Case study · CogniStream · Product engineering
Building CogniStream.
Two founders had a thesis: customer research should sound like a conversation and arrive in hours, not weeks. We built the product behind it, fully remote, in a couple of months.
One team, four cities
- Texas
- Dublin
- Oslo
- Kathmandu
Role
- All product engineering
Timeline
- A couple of months
Team
- Fully remote
01 — The thesis
Two founders, one conviction.
They had spotted something every insight team knows: talking to customers is where the truth is, and it is also slow, expensive and hard to scale. They wanted an interviewer that never tires, listens to how things are said, and hands back evidence instead of opinions.
The problem, in their words
“Qualitative research takes too long and costs too much.”
Their answer: let an AI hold the interview, by voice, with real customers, and have the report ready the same day.
They brought
- The thesis
- The market insight
- The final word on product
We brought
- Engineering expertise
- Voice AI and analysis
- The platform, end to end
02 — How we worked
Linear was the office.
We never shared a room. Two founders and our engineers, in different places, working from one board. Every idea, bug and decision became an issue with an owner, and the work moved in cycles everyone could see.
Issues, not inboxes
Nothing lived in someone’s DMs. If it mattered, it was an issue, with context, an owner and a state.
Cycles, not deadlines
Progress was something the founders could see, not something they had to ask about.
Remote, never distant
Async by default. A call when a decision needed a conversation, and the outcome written back into the issue.
Timeline
A couple of months.
From the first issue to the product.
03 — The hard part
A pause is not a full stop.
In a voice interview, silence is ambiguous. Someone might be thinking, or they might be done. Answer too early and the AI talks over the best part of an answer; too late and it feels like filling in a form. And people talk over the interviewer, so it has to know when to stop.
One exchange, as the interviewer hears it
Hold
Through a thinking pause, the interviewer waits. The considered answer is usually the one worth having.
Yield
When a respondent cuts in, it stops and listens, the way a good interviewer would.
04 — The workflow
From a brief to a report.
A study starts with a few lines about what you want to learn and ends with a report you can trust, because every insight keeps its evidence: the quote, who said it and the audio.
The study, end to end
- 01BriefWhat you want to learn
- 02ObjectivesDrafted from the brief
- 03Discussion guideQuestions, then probes
- 04Share a linkWhatsApp, text or email
- 05Voice interviewsMany at once, any browser
- 06AnalysisThemes, quotes, emotion
- 07ReportAudio behind every insight
Emotional insights
Emotion profile
Emotion over an interview
Setup in minutes
New study · Coupon expiration stress study
Why do customers let their coupons expire, and how does it feel when they notice?
The founder writes what they want to learn. The platform drafts the objectives and the discussion guide; they refine it, then share a link. No scheduling, no app to install.
05 — What we built
Everything behind the conversation.
No starter kit and no inherited code. Four systems, built from scratch.
Voice interviewer
01Live in the browser with no app to install. It holds the conversation, probes deeper and knows when to speak.
Emotion and tone
02Reads hesitation and confidence from the voice, not just the words, across languages.
Study setup and reports
03From a brief to a discussion guide; from interviews to themes, quotes, emotion and recommendations.
Platform
04Accounts, studies, sharing by link, and many interviews running at the same time.
06 — Today
What it does today.
The product, in CogniStream’s own figures.
to set up a study
Hours
not weeks, from question to report
less time and cost than traditional research
Hundreds
of interviews running at the same time
Many languages
interviews beyond English
As stated by CogniStream · cognistream.ai
For founders
You bring the thesis. We’ll build the product.
CogniStream’s founders came with a conviction; we brought the engineering. A couple of months later, the product was holding the conversations.