The short answer: A podcast agency's post-production checklist has two layers — per-episode asset work (transcript, show notes, chapters, social, clips, SEO copy) and per-client delivery work (review link, approval gate, publish coordination, archive). Most agencies have the first layer partially covered by tools. Almost none have the second layer systematized. That gap is where hours disappear and margin gets crushed. This checklist covers both, and marks which steps can be automated with current tooling versus which still require a human decision.
Why this checklist is different from a creator's checklist
A solo podcaster's post-production workflow is a personal routine. An agency's post-production workflow is a manufacturing process — it has to produce consistent output across 5, 10, or 20 shows simultaneously, hand off assets to clients who didn't build it, and survive the departure of any single team member.
That operational difference changes every step. The transcript isn't just for show notes — it's an asset the client may need for legal review, accessibility compliance, or repurposing. The review link isn't a nice-to-have — it is the approval gate that controls whether the episode ships on time or slips two days waiting on an email chain. The archive isn't cleanup — it's the audit trail for scope disputes six months from now.
Agencies that run more than five shows simultaneously describe the same bottleneck: "My bottleneck isn't production. It's everything that happens after the recording." This checklist is organized around solving that bottleneck.
Part 1: Per-episode asset checklist
This is the work that happens between "recording delivered" and "assets ready for client review." Every episode, every show.
Audio intake
- Receive raw audio from editor or recording platform (confirmed file format: WAV or high-bitrate MP3)
- Verify audio is the final edit — not a rough cut. Transcribing a rough cut wastes time and confuses the client.
- Confirm episode number, guest name, and working title in the intake log (prevents mix-ups across concurrent shows)
- Log receive timestamp — this starts your internal delivery clock
Where automation helps: RSS bulk import (Swell AI, EpisodeOps) and shared Dropbox/Drive intake folders reduce intake friction, but the human confirmation step before transcription is non-negotiable. One wrong file costs more time than the intake process saves.
Transcription
- Run speaker-diarized transcription — speaker labels are required, not optional, for show notes accuracy at agencies
- Verify speaker count matches the actual episode (transcription tools sometimes merge guests)
- Spot-check 3 random 90-second segments for accuracy — industry benchmark is >95% accuracy; flag and correct any proper nouns, brand names, or client-specific terminology
- Export clean transcript (plain text and timestamped versions)
- Store transcript in episode folder — this is the source of truth for all downstream assets
Where automation helps: AI transcription (Deepgram Nova-2 and similar) handles the heavy lifting reliably. Manual QA of proper nouns and client terminology still requires a human pass — ~10 minutes per episode, not 45. The transcript review is also your quality gate before all content generation; don't skip it.
Show notes
- Generate structured show notes: episode intro paragraph, guest bio (one paragraph), 3-5 key takeaways, timestamps for major topic shifts
- Match the client's established format — is it bullet points or full paragraphs? Does the client include sponsor callouts inline or at the bottom? Agency shows have house styles that must be consistent episode to episode.
- Verify the episode title is final before writing the intro — show notes built around a working title create rework when the client changes it (they always change it)
- Include resource links mentioned in the episode — pull from transcript; don't ask the client to supply them separately
- Add client's standard CTA block (newsletter signup, Patreon, review ask) at the bottom — this is usually the same every episode; template it, don't recreate it
Where automation helps: AI show notes generation is now reliable enough to produce a strong first draft that requires editorial review rather than rewriting. The per-client format consistency is where tools diverge. Tools that require prompt templates per client (Castmagic, Swell AI, Podsqueeze) create invisible maintenance overhead that scales linearly with your client count — that's real labor you're absorbing.
Chapter markers
- Generate timestamped chapter markers aligned to genuine topic transitions — not every paragraph
- Format for each distribution target: Spotify (WEBVTT), Apple Podcasts (chapter format in the MP3 ID3 tags or as a chapters.json), YouTube (description timestamps)
- Confirm chapter count is reasonable — fewer than 5 chapters on a 60-minute episode usually means the tool found topics, not chapters; more than 12 usually means it found paragraphs
- Export in the format the client's hosting platform requires
Where automation helps: Chapter detection has improved significantly. The manual step is the format conversion — different platforms want different formats, and your delivery SOP should document which format goes to which platform for each client.
Social content
- Generate X (Twitter) post — one punchy pull-quote or stat from the episode, with episode link
- Generate LinkedIn post — 150-250 words, more context than X, professional framing, episode link
- Generate Instagram caption — shorter, hashtag set per client's established tag list
- Verify all social posts are in the client's voice — not the agency's generic voice, not the AI's generic voice
- Check for anything in the episode that the client may not want amplified (off-hand comment, unconfirmed stat, sensitive topic) — this is the social QA gate that protects the client relationship
- If the client approves drafts before posting: route to review in this pass, not after full approval
Where automation helps: All major tools (EpisodeOps, Castmagic, Swell AI, Podsqueeze) generate social content. Voice consistency is the persistent problem — tools that require you to maintain client-specific prompt templates shift the labor to you. The social QA pass is a human step regardless of tooling.
