Intoflow is an AI platform where a single agent runs the whole UX research cycle on its own, from the brief and recruiting respondents to live interviews, transcription, analysis, and a ready report. I started it as a side project so my team could run user research fast, without spending weeks on manual work.
Good UX research takes time. You have to find the right people, schedule and run every interview by hand, then spend days reading transcripts and writing up what you found. Intoflow takes that whole cycle and hands it to one AI agent.
Instead of hiring researchers and stitching tools together, the user describes the study once. The agent prepares the interview guide, talks to respondents in real time, transcribes and translates the conversations, analyses them, and turns everything into a report with insights, quotes, and an answer for every hypothesis.
Make the UX research cycle faster and cheaper without losing quality. The product is built for product teams, researchers, and startups that need quick customer development but don't have weeks to spend on it.
This is my own side project. The whole concept is mine. I came up with the product, defined all of its logic, and mapped every user flow and journey, then designed and built it end to end.
The flow follows the same steps a researcher would, just handed to the agent, from the brief all the way to a finished report. Here's the whole cycle, the way a user moves through it.
Every study lives in one workspace, with respondents, interview counts, and processing status at a glance, and a button to spin up a new one.
A five-step brief covers context, goal, tasks, and audience. The agent then proposes the hypotheses to test, and you edit them until they fit.
From that brief, the agent writes the interview guide (the questions and themes), each one tied back to a hypothesis.
Respondents open a link, pick their language, and talk to the agent in real time. Every conversation is recorded automatically, in more than ten languages.
Each interview opens with its recording, a full transcript, an automatic translation, and the takeaways the agent pulled out.
When you're ready, the agent compiles everything into a single report (insights, respondent profiles, supporting quotes, and a verdict on every hypothesis).
The frontend is built on Next.js and Tailwind CSS, with React Query for server state, Jotai for local state, and Server-Sent Events for streaming AI chat. The AI interviews run through a voice agent, and LLM providers power the hypotheses, guides, analysis, and reports.