Koji

Koji runs AI-moderated voice and text customer interviews that follow up on every answer and auto-synthesize themes, sentiment, and quotes into a decision-ready research report.

What is Koji?

Koji is an AI-native customer research platform from Amsterdam that collapses the old trade-off between a survey's reach and an interview's depth. You start from a research goal — or an existing survey link — and Koji's study-design agent auto-builds the brief, interview guide, and methodology (Mom Test, Jobs-to-be-Done, customer discovery, and more). AI moderators then run real conversations with participants by text, voice, or phone in 30+ languages, probing every answer for depth and running hundreds of interviews in parallel with zero scheduling. As responses land, the platform auto-synthesizes themes, patterns, and sentiment into a live research report where every finding traces back to a verbatim quote, and an insights dashboard lets you query all past studies in plain language. It's priced per credit (€1 each, chat = 1 credit, voice = 3), with a quality gate so only interviews scoring 3/5 or better count against your plan, and a free tier grants 10 one-time credits to pilot the workflow. Founded in 2025 by researchers, GDPR-compliant with EU data residency and a no-training-on-your-data policy, and backed by Google for Startups. The flip side: it's a three-person bootstrapped startup with roughly $15k ARR, no managed participant panel yet, and no independent reviews anywhere — the software is genuinely well-designed, but the vendor is an open bet.

Key Features

• AI-moderated interviews — AI runs natural voice, chat, or phone conversations that listen, adapt, and probe each answer for depth
• Adaptive probing — Follows up on every response (up to 3 levels) to reach the 'why' behind a surface answer
• Study design agent — Auto-builds the research brief, interview guide, and questions from one goal, with methodology selection
• Automated thematic analysis — Clusters themes, patterns, and sentiment across every conversation with no manual coding
• Traceable quotes — Every theme and recommendation links back to the verbatim quote and source transcript
• Quality gate — Each interview is scored 1–5 on relevance, depth, and coverage; only 3+ counts against your credits
• 30+ languages — Each participant is interviewed in their own language with native phrasing, no translation step
• Insights dashboard — Query every past study in plain language ('who mentioned pricing?') and get instant visualizations
• MCP integration — Create studies and generate reports from Claude, ChatGPT, or Cursor in plain language

Who is it for?

• Product managers who want a continuous discovery habit instead of one-off feedback sprints
• UX and design researchers who need to scale qualitative interviews without a moderation or agency budget
• Founders validating a problem before building, with Mom Test and JTBD frameworks built in
• Marketers capturing real customer language for positioning, campaigns, and case studies
• Agencies and consultants running multiple research projects across clients at a fixed cost

Who is it NOT for?

• Teams that need moderated usability testing with screen observation or visual stimuli — Koji is text and voice only
• Organizations whose InfoSec requires completed SOC 2 or ISO 27001 certifications today, not 'in progress'
• Researchers who want a managed panel or recruiter — you source every participant yourself
• Anyone uncomfortable with AI-generated analysis who won't dig into raw transcripts to verify claims
• Teams already happy with a survey tool and unwilling to change their research workflow