← Ashu research analysis

CleverX · AI research analysis

Months of research,
ready to review
in minutes.

Large studies could run for months, leaving researchers to read transcripts and scrub through recordings by hand. We designed AI analysis that generated the first pass: transcripts, summaries and highlight reels, in minutes.

led end to end live in production 2025
The session view: a player with AI chapters on the left, a synced transcript in the middle with one line highlighted, and the agent's summary on the right.

One session with its player, chapters, synced transcript and AI summary.

CleverX supports human-moderated, AI-moderated and unmoderated studies. A single project can contain dozens of recordings, each fifteen minutes to an hour long. Once fieldwork ended, researchers still had to work through that material one recording at a time.

Researchers used the feature to speed up synthesis. Founders and PMs could use the same project view to read a study themselves instead of waiting for a separate report.

10–100participants in a single project
15–60minutes of recording each
monthsof manual analysis on larger studies

01 · the starting point

Why analysis took so long

Researchers had to read transcripts, jump through recordings, tag useful moments and remember how one session connected to another. On larger studies, analysis could continue for months after the research began.

The volume grew faster with AI-moderated studies. Researchers could run far more interviews, but the old analysis process still asked them to review each one manually.

02 · the first pass

AI organised the study before anyone opened it

When a study finished, the platform generated transcripts, chapters, summaries, themes and suggested highlights. Researchers no longer had to begin with a blank set of recordings; they could review an organised first pass in minutes.

AI chat let people ask follow-up questions across the study, including whether anyone disagreed or what appeared in a specific part of the research. They could keep a useful answer as a finding or continue the thread.

The Ask AI view: a researcher's question about the study, an answer, and the clips it was drawn from.

Ask AI could answer follow-up questions across the project.

03 · review

From a finding to the recording

AI chapters gave each recording a table of contents, so a researcher could jump to the relevant section instead of scrubbing from the beginning.

Each AI summary or suggested highlight linked to the participant and timestamp it came from. A researcher could jump to that moment, check what was said, then keep, edit or discard the suggestion.

The transcript stayed in sync with playback. Selecting a line opened the actions researchers already needed: highlight, tag, clip, comment or ask the agent.

A transcript with one line selected, the action menu open on Apply tags, and the project's tag list open beside it.

Selecting a transcript line opens the actions for that moment, including the project’s own tag list.

Saved moments collected into one highlight reel for the project. AI suggestions were marked separately from clips chosen by a researcher.

The project highlights view: tabs for transcripts, highlights, tags and project summary, an Ask AI button, and a grid of highlight cards from different participants.

One highlight reel, drawn from every session in the project.

04 · the summary

Insights that can be reviewed one by one

Instead of generating one long summary, we broke it into individual insights. A researcher could keep, edit, tag or discard each one without touching the rest.

The project summary: discrete insights, each carrying the participant and timestamp it came from.

The project summary is split into individual insights that can be reviewed separately.

05 · outcome

What shipped

Live in production across every project run on the platform: transcripts, chapters, summaries, themes, highlights and chat.in use
The first analysis that could take months to assemble manually was generated in minutes.minutes, not months
Researchers remained the final reviewers. AI suggested the structure and useful moments; a person decided what to keep.human reviewed
Each project produced one highlight reel that could be shared with the wider team.the deliverable