Case study Wistia
Mochine
An agentic video editor that turns a company's existing video library into new content through conversation. Mochine (formerly Remix) launched to GA as part of Wistia’s platform in March 2026, and building it changed how I run discovery and measure success when an AI produces the output.
Most video editors are built for people who edit video for a living. Wistia's customers are mostly marketers at small and mid-sized companies. They often have years of webinars, customer interviews, and demos in their Wistia library, and no time or editing skills to turn that footage into something new. Mochine lets them describe the video they want and get a finished cut back, made from footage they already own.
- Started from the library. Customers already host years of video on Wistia, so Remix focuses on repurposing that footage. A standalone AI editor has to get users to upload content before it can do anything.
- Used AI to study the AI. I built an analyzer that reads every Remix conversation and reports daily and weekly on what users asked for, what worked, and where they got stuck. It let a small team learn from thousands of sessions without reading them one by one.
- Picked metrics that fit unpredictable output. Every Remix result is different, so we tracked sentiment, job completion, and save rate.
- Launched thin on purpose. At first, chat handled nearly every interaction, including ones a normal editor puts in a toolbar. We added visual controls once usage showed where chat slowed people down.
- Ran it as a player-coach. I led the Create PM team while also acting as the hands-on PM for Remix, which kept me in the transcripts and in daily contact with the engineers.
Mochine is live and usage has doubled since launch. Between 30% and 35% of the videos it generates get saved or downloaded to be published or shared outside Wistia. Using AI to analyze AI conversations at scale is still the most unusual discovery method I've used in 12 years of product work.
Deeper Dive
Where chat fits. Chat let us learn fast, because people told us exactly what they wanted. It also showed that some jobs don't work well as a conversation. Pulling social clips from an hour-long webinar means scanning a set of options and keeping the good ones, and prompting for each clip slows that down. Since launch I've pushed Mochine toward visual controls around the chat, like suggestions it generates on its own and presets you can preview before applying. The agent still does the editing, and users now have more than one way to direct it.
Motion Kits. Motion Kits are custom motion-graphics packages Mochine applies to a video, so a generated cut looks designed and on-brand. Plenty of AI tools can cut and caption footage. I think Motion Kits will be one of Mochine’s biggest differentiators.
From clipping engine to editor. Mochine launched as a tool for repurposing and clipping. We've since shipped quick edits that let users keep refining a video in the conversation. They can clean up audio, add captions, music, B-roll, or a voiceover, and resize for a different aspect ratio. There's a lot left to build, but this is when Mochine became an editor people can work in.
Two kinds of demand. People clearly want AI to clip, caption, and repurpose their video. Whether they want to edit by conversation is a separate question, and a strong demo makes the two easy to confuse. Keeping them separate led me to put reliability ahead of broad promotion. Trying a tool once takes curiosity. Coming back a third or fourth time depends on it working every time.