the agent may ask a counter-question itself | a conversation for refining the direction as it goes; the document grows before your eyes |
| Where it arrives | in the conversation; when requested from Notion, a Google Doc, Confluence, Figma or a ticket — as a comment in the same place | in the app; requested from Notion — as a child page, from Google Drive — as a new document in the same folder, from a ticket — as a PDF attachment |
| Export | text in the conversation | PDF and Markdown |
| Time and cost | faster and cheaper: as much is read as one answer needs | slower and more expensive: two agents re-read the sources several times |
There are deliberately no figures here: the price depends on the plan’s model and on how much had to be read, and it is shown in the conversation for each question — see What a question costs.
Examples
| Question | What to choose |
|---|---|
| “What first-response time does the contract promise a client on the ‘Business’ plan, and does it match the FAQ?” | Ask |
| “What is the number of the contract with sole trader Kungur, and until when is it valid?” | Ask |
| “Compare the three versions of the commercial proposal for Romashka LLC by price, discount and term” | Research |
| “Which 2026 contracts have late-payment penalties, and at what rates?” | Research — many documents, and the result will be kept |
| “Combine the statuses from three departments’ September reports and name the discrepancies” | Research |
| “Check this text against the ‘How we talk about the product’ document” — with the text as an attachment | Ask — research does not accept attachments |
The rule
If the answer fits on a screen and you will read it yourself — “Ask”. If the result will be forwarded or kept, or it has to reconcile several documents with each other — research. In doubt, start with a question: the answer will show how many documents are involved, and it will become clear whether you need a document.
Deep research
Two agents prepare a verified document from code or company documents: the analyst writes, the reviewer checks every claim against the sources.
Ask from a document or ticket
How to address the AI agent in Notion, a Google Doc, Confluence, Figma or a tracker: forms of address, where the answer arrives, what the bot does not do.