Newsletter copy
- Write episode-specific newsletter section — 100-200 words, what listeners will get from this episode, episode link, show notes link
- If the client runs a standalone newsletter (not just episode announcements): generate a longer newsletter edition draft from the transcript — 400-600 words, email-native format, no podcast-listener assumptions
- Match the newsletter's established tone — B2B clients write very differently for email than they do for show notes
- Confirm embed link format the client's email platform supports (Mailchimp anchor links behave differently from Klaviyo, etc.)
Where automation helps: Newsletter generation from transcript is available in EpisodeOps and Castmagic. Capsho lists it as "coming soon" as of mid-2026. The format/tone calibration for each client's audience is a human editorial step.
Blog draft / SEO landing page
- Generate a blog post from the episode — not a transcript dump, but a structured article with H2 sections, a clear thesis, and a search-intent frame
- Optimize for the episode's primary topic keyword — what would someone search to find the conversation this episode covers?
- Include guest name in the title and H1 if the guest has search profile (guest name searches drive meaningful traffic for B2B shows)
- Add internal links to the client's other content where relevant
- Format for the client's CMS (WordPress shortcodes, Webflow rich text, Substack prose)
- If the agency manages the client's SEO: add meta title, meta description, and featured image spec before delivery
Where automation helps: Blog post generation from transcript is one of the more reliable AI outputs because the source material (the transcript) is rich. The SEO optimization layer — keyword research, internal link selection, meta fields — is a human step that requires knowing the client's content strategy.
Vertical video clips
- Identify 2-5 clip candidates from the transcript — moments that are punchy, complete, and under 90 seconds
- Confirm clip start/end with the actual audio (AI clip selection is good but not infallible — a clip that reads well sometimes has an awkward audio lead-in)
- Add captions — required for LinkedIn, Instagram Reels, TikTok. Caption accuracy matters: the captioning on a B2B client's clip will be seen by their buyers.
- Export in the aspect ratio the client's team actually posts from (9:16 for Reels/TikTok, 1:1 for some LinkedIn accounts)
- Label clips by topic for the client — "Guest name on [topic]" so they know which clip goes where
Where automation helps: Clip selection and captioning are now solid AI functions. The clip label naming and the confirmation QA pass are human steps.
Part 2: Per-client delivery checklist
This layer is where most agencies' workflows are informal — handled through Slack, Notion, email, or whatever the client prefers. That informality creates the margin-killing delays. These steps should be as systematized as the asset production steps.
Pre-delivery quality gate
- All assets from Part 1 are complete — do not deliver a partial pack
- All assets use the client's name, show name, and episode title correctly — search the full pack for the most common mis-spellings of the client's brand before sending
- Social posts have been QA'd for voice, tone, and sensitive content
- Transcript is the clean, reviewed version — not the raw AI output
- If this is a premium client or enterprise show: internal second-pair-of-eyes review before delivery
Client review delivery
- Send the full pack through a single shareable review link — not a Dropbox folder, not a Notion page, not a Zip file attachment. A single URL the client can bookmark and return to.
- The review link should show: episode title, transcript, show notes, chapters, each social post, newsletter copy, blog draft, and clips in one view
- Include a clear approval call-to-action in the review — "Reply here" or "Click approve" — not just "Let me know what you think"
- Set an explicit response deadline: "Please send any revisions by [date] so we can meet the publish window." Clients who don't know there's a deadline treat it as open-ended.
- Log delivery timestamp — your delivery SLA clock starts here
The approval bottleneck is the single biggest margin killer in podcast agency work. Industry data points to 3+ hours of rework per episode when the review cycle isn't structured. A 4-day approval lag across 10 shows means ~160 hours of delayed pipeline per month — that's real dollars sitting idle. The mechanism that fixes this is a purpose-built per-episode review link with a clear approval action, not a better email subject line.
This is the capability most post-production tools don't provide natively. Podsqueeze has a content-share page. EpisodeOps ships per-episode client review links with comments as a first-class agency feature on its Studio plan ($199/mo, $159/mo billed annually) — the client gets a clean URL, sees the full Release Pack, and approves without logging into anything. Studio also includes white-label review pages, 5 team seats, and multi-show brand-voice adapters — the tier is purpose-built for agencies running multiple client shows.
