The AI assistant can write content for you and put it straight where it belongs: as cards on a board, as sections in a document, or as an answer in the chat.

Where to generate

Open the assistant with the AI button in the top navigation. There is one assistant for the whole app, and what it can do adapts to where you are:

• On a board: it creates and edits real canvas elements.
• In a document: it writes into the document itself, section by section.
• On the dashboard: it answers and plans across your projects.

What it can generate

• Text: scripts, outlines, summaries, descriptions, section drafts
• Ideas: concepts, angles, alternatives, names
• Structures: plans, frameworks, checklists, schedules, breakdowns
• Rewrites: shorter, longer, different tone, different audience

Getting better output

Give it something to work from. The single biggest quality difference is context. A board that already holds the brief, the transcript or the reference material produces far better content than a prompt describing them from scratch.

Attach the right things. Use the + button in the chat composer to add context or a Tactic to your message, and select elements on the canvas before you ask so "these" and "this" mean exactly what you intend.

Ask it to research first. For anything factual, current or dated, telling it to look things up before writing gives you specifics instead of generalities.

Say what "good" means. Length, tone, audience and format are the four things worth stating up front. "Three bullets, plain language, for a client who has not read the brief" gets a usable answer the first time.

Tips and tricks

Iterate on the card, not in the chat. If a card is nearly right, ask for the change directly ("shorten this to two sentences", "make this friendlier"). The assistant edits the existing element rather than making a second copy beside it.

Generated content is normal content. Everything it writes is a real element or document section you can edit by hand afterwards, and undo works the way it always does.

Good to know

Treat the first output as a strong draft rather than a finished one. The assistant is quickest to improve when you tell it what specifically is wrong with what it just gave you.