Turn a single webinar transcript into weeks of multi-channel content using Claude Cowork and Claude Code. Upload a raw transcript, connect your brand context docs, and generate LinkedIn posts, blog articles, newsletter recaps, email sequences, and short-form video scripts, all in your brand voice. Includes both a manual high-quality workflow and an automated tool you can build yourself in 2 to 3 hours.
Workflow Description
This workflow takes a raw, unedited webinar transcript and repurposes it into a full content package across multiple formats and channels. The system uses Claude's context doc folder to maintain brand voice consistency across every output.
There are two versions: a manual process where you work directly with Claude Cowork through multiple editing rounds (higher quality, used for client work), and an automated tool built in Claude Code that generates a Google Doc with all content types in one pass (faster, used as a starting point).
An hour-long webinar produces 8,000 to 12,000 words of original, conversational content. That is substantially more raw material than a blog outline or content brief, and it is inherently original because it came from a live discussion that nobody else can replicate.
Before You Begin
Tools you'll need open
- Claude Desktop App (Cowork or Code, not the browser version)
- Wispr or another voice transcription tool (optional but recommended)
- Riverside, Zoom, or any webinar platform that exports transcripts
- Google Docs (for the automated tool output)
What you'll need before starting
- A raw webinar transcript downloaded directly from your recording platform. Do not clean it up or edit it. Claude handles messy transcripts without issue.
- A context doc folder on your local machine containing: brand and writing guidelines, ICP documentation, previous content examples, presentation decks or sales materials with branding, product info and competitor positioning.
- A "don't do this" list for your AI writing. Ours bans em dashes, staccato sentences, passive voice where active works, and overused AI words like "shift," "gap," and "leap."
How It Works
The manual workflow follows four stages. Each stage feeds into the next and requires human judgment at every transition.
Transcript upload → Idea extraction → Draft generation → Multi-round editing → Final content package
The automated version compresses all of this into a single tool that accepts a transcript and outputs a Google Doc. The quality trade-off is real: the automated version gets you about 70% of the way there. The manual process, with a human editing each draft, gets you to publishable.
Build Instructions
Create a dedicated folder on your local machine for each client or brand. Add brand guidelines, ICP documentation, writing style examples, previous high-performing content, and any presentation decks with visual branding. Connect this folder to Claude Cowork or Code at the start of every content session.
Download the unedited transcript from Riverside, Zoom, or your webinar platform. Drop it directly into the Claude session. Do not clean it up. The messier the input, the more context Claude has to work with.
Ask Claude to outline the most engaging and insightful ideas from the transcript, taking into account the ICP and company info from the context docs. Use voice transcription (Wispr) instead of typing. When you speak, you naturally over-explain and give background context. That extra context produces dramatically better output than clean, trimmed typed prompts.
Ask Claude to write drafts for each content piece using the outline it produced. Specify: LinkedIn posts, blog articles, newsletter recap, email follow-up sequences, and short-form video scripts. Claude pulls from the context docs to match the brand voice automatically.
Read every draft. Identify every sentence that sounds like generic AI writing. Tell Claude exactly what is wrong: this opening line needs to work without context, this paragraph uses a banned sentence structure, this CTA points to the wrong page. For important posts, this process can take up to an hour per piece. That editing time is the entire value proposition.
Take one of your full manual content sessions and ask Claude to summarize your workflow. Then go into Claude Code and say "build me this." The tool took us 2 to 3 hours to build. We are not developers. The tool generates a Google Doc with all content types. Subscribe to our Substack to get the markdown file and build your own version.
Quality Checklist
Before publishing any content piece
- Does the opening line work without any additional context?
- Have you removed all banned sentence structures and overused AI words?
- Does the tone match the brand voice from your context docs?
- Is the CTA pointing to the correct page?
- Would you be comfortable if a client saw this as a first draft?
For the automated tool output
- Have you reviewed every piece before publishing? (Do not blindly publish AI output)
- Does the Google Doc contain all requested content types?
- Are the LinkedIn posts under the character limit?