Revision handling
- Log all revision requests against the specific asset (show notes, not "the episode") — this prevents scope drift
- Determine if the revision is within scope or a change order — client asking to reword a sentence is in scope; client asking to rebuild the blog post in a new structure is a scope conversation
- Set a revision turnaround expectation before starting — "I'll have revised show notes back to you by [time]" prevents the client from checking in every 30 minutes
- After revisions are delivered: confirm client has approved the specific assets that changed, not just replied "looks good"
Publish coordination
- Confirm the publish date — not the client's "ideal" date, the actual scheduled publish date the hosting platform is set to
- Deliver all platform-specific formatted assets by at least 24 hours before publish: episode page copy to the hosting platform, social posts in the scheduler, newsletter in the ESP, blog post to the CMS
- Confirm the hosting platform has the correct episode artwork, title, description, and chapters uploaded (this is the step that most often slips through)
- Send a pre-publish confirmation to the client: what's going live, when, where — one message, all platforms
Post-publish archive
- Save the full delivered pack (all assets, final versions) to the episode folder in your project management system — not just the review link
- Log the publish URL, podcast episode page URL, and each social post URL in the episode record
- If the client tracks downloads or social performance: schedule a 7-day performance pull
- Close the episode record — mark it delivered and published with timestamps
Where agencies lose the most time (and what to do about it)
The per-tool context-switch tax
An agency running 8 shows across Otter for transcription, Castmagic for show notes, Opus Clip for clips, Notion for delivery, and email for approvals is managing 5 tool contexts per episode per show. At 8 shows × 4 episodes/month, that's 160 per-episode tool transitions per month — each one a small decision about where something lives, which window to open, which login to use. This isn't a theoretical efficiency loss; it's the reason your team finishes by Tuesday and the episode doesn't go live until Friday.
The architecture that solves this is one upload → full pack → single delivery URL. See what that looks like at /p/demo-episode — no signup required.
The per-minute credit ceiling
Castmagic's Starter tier gives 20 hours of transcription per month at $79/month. Swell AI's Agency plan gives 600 minutes (10 hours) at $49/month. For an agency running 8 shows × 4 episodes × 45-minute average episode length, that's ~1,440 minutes of audio per month. Swell's Agency plan runs out in the first two weeks. You either pay enterprise pricing, ration uploads (which defeats the automation value), or you're doing mental math about which episodes "count" this month.
The math only works in your favor on a flat unlimited plan. See EpisodeOps pricing — Pro ($49/mo, $39/mo billed annually) is unlimited episodes, no per-minute tracking. For agencies who also need client review links, team seats, and white-label delivery, Studio ($199/mo, $159/mo billed annually) is where those features live.
The prompt-maintenance overhead
Every tool that generates client-specific content from a configurable prompt requires you to maintain one prompt template per client. For an agency at 10 client shows, that's 10 prompt configurations — and when a client changes their format preferences, you update the template, re-test the output, and verify it didn't break the other 9. That maintenance overhead is real and invisible in the per-tool billing. Tools that generate consistent, client-appropriate output from the episode content itself (not from a generic template you maintain) scale without this overhead.
The approval lag
Clients who receive assets in a Dropbox link respond slowly. Clients who receive a clean review URL with a clear "approve here" action respond faster — measurably. The mechanism isn't persuasion; it's friction reduction. A purpose-built per-episode review link with all assets visible in one view and an approval action that doesn't require a login is the operational equivalent of shortening the distance between "client receives assets" and "client clicks approve." This capability, along with client comments and white-label review pages, is available on EpisodeOps Studio ($199/mo).
For a deeper look at how agencies handle the full release operation at scale, see how podcast networks scale operations and the agency use case breakdown.
The per-client onboarding checklist (do this once, not per episode)
This section is a separate checklist from the per-episode workflow — these are the standing configurations that make per-episode work consistent.
- Capture show format: solo, interview, co-host, narrative — this determines the transcript structure and social framing
- Document the show's audience — not demographics, but: what does the listener do with this show? (Background listening vs. note-taking vs. learning a skill affects the show notes format)
- Capture the show notes format: bullet points or prose? One paragraph of intro or three? Guest bio placement?
- Capture the social voice: professional, conversational, technical, casual — one clear adjective per platform for each client
- Get the full client brand asset set: logo, brand colors, show artwork dimensions — you'll need these for clip captions and social card templates
- Confirm the hosting platform, CMS, and social scheduler the client uses — and which of those your agency manages versus which the client manages
- Document the review SLA: "We deliver by [day] and need approval by [day+2]" — get this in writing in the contract, not just in onboarding
- Establish the revision policy: how many revision rounds are in scope? What counts as a revision vs. a new deliverable?
- Confirm the client's approval contact — the person who actually clicks approve, not their assistant, not their co-host
Use this checklist
Print it, adapt it to your stack, put it in your project management template. The goal isn't to follow every line literally — it's to have every line answered before an episode ships. When something falls through the cracks, it's because no one was responsible for a specific step at a specific time. This checklist makes that visible.
If you want to see what the asset side of this checklist looks like when it's fully automated — transcript through clips through client review link, from a single upload — try it free at /p/demo-episode. One episode, no card, full Release Pack.
And if you're running more than five shows and evaluating tools, the EpisodeOps agency overview and the ROI calculator will give you the numbers to make the case internally.