An AI podcast copilot that researches your guest, builds your episode plan, and suggests smart questions live while you record.
- Client
- Wisecast
- Role
- Product builder and lead developer
- Year
- 2026
- Status
- Launched to the public
The problem
Prepping an episode takes hours across a dozen tabs. LinkedIn, Google searches, a doc of questions. Then recording starts and none of it helps you in the moment.
Features
Plan Episode (the core)
- A short guided flow — guest, episode length, audience, topics to prioritise
- Research runs in the background while the host answers, so the plan is ready the moment they finish
- Outputs a full run of show: episode promise, segments with goals and questions, follow-ups, transitions
- Edited by chatting — "add two questions to section 1" changes only that part
- Every element has a stable ID, so the AI patches in place and the page never jumps
Guest research
- Paste a LinkedIn URL, get a cited profile back
- LinkedIn scrape plus web search, with every claim linked to its source
- A progress bar showing the real pipeline stages, not a fake spinner
- Asks which person you mean when two share a name
- A validator rereads the summary to catch facts belonging to the wrong person
Live copilot
- Transcribes the recording as it happens
- Surfaces follow-up questions at the moment they land
- Waits for a complete thought, and never fires more than once every 22 seconds
- Stays silent after something vulnerable — that moment belongs to the host
Guest review
- The host shares a link; the guest comments on any line, no login
- "Resolve with AI" applies the note to the exact question it was about
- Comments stay pinned to their text as the plan changes underneath
Summary and export
- A structured summary after recording, with timestamped clip candidates
- Edited by chatting, or ask the transcript questions directly
- Exports a clean PDF
Billing and launch
- Stripe billing with a card-required trial and a hard paywall
- Analytics, plus the landing and demo pages
How it was built
A weekly loop with real hosts testing on live episodes. Bugs arrived as WhatsApp screen recordings, fixes shipped within days. Many features exist because a host hit a wall and said so.
Outcome
From prototype to a paywalled product in private beta.