- Do the email sequences have correct subject lines and personalization tokens?
For the context doc folder
- Is the writing style guide up to date with your latest "don't do this" list?
- Do you have at least 5 examples of high-performing content for the brand?
- Are the ICP docs specific enough (job titles, company sizes, pain points)?
Common Mistakes to Avoid
- Skipping the context docs. If you paste a transcript with zero context, you will get mediocre output. That is not the tool failing. That is you skipping the foundational work.
- Publishing first-draft AI output. If your agency delivers what Claude spits out on the first pass, your clients can do that themselves for $20/month. The human judgment layer is what they pay for.
- Typing prompts instead of speaking them. When you type, you self-edit and trim context. When you speak, you ramble, over-explain, and give background. All of that extra context makes the output dramatically better.
- Using vague instructions. "Optimize for engagement" and "make it good" mean nothing to an AI. Say exactly what you want: avoid staccato sentences, do not use the word "bank" because this is a finance startup, the CTA should drive to a demo page not a blog post.
- Following the same template every month. If you use the automated tool without adjusting for what made each webinar unique, every content batch will feel identical. Sit down and ask: what was special about this one?
- Sending follow-up content too late. If you send the post-webinar email a week after the event, it is over. People have already forgotten. This workflow exists to get content out within hours, not days.
Handling Special Situations
Highly regulated industries
If you work with finance, healthcare, or legal clients, your context docs need to include every compliance constraint. What words you cannot use, what claims require disclaimers, what needs legal review. Claude will follow these rules if they are clearly documented in the folder. We work with multiple finance startups where you literally cannot say the word "bank" in certain contexts.
New clients with no content history
Spend 2 to 3 hours upfront compiling brand docs, writing guidelines, and gathering examples of content that represents the voice you want. The first few batches will require heavier editing. By the third or fourth webinar, the context library is strong enough that first drafts are noticeably better.
Multiple ICPs attending the same webinar
Download the RSVP list before generating content. Ask Claude to segment attendees by role, industry, company size. If you identify 3 to 4 core ICPs, consider creating targeted email sequences and specific social posts for each segment. You can even turn these into paid ad creatives for each ICP.
When the webinar content is mediocre
Not every webinar produces great content. If the conversation was surface-level, you might only get 2 to 3 usable content pieces instead of 5 or more. That is fine. Do not force mediocre insights into more posts just to hit a content quota.
Claude vs ChatGPT for this workflow
We used ChatGPT exclusively through 2025 and switched to Claude in 2026. The biggest improvement: Claude maintains context from the docs folder throughout the entire chat. With ChatGPT, we would constantly re-remind it to reference the brand guidelines. By the end of a long content session, ChatGPT would drift from the brand voice. With Claude Cowork and Code pulling directly from local context docs, that drift does not happen.
Measuring Success
Content output per webinar
- 5+ LinkedIn posts (each with a unique angle from the transcript)
- 2 to 3 blog articles (long-form pieces based on specific topics covered)
- 1 newsletter recap
- 1 email follow-up sequence (3 to 5 emails)
- 2 to 3 short-form video script outlines
Time benchmarks
- Manual workflow (high quality): 3 to 4 hours per webinar for a full content package
- Automated tool (good baseline): 20 to 30 minutes to generate, plus 1 to 2 hours editing
- Before this workflow: 2 to 3 days for a content team to produce the same output
Quality indicators
- Content passes brand voice review without major rewrites
- Follow-up emails sent within 24 hours of the webinar (not a week later)
- LinkedIn posts generate engagement comparable to manually written posts
- Clients cannot tell which posts were AI-assisted vs. fully manual
The Prompts
Prompt 1: Idea extraction
Prompt 2: Draft generation
Prompt 3: Editing round
Expected Results
- A full multi-channel content package from a single webinar, produced in hours instead of days
- Content that maintains brand voice consistency because it pulls from your context doc library
- An automated tool you own and can customize, built in 2 to 3 hours with no coding knowledge
- A repeatable weekly workflow that scales across multiple clients
- Follow-up content published while the webinar is still fresh in attendees' minds
- Significantly reduced reliance on freelance writers for routine content repurposing
Frequently Asked Questions
How many content pieces can you realistically get from one webinar?
