Qwin: Your AI research partner

Qwin is Xperius's built-in AI research partner — part librarian, part analyst, part first-draft machine. You can activate it across the platform and showcase to surface what you already know, accelerate what you're actively building, and help stakeholders find that needle-in-a-haystack insight they actually need.

That said, Qwin works best when you're in the driver's seat: it opens doors, but you decide which ones are worth walking through. Qwin can process, transform, and change your research if you want it to, but requires human approval for those changes to go-live always.

Qwin lives in a chat panel that opens on the right side of the screen from anywhere in the platform. Look for the floating stars button in the bottom right corner of the screen . You may also use ⌘(Command)-K on Mac or ^(Control)-K on PC.

Inside the Xperius platform (researchers, admins, and collaborators) Qwin works to accelerate your work and also find connections that may be hard to make if you can't remember them. Qwin's context shifts depending on where you open it. Inside a study or session, it's aware of that study's content. On the synthesis canvas, it can reach into your findings and pull connections. Across the platform generally, it can draw on your team's broader repository of published research.

Inside the Showcase (stakeholders) If you access Xperius exclusively through the Showcase portal, Qwin is available to you there too. Your version of Qwin is focused on published research: it can help you find, summarize, and make sense of what your organization's research teams have shared. What it can't do from Showcase is access unpublished studies, session data, or anything outside the published research repository. If it's not published, Qwin can't see it from where you are.

Before you start prompting, it helps to know what you're working with. Here's what each part of the panel does.

Before you kick off a new study, it's worth asking what the org already knows. Qwin can sweep across your repository — studies, insights, findings, reports — from you, your team or others. It can reveal whether this territory has been covered before, who covered it, and what they found. It's the fastest way to avoid re-running research that already exists.

Qwin can translate a body of published research into plain-language summaries oriented around a topic or timeframe — the kind of digest that doesn't require the reader to have followed the research in real time. For stakeholders in Showcase, this is Qwin's home turf.

Working inside a study? You can direct Qwin to analyze attachments for a session. Qwin can operate as a junior researcher and apply the taxonomy defined for the study. Qwin will transcripts, session notes, and others and let you generate findings. All findings generated in these attachments will need to be confirmed to leave their draft mode.

Hand Qwin a research plan document and ask it to build out a study structure and sessions around it. Reference patterns that a study should follow. Qwin will give you a starting point; you bring the judgment about whether the sessions actually match your methodology and your participants.

When building a new collection in Showcase, you can enable Qwin-powered curation to let Qwin draft the contents for you. Give it a theme and a project team, and it'll identify the published items that fit and build a first draft. You review and refine from there.

Qwin supports integration with external tools via MCP (Model Context Protocol). This means you can access all the research data you have accumulated in Xperius directly through your LLM of choice. Much of the prompting advice and tips will still apply in this environment. Your LLM may have additional guidance for working with external integrations such as Xperius.

Once set, the LLM you have selected becomes the one that is prompted in the chatbot on the product platform as well.

Be specific about what you want, not just what you're looking for. "Tell me about onboarding" is going to get you a general response. "Summarize what our research has found about new user onboarding friction, focusing on mobile, and format the output as bullet points by theme" gives Qwin a real target.

Scope it or Qwin will scope it for you — probably too broadly. Add a timeframe ("in the last year"), a product area ("for our mobile app"), or a team ("from the enterprise research team") to pull more focused results. The more specific your scope, the less noise in the response.

Tell Qwin what format you want. Want a bulleted list? Say so. Want three insight statements? Ask for exactly three. Want a table comparing findings across studies? Describe the table. Qwin follows formatting instructions well; use that.

When you have source material, make sure Qwin has it in context. If you want Qwin to work from a specific study or document make sure you are on that page, reference it in the prompt, or attach it to the conversation.

Treat the first response as a draft, not a deliverable. Qwin is good at first passes. It's not always good at knowing when to stop. If the first response is close but not quite right, push back: "The second and fourth points aren't relevant — revise with a tighter focus on [topic]."

These prompts are ready to use. Copy one, paste it into the Ask Qwin input, and hit send.

Qwin returns "I could not process that request: Failed to fetch" or a similar error. This usually means Qwin couldn't reach the data it needed — a temporary connectivity issue or a backend hiccup. Wait a moment and try again. If it keeps happening, reload the page. If the error persists after reloading, contact your admin or submit a support request.

Qwin's response doesn't match the research you know exists. Qwin draws from published, indexed content — if research wasn't published, wasn't tagged well, or was ingested recently, Qwin might not surface it. Try rephrasing your prompt with more specific terms (study names, topics, timeframes), and manually verify against the repository if the stakes are high. Don't rely on absence of results to mean absence of research.

Qwin's output seems shallow or generic. This almost always means the prompt was too broad. Add specificity: a topic, a timeframe, a scope, a desired format, a point of view you want explored. If you're getting surface-level answers to deep questions, you're asking for a skim when you need a dive.

An attachment doesn't process as expected. Make sure the file is fully uploaded before you send your prompt — look for confirmation that the attachment loaded in the Ask Qwin panel. If the % complete loaded is visible, it isn't yet attached. If Qwin returns a response that ignores the file entirely, try re-attaching and reprompting. If a specific file type consistently fails, check whether the format is supported (i.e., GIF, JPG, PNG, WEBP, MP4, PDF, DOCX, TXT, PPTX).

Qwin navigates to a part of the platform you didn't expect. Qwin can perform actions and navigate within Xperius — that's by design. If it takes you somewhere you didn't intend, use your browser's back button or return to your previous location via the breadcrumb. To prevent unexpected navigation, be explicit in your prompt: add "do not navigate, just give me a response" if you want a text answer rather than an action.