The Pros and Cons of Writing a Talk with AI
As speakers and speaker coaches, we spend a lot of time crafting talks—both for ourselves and with our clients. It’s an obvious exploration, therefore, for us to look into how we can leverage AI to craft these talks more quickly (and ideally, effectively). Recently, Stephanie started giving a talk entitled Stories That Sell. In the past several weeks, Stephanie has delivered this talk three times and here are the things she learned while leaning into AI as a tool to help her.
AI is great for structure.
Stephanie started by asking AI to generate a talk about stories that sell. This is a super generic prompt that leaves things wide open for interpretation. What came back to Stephanie were a few ideas for how to structure the talk, how to open and close powerfully, and what key points to make in between. This was helpful in giving Stephanie a place to start, but it was just that: a start..
Your point of view should focus and guide the AI.
In our Impactful Presentations and Influential Storytelling programs, we spend a LOT of time discussing message clarity. This is because you can only be an effective and influential communicator if you are crystal clear about what you’re trying to communicate. Well… that remains true if you want to leverage AI to help you draft your talk from a blank sheet of paper.
After crafting an outline, Stephanie asked AI to generate a script that would fit the outline. AI came up with a lot of verbiage, but the point of view was vague. Functionally speaking, it just regurgitated back to Stephanie a bunch of stuff other people on the internet have already said about storytelling and sales. AI didn’t offer anything uniquely different that would make Stephanie’s talk special or interesting.
Once Stephanie put into AI a prompt that was more specifically targeted, with a strong point of view, and clear opinions about what to focus on, the software was able to generate language that was much more compelling.
You still need to add your own personality.
One of the key qualities of AI (at least at this point in its development) is that it needs a data bank to pull from. Since Stephanie doesn’t have 100 talks already out there in the world to analyze, AI is absolutely useless at figuring out how to make a talk in Stephanie’s voice. It doesn’t know the personal stories that Stephanie can weave into the talk. It doesn’t know Stephanie’s delivery style, so it can’t write a script that will feel natural for her to deliver.
One of the most crucial components of being a compelling communicator is that you have to figure out how to say what you want to say in such a way that it could only come from you. You need to embed stories that are yours and yours alone. You need to deliver the talk in a way that is authentic to you. It must be informed by your experiences and perspective, and showcase your personality. What is compelling to an audience, beyond the information you share, is what is unique and human about you.
If you’re one of those people that already has a ton of content (articles, videos, prior talks, interviews, etc) out there for AI to analyze, then you can prompt it to use the data already available to put together something that is unique to you and your style. That’s why you can ask AI to write a short story in the style of Dr. Seuss or create a song in the style of Taylor Swift. There’s data already available that AI can analyze to figure out the specific look and feel that is their vibe. For Stephanie, this isn’t the case, and that won’t be the case for most of you. That means that after AI spits out a script for you, you still have to go back and make it your own using your blood, sweat and tears.
AI can iterate really well.
We’re huge fans of iteration. We always say that a story or a presentation isn’t done… it just evolves closer and closer to perfection the more you work it, deliver it, collect feedback, and iterate.
After Stephanie delivered her talk for the first time, she asked the audience to complete a survey so she could collect some feedback. Then she input the transcription of the talk recording along with the audience feedback and asked AI to come up with ten concrete recommendations to improve the talk for next time based on the feedback and what it knows about great talks. The output was amazing! AI generated 10 very concrete ways to level up the talk, even offering specific scripts to replace certain parts of the talk. It looked at the relative weight of each component of the talk based on length of time spent discussing it, and suggested a few tweaks to emphasize specific points more thoroughly and de-emphasize less important points. It even went so far as to suggest metaphors to use in lieu of technical jargon, and how to finish even more powerfully.
The learning Stephanie pulled from this is that AI is an amazing co-pilot when it comes to crafting a talk or presentation. Stephanie still had to know what she wanted to say and have her own (hard won) point of view, and she still needed to be able to turn the talk into something that was uniquely hers. But in the end, AI is a fantastic partner in crime. It helped her generate an initial outline and was REALLY good at leveling up a draft into something much more polished. It’s also motivated her to be more intentional about starting to build a library of data on herself that she might one day be able to feed to AI, so that AI can model her style, suggest personal stories, and more effectively support her in the future.
What’s your take on using AI to generate a talk? What have you tried that really worked or really didn’t?