Severity you can defend.
Watch real users attempt tasks in your product, live on a call.
- For
- Teams running moderated usability sessions — the same activity as internal testing, with a different person at the keyboard.
- Leads with
- Invite to call
- Preset goals
- 5 — the most a project holds, all editable
- Also uses
- The Sense Bot notetaker on Zoom, Meet, and Teams
What someone does and what they say happen at the same moment, so the ways in capture the session itself — screen and voice together.
A usability finding is worthless without the moment it came from. Someone hesitating for nine seconds and then giving up is the evidence; a note reading "invite flow is confusing" is not. So capture keeps the screen and the voice on the same clock.
The capture routes this project puts in front of you.
All seven capture methods work in every Sense project. This template only decides the order — and for user testing, that order starts here.
Paste a meeting link and the Sense Bot joins the Zoom, Meet, or Teams call, records it, and transcribes it.
Capture your screen and your voice together, so what you say lands next to what you did.
Record a voice note in the browser — the fastest way to get a thought down right after a session.
Drop in audio, video, or a document you already have and we will transcribe it.
5 preset goals, written for this situation.
A goal is an extraction prompt: it tells the AI what to pull out of every transcript in this project. These ship with the template and are yours to rewrite — a project holds up to five.
How each attempt ended: unaided, with a hint, abandoned, or failed without the person realising. Records what happened, not whether the design is good.
Stalls, hesitation, backtracking, repeated actions; what they searched for and where; whether they recovered. A pause is not an abandon.
The gap between the mental model and the product — predictions said aloud, surprise, and the words they use for things your interface names differently.
The long way round, features used for something they were not built for, and leaving the product entirely to get the job done.
Unprompted delight, frustration, confusion, and explicit asks — with the trigger and the underlying need. Weighs what they did over what they said they would do.
One line becomes an action item, without losing the words.
Nothing here is a summary. Each hop keeps a link back to the one before it, so the item at the bottom still opens onto the sentence at the top.
- 01Capture
The meeting link goes in ahead of time and the Sense Bot joins the call, records it, and transcribes it. A moderated session, one task in.
DRSKRECSenseSense Bot joined the call - 02Signal
It lands as a transcript — timestamped, searchable, with the recording attached.
0:04LenaGo ahead and try to invite a teammate to this workspace — talk me through what you're thinking.0:12OmarI'd look under Settings. It's not there. Maybe under the workspace name? No. I'd probably just email them at this point.0:29OmarOh — it was in the sidebar the whole time. I didn't read that as a people thing. - 03Anchor
Read against this project's 5 goals, the passage that answers one gets lifted out verbatim and tagged. The anchor keeps its offsets into the transcript.
0:04LenaGo ahead and try to invite a teammate to this workspace — talk me through what you're thinking.0:12OmarI'd look under Settings. It's not there. Maybe under the workspace name? No. I'd probably just email them at this point.0:29OmarOh — it was in the sidebar the whole time. I didn't read that as a people thing.Tagged toWhere they got stuck - 04Insight
Anchors group into a finding. One insight often draws on more than one goal.
People look for invites in Settings first; the sidebar entry doesn't read as people-related
3 anchorsWhere they got stuckExpectation mismatches - 05Action item
The finding becomes a backlog item on the workspace kanban, still linked to the anchor it came from.
Add an invite entry point to workspace SettingsTodoTraces back to Omar at 0:12
Every session becomes a signal, and the usability problems land in your backlog ranked by what actually blocked people.
Action items live on the workspace backlog, not inside a single project — so a fix that came out of a testing session sits next to one that came out of an interview, each carrying its own evidence.
Stop arguing about design in the abstract.
- Findings arrive ranked by what actually blocked people, with the stall itself as the evidence — not a reviewer's opinion about the layout.
- Five goals, the fullest set of any template, because a session produces more kinds of evidence than any other capture: outcomes, stalls, mismatches, detours, and asks.
- What someone did outweighs what they said they would do. "I'd probably just email them" is recorded as the detour it is.
All seven capture methods are available in every project. A template only decides what you see first.
Rewrite, remove, or add to the preset goals at any time. A project holds up to five.
A new workspace is seeded with one project per template. Nothing is locked behind a choice you make on day one.
Every anchor carries its citation offsets into the transcript, so each claim traces back to the words that were said.
Every workspace starts with this project already in it.
Create a workspace and all six projects are seeded for you — this one included, goals and capture order set up. Nothing to configure before your first capture.
Your first run in this project ends on “Paste a session link”. The project tour spotlights the composer, ready for the link to your next session.