Watch participants do a task, then ask about the moments that matter

Ask people how they shopped for a blender and you get a tidy story. Watch them do it and you see the real one: they started in an AI assistant, opened three retailer tabs and gave up on a filter halfway. Asking about it during the task changes how people do the task. Asking afterward, from memory, brings the tidy story back.

Short on time? Watch the one-minute version.

How it works

• Mark the task. On a question in the topic guide, open Live vision and choose Task observation.

• Tell it what to watch for. Write one guidance box on what matters and how to probe. Under Observation settings, set how many moments to ask about (up to five, three by default) and how many follow-ups each one gets.

• The participant does the task, thinking aloud. They share their screen and see "Your screen is being recorded". The AI research assistant watches without interrupting. If nothing happens, it nudges them to carry on, at most twice.

• Follow-ups are ready when they're done. While the task runs, it picks the moments worth asking about and writes the questions. It skips what the participant already explained and goes deeper on what they only touched on.

Every answer points to the moment it came from

The follow-ups appear in the transcript as "Moment 1 · Task observation", "Moment 2". They count as normal answers, so they flow into analysis, reports and quote reels.

Where it works

Live websites, apps, clickable prototypes and journeys across several sites or apps, in desktop browsers and in the Conveo app on iPhone and Android. Phone browsers can't share a screen, so mobile participants are sent to the app. Camera mode runs the same flow for real-world tasks and shop-alongs.

Ways teams use it

• Category shopping journeys. Where people start (search, marketplace, social, AI tools), how they compare and how they reach the cart.

• AI-assisted purchase research. Which assistant people use, when they switch, which sources they click and whether your brand shows up in the answer.

• Several tasks in one interview. A food-ordering app: find a location, customize an order, sort out a missing item.

• Prototype tests. Whether people notice a new feature in a mobile app and read the pop-up.

• Checkout walkthroughs on a brand's own site.

What this means for your research

You see what people did and hear why in the same video conversation, without interrupting the task.

Know the limits. It asks nothing about what it sees until the task ends. It does not score task success or measure time on task, clicks, heat maps, scrolls or funnels. It can miss moments under about a second or off screen, and can't tell how someone felt, which is why it asks. It only looks at the shared screen, not background tabs or the Conveo window. Nothing on screen is blurred or masked, so ask participants to close anything private first.

Task observation is in beta and switched on per organization. Reach out to your Conveo contact to enable it.

Task observation in one minute

A shopping task, three moments marked, three follow-ups in the transcript.