Using Ask
The full Ask reference — context awareness, modes, agent sources, web search, slash actions, and citations.
Ask is the resizable AI chat column to the left of your lists and document tabs. It answers from your project's research and experience graphs, not from a blank model chat.
Context awareness
Ask adapts to what you are looking at:
- List focus — questions cover the whole Research or Experience tab.
- Detail focus — with a document tab open, questions target that note, research item, or experience entity.
- Suggestions — starter prompts change with the current surface.
You never paste document text into the chat; the focus context is sent with each question.
Modes
| Mode | Behaviour |
|---|---|
| Auto | Ask routes the question to the research graph, the experience graph, or both |
| Research | Evidence-first answers from files, notes, highlights, and tables |
| Experience | Answers about pages, features, content, menus, and structure |
Pick a model for the turn
The composer's model control offers Fast for everyday work, Sharp for harder reasoning, and Capable for a deeper pass on Pro and Teams plans. Fast is the default; switch up when a question deserves more thinking and back down when you are iterating. The choice sticks per project and can change mid-thread. See AI requests and models.
Agent sources
The Agent sources control in the composer lets you choose which knowledge Ask draws on for a turn — useful for narrowing an answer to specific material rather than the whole project.
Web search
Toggle Web search on when a question needs current information from outside your project. Live web results are attached to the turn and cited separately, so you can always distinguish them from your own evidence. See AI requests and models.
Slash actions
Type / in the composer for direct actions. Each one is a skill: it phrases the request properly, asks for a format the Agent can save, and puts save controls under the reply. The menu is filtered to what makes sense where you are, and typing after the slash searches it.
| Skill | What it does |
|---|---|
| Analyse & Label | Scans the open item for highlights and drafts an Analysis frame — see Analyse & Label |
| Create types from open item | Pulls the questions, assumptions, evidence, and gaps already written in a note into saveable research items |
| Generate note / Generate notes | Drafts one focused research note, or several from a single brief |
| Generate table | A comparison matrix or scorecard you can save as a research table |
| Generate persona / Generate personas / Generate audiences | Audience and persona work grounded in your evidence |
| Generate journey / Generate user flow | A journey or a step-by-step flow you can save to Research |
| Generate IA map / Generate page from URL | Menus, pages, and hierarchy for the experience graph |
| Visualise diagram | An infographic of a journey or a table |
| Usability scan | Scores a live page against the usability benchmark — see Usability scan |
| Break down a goal | Milestones, the activities under them, and the to-dos that deliver them — see Plan with the Agent |
| Generate to-dos | Concrete next steps you can add to the project list |
| Summarise / Ask a question | A concise read of what is known, or a plain question against the current focus |
They are a faster route to a specific outcome than phrasing a request in prose.
Mention an item with @
Type @ to pull something specific into the question and it is inserted as a chip that travels with the turn as context. The picker covers:
- Research items — files, notes, personas, journeys, tables, user tests.
- Experience entities — pages, features, content, menus, listed with their short ID.
- Milestones and activities — each activity shows the milestone it sits under, so you can ask about a piece of work and the plan around it in one turn.
If what you need is not in the project yet, Upload a file sits at the bottom of the same menu: pick a file from your computer and it is uploaded to the project library and inserted as a chip once it has landed, without leaving the conversation.
Chat history
Conversations are kept, and each workspace surface remembers the thread you were last on there — along with anything you typed but never sent.
- History in the Agent header searches everything you have asked in the project. Threads held on the view you are on now are listed first, under This view, with the rest under Elsewhere in this project.
- New chat starts a clean context when you want a fresh line of questioning rather than a continuation.
Citations
Every answer lists the items it drew from under Sources. Follow a citation to verify the claim in context before you rely on it.
Next
Continue with Research and experience graphs, or jump ahead to Applying and materialising answers.
Related articles
Documents: Ask