Abstract painting: one glowing line of pink voice waveforms across a dark field

Customer interviews that ask why, in their language.

The go-to-market OS after the CRM era: ask real people, follow up on every answer, and decide where to launch and who to sell to first.

Free plan. No card.

A real interview

Interviewer: How have the delivery delays affected your weekly meal planning?
Respondent: They’ve made it hard to plan around the kit, so I’ve had to keep backup groceries on hand and make last-minute trips to the store. It’s also made the subscription feel less convenient and less worth the rising cost.
A real interview CURI AI ran. Not a script.

Teams that ran research with us

  • LG Electronics
  • Kakao Bank
  • Kakao Mobility
  • Google for Startups
  • Metabranding
  • Biomoa
  • SYM Office
  • Toonsquare
  • Hyggee
  • Oing

The problem

Go-to-market decisions run on thin evidence.

  1. 01

    Surveys tell you what, not why.

    A checkbox says 32% won’t buy. It doesn’t say what would change their mind.

  2. 02

    Interviews don’t scale.

    Recruiting, scheduling and moderating 30 interviews takes weeks, and usually in one language.

  3. 03

    Bots answer surveys now.

    In a 2025 study in PNAS, an AI agent passed 99.8% of the checks surveys use to catch bots.

    Westwood, PNAS 2025

See it run

From one question to the reasons behind it.

  1. 01

    You write the question.

    Start from your own words. Run it as a survey, a voice interview or a focus group.

    How have the delivery delays affected your weekly meal planning?
  2. 02

    It asks why.

    After each answer, the interviewer decides whether to follow up, up to five times per question.

    How have the delivery delays affected your weekly meal planning?
    They’ve made it hard to plan around the kit, so I’ve had to keep backup groceries on hand and make last-minute trips to the store. It’s also made the subscription feel less convenient and less worth the rising cost.
    Follow-up
    What would need to change for the subscription to feel worth keeping?
  3. 03

    Every answer is judged.

    A separate model marks what each answer shows, such as a root cause or how likely they are to buy.

    They’ve made it hard to plan around the kit, so I’ve had to keep backup groceries on hand and make last-minute trips to the store. It’s also made the subscription feel less convenient and less worth the rising cost.
    INTENT_HIGH

    Reliable deliveries and fairer pricing would persuade the respondent to retain the subscription.

  4. 04

    You read the reasons.

    Answers are counted by reason, and each reason is backed by what people said.

    CURI AI results screen for the same 20 sample interviews. What decides it for you: I can see results 40%, The price for the size 30%, It is gentle on my skin 20%, I found another product 10%.
Abstract painting: a magenta field above a deep blue field, both made of thin voice waveforms

Voices, layered into insight.

How it works

One interviewer, three formats.

The same engine runs all three. It reads each answer and decides what to ask next, up to five follow-ups per question.

Survey
Your questions, plus open follow-ups where an answer needs one. Text, on any device.
Voice interview
One-on-one and spoken, in the respondent’s language. Every answer comes back as text you can search.
AI focus group
A moderated text room with several respondents. The AI keeps the topic and asks each person why.

Try it with your topic

AI drafts 3 opening interview questions in a few seconds. Treat them as a first draft.

Or try one of these:

Where conversations become data

CURI AI counts every answer by reason. Under each reason, you read what customers said in their own words.

  1. I can see results8 of 20 · 40%
    “My dark spots faded in about three weeks. That is why I keep buying it.”
  2. The price for the size6 of 20 · 30%
    “It is good, but $38 for 30 ml runs out too fast.”
  3. It is gentle on my skin4 of 20 · 20%
    “It does not sting, even on days my skin is red.”
  4. I found another product2 of 20 · 10%
    “My sister gave me a cheaper serum that works about the same.”

Based on a sample of 20 interviews about a serum. The answers are synthetic. They do not come from a customer.

CURI AI results screen for the same 20 sample interviews. What decides it for you: I can see results 40%, The price for the size 30%, It is gentle on my skin 20%, I found another product 10%.
The same 20 answers in the CURI AI results screen.

Why trust it

Answers you can defend.

  • Real people, recruited for your study

    Invite your own customers, or we recruit through panel partners and paid social, screened against your criteria. We don’t sell synthetic respondents.

  • Every answer is read

    A separate model reviews each answer and decides whether it needs a follow-up. Short or off-topic answers get one.

  • Your data stays yours

    You own your respondents’ data. We process it on your behalf and don’t use it to train models.

  • Security, stated plainly

    Our security program is built toward ISO/IEC 27001, ISO/IEC 27701 and Korea’s ISMS-P. We are not certified yet.

Use cases

What teams use it for.

  • Interviews in several countries at once

    Run the same voice interview in each market’s language and compare the answers side by side.

  • Concept and price checks before launch

    Test a concept or a price with real buyers, and hear why they would or would not pay.

  • Employee interviews

    Talk to every employee one-on-one and in private, at the scale of a survey.

  • Recruiting in a new market

    Screen people with a short survey, then invite the ones who qualify to the main study.

  • Voice intake

    A voice agent takes a spoken request on your site and fills in the form for your team.

  • Daily voice check-ins

    A short spoken conversation each day that tracks how someone is doing over time.

Cards show demo topics, not customer data.

Pricing

Start free. Upgrade when you need more.

Self-serve plans for teams that run their own studies. One unit is one text response; a voice interview is 20.

  • Free

    Trying it out on your own

    $0/mo

    100 units a month

    Start free

  • Starter

    Small teams running a few studies a month

    $149/mo

    1,500 units a month

    Talk to us

  • Growth

    Teams asking across markets and languages

    $499/mo

    6,000 units a month

    Talk to us

  • Scale

    Organizations running Curi at scale under their own brand

    $1,490/mo

    20,000 units a month

    Talk to us

USD card checkout opens soon. To start a paid plan in USD now, talk to us.

Enterprise

Managed studies, recruiting in your markets, invoicing and security review. From $20,000 a year.

Talk to us

Compare plans

FAQ

Questions

What is CURI AI?

A research platform that runs surveys, voice interviews and AI focus groups with real people. It asks a follow-up after each answer and counts the reasons behind what people say.

How is this different from a survey tool?

A survey asks only the questions you wrote. CURI AI asks those, then follows up on each answer, up to five times, so you get the reason and not just the choice.

Where do respondents come from?

Invite your own customers with a link, or ask us to recruit them through panel partners and paid social, screened against your criteria.

Which languages do you support?

Interviews run in the respondent’s language. Teams have run them in Korean, English and several European languages. Tell us your market and we will confirm.

Can I try it before paying?

Yes. The Free plan includes 100 units a month and every AI feature except AI focus groups. No card is needed.

Do you train AI models on our data?

No. You own your respondents’ data. We process it only to run your studies.

Hear the why before you decide.