Automatic episode QA checks for a podcast agency publishing to 99 shows
Watches raw recordings arrive, checks every published episode against what was approved, and alerts the team when something doesn't match.
We built Fame guardrails around every episode: a bot that spots raw recordings landing in Drive and tags the right editor, a watcher that checks each newly published episode on 99 shows against the approved version, and audio checks that confirm the right file. Main result so far: 99 shows' public feeds checked every 15 minutes.
Who we built it for
Fame is a B2B podcast agency that has been running for six years. Its team finds and books guests, produces episodes and promotes each show for around 100 clients, and Fame Crew finds local videographers for shoots and handles the editing. Team size: about 97 people (estimate), most of them freelancers and contractors. Fame is Forge's own sister company, so we built these tools for ourselves first and still run every one of them.
Who uses it: the ops team, editors.
Mistakes were found by listeners
Raw recordings landed in Drive and someone had to notice and tell the editor. After approval, the final upload to the podcast host happened by hand, and nothing checked that what went live was what the client approved.
The ops team spent around 20 hours a week watching folders and checking uploads (estimate).
Checks at each hand-off
Raw intake bot
Spots new raw recordings in Drive, tags the right editor on Trello and Slack.
Feed watcher
Every 15 minutes, reads each show's public feed and compares new episodes to the approved version.
Audio identity check
Confirms the audio file matches the episode it's meant to be, and retries until it can.
What the ops team and editors used before
Someone watching Drive folders and spot-checking published episodes by hand.
Monthly cost of what came before: £0 a month (estimate)
What changed for the ops team and editors
How we look after the episode qa guardrails
Forge built this for Fame and looks after it: errors are reported automatically to an error tracker, and anything that stops someone doing their job is fixed the same day.
These guardrails exist because of a real mistake, and we say so: we built the feed watcher after a wrong episode reached a show's podcast page.
What broke, and how we fixed it
- October 2026: We found the audio identity check had reached a verdict on only 7 of 1,314 audio checks, because it had 60 seconds to finish work that usually takes longer, and skipped checks were never retried. We moved it to a retry lane that keeps trying for up to six attempts, and added the feed watcher as a second check after publishing.
What it's built with
- Next.js
- Vercel
- Upstash Redis
- Google Drive API
- Trello API
- Slack API
- RSS
- AssemblyAI
Connects to: Google Drive, Trello, Slack, each show's public feed.
What it would cost a client today
Start with the 2-week ops audit ($2,500 / £2,000, credited to your first month). A tool like this then fits the growth plan at $3,500 / £2,800 a month, with this as the first build included. Guardrails watch several systems at once, so they usually fit growth.
Plans have a 3-month minimum. AI usage is passed on at cost, and the tool runs on your own accounts, so you own it. See pricing for the full breakdown.
Would this work for your firm?
If a mistake reaching a client is expensive for you, automatic checks at each hand-off give your team peace of mind and the checking time back. It suits content agencies and any firm publishing on clients' behalf.
See what we build for agencies, or read about our business automation work.
Questions about building a episode qa guardrails
Can you automatically check a published podcast episode is the right one?
Yes. Fame's feed watcher reads each show's public feed every 15 minutes and compares new episodes to the approved version, alerting the team on a mismatch.
Why check after publishing instead of before?
The upload to the podcast host happens outside Fame's tools, so the public feed is the earliest point a check can see. It catches a mistake within minutes rather than when a listener does.
Who gets told when a check fails?
The operations channel in Slack and the episode's Trello card.
Who owns the tool once it's built?
You do. It runs on your own accounts and you pay those tool costs directly; we monitor it, fix it and keep improving it on the monthly plan.
Want a episode qa guardrails shaped to your firm?
Start with a 2-week ops audit - $2,500 / £2,000, credited to your first month.