Skip to content
Earshot
Sign in

Your customers will tell you everything.Someone has to ask.

Earshot is an AI interviewer. You write the questions; it talks with each customer out loud, follows up where an answer is thin, and brings back what they said, sorted by question and backed by their own words.

One sentence is enough.

Free to start: 30 minutes of interviews included. No card.

A sample of an interview in progress, shown without sound. Interviewer: What made you start looking for a new tool? Participant: Honestly? The reports. Every Monday I lost two hours pulling numbers together by hand. Interviewer: Two hours, every Monday. What happened the week you finally had enough? Participant: My manager asked for a number in a meeting, and I didn't have it. Interviewer: What did you need to be able to say, right then? Participant: Just where we stood. One screen I could trust.

Interviews in 12 languages0 to 4 follow-ups a question, as you set themevery quote read from the transcript, never written by a model

How does it work?

Ten interviews in the same hour, and nobody booked a slot.

Talking to customers is the best research there is, and the first thing a busy week drops. Earshot holds the conversations for you, several at once, at whatever hour suits each person. Below, one study from its first sentence to its findings.

  1. Write what you want to know

    Start from one sentence and the assistant drafts the guide: topics, questions, and how far to dig on each. Or write it yourself. You decide what counts as a full answer, and where a follow-up is worth the time.

    1. 1Must-ask questions are never skipped.
    2. 2The editor estimates the time as you write, so the guide fits the call.
    3. 3Tone, voice and language are set for each study.
    Fig. 2. One sentence, and the guide drafted from it: a topic, its questions, and how far to dig on each.

  2. Share one link

    No calendar, no app to install. Each customer opens the link when it suits them and talks for as long as you asked, in their own language.

    1. 1They are told up front that it is an AI, and that the call is recorded.
    2. 2They can type instead of speaking; captions are on.
    3. 3The study closes itself at the number of interviews you set.
    Fig. 3. The study's link, and the interviews that came in through it: at night, over breakfast, in a lunch break.

  3. Read what they said

    Every interview comes back with the answers already sorted under your questions. The transcript and the recording sit under each summary, for when you want the whole conversation.

    1. 1Each summary shows the words it rests on, and where in the call they were said.
    2. 2How much of the guide each interview covered is counted.
    Fig. 4. Ana's interview: the recording, and under each question a summary of her answer with the words it rests on.

Why not send a survey?

It asks why, and knows when to stop.

A survey takes the first answer. A good interviewer hears that it was thin and asks for the story behind it. Every question in your guide carries a follow-up budget, so the time goes where the answers are worth it. Try it on one.

Follow-ups on this question

Up to two follow-ups to get an example and the why.

You learn that the dashboard isn't trusted.

The question

Interviewer: How do you decide which reports to send each week?

Participant: It depends. Mostly whatever the team asks for.

Follow-up 1 of 2

Interviewer: What did they ask for last week?

Participant: Churn by plan. Again.

Follow-up 2 of 2

Interviewer: Why does that one keep coming up?

Participant: Because nobody trusts the number on the dashboard.

Moves on

Interviewer: Thank you. Let's talk about how you share them.

Fig. 5. One thin answer at four depths. Every question in a guide carries its own.
It keeps to time.
It knows how long is left and what is still uncovered. Running late, it trims follow-ups, drops optional questions and closes properly.
It waits.
A pause to think is not the end of an answer. It gives silence room before it speaks, and keeps its own turns to a sentence or two.

Can I check the findings?

Every finding leads back to something a person said.

Across a study, Earshot writes up what the interviews say together: themes stated as claims, signals too weak to call, answers to your own questions, and what the interviews can't tell you. Each quote opens the moment it was said.

Fig. 6. The write-up across seven interviews. The small numbers refer to the three rules below.
  1. 1

    A theme with no quote behind it is dropped.

    What a model says the interviews mean is not kept unless a participant's own line supports it.

  2. 2

    Quotes are read from the transcript.

    The analysis points at a line; the words you see are fetched from what was said, never written anew.

  3. 3

    Prevalence is counted, not estimated.

    “Five of seven” means five interviews hold a line that supports the theme.

What does it cost?

You pay for minutes of conversation.

Interview time is the only thing that is counted. Studies, teammates, typed test runs, the assistant and the write-ups come with it.

  • To start

    Free

    30 minutes, about 2 interviews

    Once, with the first workspace you set up. No card.

     Start free

  • Starter

    $99

    300 minutes, about 20 interviews

    A first study, start to finish.

    $0.33 a minuteAsk for it

  • Team

    $299

    1,000 minutes, about 66 interviews

    Several studies, or one with many people.

    $0.30 a minuteAsk for it

  • Your own terms

    Ask

    More minutes, or no limit

    For a pilot, an agency, or a team that interviews every week.

     Write to us

A minute is a minute of spoken interview, test calls included. Typed test runs, the assistant and the write-ups don't draw on it. Packs are paid by invoice for now: write to us and the time is added to your workspace once it is settled. Minutes are good for 12 months.

Anything else?

Before you try it.

Do participants know they are talking to an AI?
Yes. The page they open names the interviewer as an AI before the call starts, and a line you can edit tells them the conversation is recorded and transcribed and who will read it.
What if someone would rather not talk out loud?
They can type their answers instead, at any point in the call. Captions of what the interviewer says are on from the start.
Which languages does it interview in?
English, Spanish, German, French, Portuguese, Italian, Dutch, Polish, Ukrainian, Russian, Turkish, Japanese. You choose one per study; the guide, the interviewer and the participant's page all use it.
How long is an interview?
As long as you set, from five to forty-five minutes. The editor estimates how long your guide takes while you write it, and the interviewer paces itself to the time once the call is under way.
Can I try it before sending it to customers?
Yes. Every study has a test call: you talk to the interviewer yourself, on the draft, as often as you like. Test calls are kept apart from real interviews and never count towards a study's limit; a spoken one uses interview time like any other call, a typed one uses none.
Can an agent set a study up for me?
Yes. The built-in assistant drafts a study from a few sentences, makes the edits you describe and answers questions about the results; it asks before publishing or starting an analysis. Your own agent can do the same over MCP, with a key you can revoke.
What happens to what people say?
Interviews, transcripts and reports belong to your workspace, and only its members can open them. Nothing is sold, and we do not train models of our own on interviews. The privacy notice says what is collected, who handles it and how to have it removed.

Ask the question you have been putting off.

One sentence is enough to start. Your first guide is drafted from it as soon as you have signed in.

One sentence is enough.

Free to start: 30 minutes of interviews included. No card.