Start with the source, not a blank conversation

Imagine evaluating a new email tool. Its product page has a description, screenshots and a feature list. Copying the page into a separate chat loses the relationship between a question and the exact item you were reading.

In dodu, shake to enter selection mode, choose the relevant product card and ask: “Who is this for, and what should a five-person team verify before a trial?” The selected context gives the question a concrete starting point. If the linked page cannot be read, the answer should be treated as limited to the available context.

Ask for a decision aid, not a confident verdict

A useful brief separates source claims, open questions and next actions. Ask for three checks you can actually perform: whether your email provider is supported, which permissions the tool requests and how pricing changes with team size.

AI output can be wrong or incomplete. Revisit the source for dates, prices and important claims. An agent is most helpful when it reduces the work of checking a decision, without pretending the check is already complete.

An answer is only one step

Save the brief as a PDF in SPACE. Take a separate full-page screenshot when you need the source layout. Later, open Workspace, preview the PDF in SPACE and move to DODU to turn the reviewed information into a trial checklist.

The landing demo shows these steps with real app captures and real saved output. Its replay is edited to explain the interaction; it is not a live model run or proof of autonomous account access.

Keep assistance easy to enter and easy to leave

Not every visit needs a chat. On-demand AI lets you ask one question, keep the useful result and return to the page. Longer conversations belong in Workspace. This separation keeps quick browsing help close to the content and gives deeper work a place to continue.