It depends on the depth of the conversation. A one-hour webinar where speakers go deep on specific topics and share concrete examples will produce significantly more than one that stays at a high level. From a strong webinar, we routinely pull 10 or more individual content assets. From a weaker one, sometimes only 2 to 3 pieces are worth publishing. Trying to force content from a surface-level conversation is how you end up with generic posts that nobody engages with.
Will the AI-generated content sound robotic or generic?
That depends entirely on you. If you hand it a transcript with no brand guidelines, no tone of voice documentation, and no examples of previous content, the output will sound like every other AI-generated LinkedIn post. If you spend the time building a context library with writing style docs, ICP information, and examples of content the brand has published before, the output is a completely different quality level. The AI is only as good as the context it has to work with.
Can I use this workflow if I don't have a transcript?
You need a transcript. Most recording platforms generate them automatically (Zoom, Riverside, Google Meet all have this built in). If yours does not, run the recording through a transcription service like Otter.ai or Rev, or upload the audio file directly to Claude and ask it to transcribe. The transcript does not need to be polished. Timestamps, speaker labels, filler words, none of that matters. Raw is fine.
Is it legal to repurpose webinar content into social media posts and articles?
If you hosted the webinar, you own the content. Repurpose it however you want. If you were a guest speaker, check the terms of your appearance, but most hosts are happy for guests to repurpose their own contributions since it extends the reach of the webinar. The AI is not generating new ideas from nothing. It is restructuring and reformatting content that already exists in your transcript.
Does this work for podcast episodes, YouTube videos, and conference talks?
Yes. Any long-form content that produces a transcript works with this exact workflow. We have used it for podcast episodes, YouTube interviews, internal all-hands recordings, customer discovery calls, and conference keynotes. The format of the original content does not matter. What matters is that you have spoken words captured in a transcript, and that those words contain ideas worth repurposing.
What if the webinar had multiple speakers from different companies?
You can repurpose content from your own speakers without issue. For guest speakers from other companies, let them know you are repurposing their quotes or insights, especially if you are attributing specific statements publicly. Most guests appreciate the additional exposure. Sharing drafts before publishing also makes them more likely to reshare the content.
Can I use ChatGPT instead of Claude for this?
You can use any large language model. The workflow is model-agnostic. We switched from ChatGPT to Claude because Claude maintains context from local files more reliably during long content sessions. If you already have a working setup with ChatGPT and it holds context well enough for your needs, the same prompts and process will work.
How do I handle webinars in languages other than English?
Claude and most major language models handle multilingual content well. You can feed in a transcript in one language and ask for output in another, or keep everything in the original language. If your audience is multilingual, you can generate the same content package in multiple languages from a single transcript. The context docs should be in the target output language for best results.
What happens if my webinar transcript is really long (2+ hours)?
Longer transcripts produce more raw material, which is a good thing. The idea extraction step becomes more important because you need to filter down to the most valuable insights. For transcripts over 15,000 words, consider breaking the extraction into segments (first hour, second hour) so the AI can focus on each section with full attention rather than processing the entire document at once.
Can a team use this workflow or is it just for solo marketers?
Teams can absolutely use this. In a team setting, one person typically handles transcript upload and idea extraction, then distributes content outlines to different team members for drafting and editing. The context doc folder can be shared via a team drive. The automated tool version is especially useful for teams because it produces a Google Doc that multiple people can edit simultaneously.
How often should I update my context docs?
Every time you publish content that gets a positive response, add those pieces to the context folder. Every time you get feedback that something was off-brand, update the writing guidelines with a specific note. The context library is a living document. Teams that update it after every content cycle get noticeably better output within a month. The ones who set it up once and never touch it see diminishing returns.
Want more like this?
One email per week. Recaps, prompts, and lessons from every live session.