You gave a conference talk. The audience loved it. The feedback forms said "this should be a course." And now you're looking at your slide deck and recording wondering how to make that leap. The good news: you've already done the hardest part — developing the ideas and testing them with a live audience. The transformation from talk to course isn't about creating new content from nothing. It's about restructuring what you already have for a different learning context, using ChatGPT and Descript to accelerate the process.
Quick Answer: How to Use AI to Create a Course From a Conference Presentation
- Transcribe Talk: Execute this step in your course creation workflow.
- Restructure for Self-Paced: Execute this step in your course creation workflow.
- Expand Key Points: Execute this step in your course creation workflow.
- Add Practice Elements: Execute this step in your course creation workflow.
- Build Course Shell: Execute this step in your course creation workflow.
What you’ll walk away with:
- A self-paced course from your conference presentation
- Expanded explanations for concepts that went fast live
- Exercises and assessments a talk can’t include
Why conference talks aren't courses (and how to bridge the gap)
A conference talk and an online course are optimized for different things. Your talk was optimized for a one-time experience — building energy in a room, fitting a time slot, creating a memorable impression. An online course is optimized for lasting change — students need to understand, practice, and apply your ideas over time, without you in the room.
The mistake most experts make is treating the conversion as a recording problem: "I'll just record the talk again in better quality." That produces a lecture, not a course. The work is designing practice around your talk's ideas — giving students opportunities to apply each concept before moving to the next.
Step 1: Transcribe and extract principles
Upload your conference recording to Descript and let it transcribe. Then feed the full transcript to ChatGPT with a specific extraction prompt. You're not asking for a summary — you're asking ChatGPT to identify the 3-5 core principles or frameworks you presented and separate them from the connective tissue of a live presentation.
Most 45-minute talks contain 3-5 core ideas surrounded by stories, transitions, audience engagement moments, and contextual framing specific to the event. The principles are your course content. Everything else is either dropped or repurposed.
Step 2: Review the audience feedback
If you have feedback forms, post-talk emails, or social media responses from your presentation, feed those to ChatGPT too. Ask it to identify which aspects of your talk generated the most engagement and which questions people asked afterward. This data tells you where to go deeper in your course — the moments that resonated most in person are the ones worth expanding into full lessons.
Step 3: Strip live-only elements
Go through each extracted principle and identify what only works in a live context:
- "As we discussed in the previous session..." — references to other conference talks
- "Raise your hand if..." — live audience interaction that doesn't translate
- "Given our limited time today..." — time constraints that don't apply to a self-paced course
- "Here at [Conference Name]..." — event-specific context
ChatGPT can flag these automatically if you ask it to identify "live presentation elements that won't work in a self-paced online course." Replace each one with something that works asynchronously — a reflection question instead of a show of hands, a deeper explanation instead of a time-constrained overview.
Step 4: Apply Tell-Show-Do
This is the framework that turns a principle into a lesson. For each of your 3-5 core ideas:
- Tell: Explain the concept clearly. Your conference talk already did this — use the transcript as your starting point, then expand where you were constrained by time.
- Show: Demonstrate the concept with a specific example. Your talk probably included one example per point; for the course, add 2-3 more. Examples from different contexts help students see how the principle applies to their specific situation.
- Do: Give students a practice activity. This is the piece your conference talk didn't include — there's no time for practice in a 45-minute session. Design an exercise that lets students apply the concept to their own work. This is where the learning actually happens.
The "Do" component is what separates a course from a lecture series. Without practice, students understand your ideas intellectually but never implement them. I've seen this pattern countless times with course creators on Ruzuku — the courses with built-in practice activities have dramatically higher completion rates and better student outcomes.
Step 5: Build the course structure
With your principles extracted and Tell-Show-Do applied, you have a course outline:
- Lesson 1: Introduction and overview (adapted from your talk's opening)
- Lessons 2-6: One lesson per core principle, each with explanation, examples, and a practice activity
- Lesson 7-8: Integration and next steps (adapted from your talk's closing, expanded with implementation guidance)
Use ChatGPT to draft the lesson scripts from your transcript, following the Tell-Show-Do structure. Each script becomes either a video you record or a written lesson — your choice of format.
Prompts to try
Here's the transcript of my 45-minute conference talk on [topic]. Extract the 3-5 core principles or frameworks I presented. For each one, provide: (1) the principle stated clearly in one sentence, (2) the key example or story I used to illustrate it, (3) which elements are live-presentation-specific and need to be adapted for a self-paced course. Ignore transitions, audience engagement moments, and event-specific references.
This extraction prompt produces the raw material for your course outline. The distinction between core principles and presentation elements is the key first cut.
I'm turning this conference talk principle into a course lesson using the Tell-Show-Do framework: Principle: [extracted principle] Original example from talk: [the example you used] For the TELL section: expand this explanation for someone learning at their own pace — no time constraints. For the SHOW section: suggest 2 additional examples beyond my original one, from different professional contexts. For the DO section: design a 15-20 minute practice activity that helps a [your audience] apply this principle to their own work.
Run this prompt once per principle. The practice activity design is where ChatGPT adds the most value — it generates ideas for exercises that you can then refine based on your knowledge of what works with your audience.
Here's my conference talk transcript and the audience Q&A that followed. Extract the 5 best questions from the Q&A — the ones that are most likely to come up again from future students. For each question, identify which lesson in my course outline it relates to, and draft a concise answer I can include in that lesson.
The Q&A mining prompt turns audience interaction into course content. Real questions from real practitioners are more valuable than hypothetical FAQs.
The human layer
Your conference talk worked because of you — your energy, your stories, your ability to read the room and adjust. An online course works differently. The energy comes from clear structure and well-designed practice activities. The stories still matter, but they need to work for students watching alone at 11 PM, not in a ballroom at 2 PM.
AI can help you restructure the content, but the pedagogical decisions are yours. Which examples will resonate with self-paced learners? What practice activities will actually get done? How deep should each lesson go? These decisions require your teaching instinct, not an algorithm. Use ChatGPT for the mechanical work — restructuring, drafting, generating exercise ideas — and apply your judgment to every output.
What it gets wrong
- It treats your talk as a transcript, not a performance. The energy shifts, the pauses, the moments where you emphasized something by slowing down — ChatGPT can't detect any of that from text. It gives every sentence equal weight. You need to manually identify which moments were the climactic points of your talk and ensure they're featured prominently in the course.
- Practice activities can be generic. ChatGPT generates reasonable-sounding exercises, but they're often too broad ("reflect on how this applies to your work") or too narrow ("complete this specific worksheet"). Push for activities that are concrete enough to complete but flexible enough to apply to different student situations.
- It doesn't know what you cut for time. Every conference talk has material that didn't make it in because of the time limit. Your course is your chance to include it. ChatGPT only works with the transcript it's given — you need to add back the depth and nuance that the conference format forced you to leave out.
Related guides
- How to Turn a Workshop Into an Online Course Using AI — similar transformation from a different live format
- How to Outline Your Course Using ChatGPT — refine your extracted outline into a complete curriculum
- How to Record Course Videos Using Descript — record your new lesson videos from the restructured scripts
- How to Create Your First Online Course — the complete course creation guide
Your conference talk already proved the ideas work. The audience response already validated the demand. Now it's about designing the learning experience that helps students apply those ideas on their own. Start free on Ruzuku and build the course your conference audience asked for — with lessons, practice activities, and community discussion all in one place.