Totem

How it works

One line in. Synthesis out.

No metaphors, no theatre. Here's what actually happens between the moment you type a prompt and the moment the synthesis lands on your desk.

Silent betaFully free in beta

Already convinced?

Skip ahead, drop your prompt

The five steps

From your line, to the synthesis on your desk.

  1. 01

    Describe the conversation in one line.

    You type what you want to learn and who you'd ask. Totem doesn't need a script, a discussion guide, or a slide deck. The line in the prompt box is the brief.

  2. 02

    Totem designs the scenario.

    It generates the question set, the follow-up probes for vague answers, the structure of the call, and the rubric for synthesis. You can edit any of it before invites go out, or hit launch and let it run.

  3. 03

    Participants join when it suits them.

    Each participant gets a link. They show up async, on voice or video, in their language. No calendar. No 30-minute Zoom. Most sessions land between 10 and 20 minutes, on the participant's schedule.

  4. 04

    Totem hosts every conversation.

    Asks the questions. Listens. Probes "fine" until it's specific. Reads the room enough to skip questions that have already been answered. Records everything, transcribes verbatim, cites every line back to the speaker.

  5. 05

    You wake to the synthesis.

    Themes ranked by how often they came up. Verbatims that prove each one. The people behind every signal, with the moments that mattered. On your dashboard, minutes after the last call ends.

What you wake to

Step 05, in one screen. The synthesis.

Themes ranked by how often they came up. The verbatims behind each one. The participants behind every signal. The room you didn't have to be in.

Onboarding Friction, Week-1 Users

Onboarding research. Totem ran this conversation across 14 participants and synthesized what came back into the takeaway below.

14participants

Key findingsAI-synthesized from 14 participants

The empty state is the bottleneck, 10 of 14 stalled there before finding their first use case. Six found templates helpful once discovered; four missed the Templates tab entirely. Three said they'd have bailed without "see an example", and five want a guided first-workflow tour from cold start.

  • Stalled at empty state
  • Templates helped
  • Missed Templates tab
  • Wants guided first-run tour
  • First workflow in week 1
Sentiment mix
  • 40%Positive
  • 40%Friction
  • 20%Focus
Responses
0
Fully completed
0/0
Avg. completion
0%
Avg. length
0m

Under the hood

What Totem does, and what it doesn't.

  1. 01

    Every claim cites a participant.

    If a theme appears, you can click it to see the verbatims behind it. No synthesis without sources; no quotes without speakers.

  2. 02

    Totem doesn't impersonate.

    The host introduces itself as Totem at the start of every call. Participants know they're talking to an AI. No deception about who they're talking to.

  3. 03

    Recordings stay yours.

    Every call is recorded with consent. The recordings live in your dashboard, exportable, deletable. Nothing trains a foundation model.

  4. 04

    It works in any language.

    The host speaks the participant's language. The synthesis is delivered in yours. Translation is part of the run, not a separate step.

  5. 05

    What it isn't great at yet.

    Group conversations with multiple participants on the same call (the host is one-on-one for now). Sessions above 75 minutes (the probes get less crisp). Regulated audit-grade trails for HIPAA-bound work, we're working on it; talk to us if that's you. We'll tell you when something doesn't fit, not after.

Common questions

How Totem works, in your questions.

How does Totem work?
Five steps: (1) describe the conversation in one line; (2) Totem designs the question set, probes, and rubric; (3) participants join async on their own schedule; (4) Totem hosts every call, asks the questions, probes hedges, transcribes verbatim; (5) the synthesis lands with themes ranked, verbatims cited, and the people behind every signal.
What does an AI-moderated conversation actually feel like?
Like a recruiter call, not a video form. The participant speaks naturally; Totem asks open questions, probes "fine" until it's specific, follows the participant down a thread when the thread is interesting, and skips questions that have already been answered. Most sessions land between 10 and 20 minutes, on the participant's schedule.
How long does setup take?
Setup-to-first-result is under a day in most studies. The prompt is the brief — no question guide, no slide deck, no recruiter handoff. You can edit Totem's generated protocol before launch or hit launch and let it run. The longest part is reviewing the synthesis.
What does the synthesis look like?
Themes ranked by participant count ("7 of 9 mentioned tier confusion"), sentiment-coded, with verbatims grounding every claim. The synthesis fits in a Slack message; the depth is one click away. The output is one cross-participant takeaway per study, not one summary per call.
What does Totem not do?
Totem doesn't recruit participants — you bring them. It doesn't replace live moderated interviews when novel-rapport-required depth is needed. It isn't HIPAA-certified yet, so don't enter PHI in healthcare settings without your own compliance review. It doesn't analyze video for body language; the conversation is voice-first.
How is Totem different from a survey or interview-AI tool?
Surveys collect closed-ended answers; AI-meeting-recorders capture calls you already made. Totem hosts the call itself: AI-moderated, async, with adaptive probes that match what each participant says. Different unit of work, different unit of output.

Your turn

Now you've seen how. Try the line.

Drop a prompt. Totem will set up the conversation, run it with the people you point it at, and wake you with the synthesis. Free during the beta.

Drop your prompt

Free during the betaNo cardYour prompt, your data