"The thing that AI is really good at right now is giving us adequate stuff," says Ines Lee, Head of Content at Ali Abdaal. "It's like a cover band that really hits all the notes perfectly, but you're never gonna feel that element of personality, that element of jazz."
In this episode of The Animalz Podcast, Ines shares how she went from teaching economics at Oxford to running the content machine behind one of YouTube's biggest educational creators. As Ali Abdaal's channel grew to over 6 million subscribers, Ines developed workflows that balance AI efficiency with human creativity.
Ines explains how her team uses AI across their content pipeline — from video editing to newsletter creation — while preserving the distinctive personality of Ali and their content.
Ines Lee is the Head of Content for Ali Abdaal, a top YouTuber with over 6 million subscribers. Her unconventional career path took her from academia to content creation. She started with a PhD in economics at Oxford and a postdoc at Cambridge before joining Ali's team in 2021.
Ines initially balanced academia with content writing as a side gig. In late 2023, she fully committed to creative work when she stepped into her current role. She now leads a team responsible for Ali Abdaal’s YouTube videos, newsletters, and social media content that generate revenue through AdSense, partnerships, and sponsorships.
Ines recently launched her own Substack Side Road 38, where she explores ideas beyond her work with Ali's content empire.
Ines' economics training and content expertise deliver practical and thoughtful insights, offering a unique perspective on how content teams can integrate AI.
Instead of using AI to generate content directly, Ines describes a reverse-engineering approach: feeding AI examples of Ali's content to identify patterns and create style guidelines.
Ines predicts that human elements — personal experiences, distinctive voices, and authentic perspectives — will become more important differentiators.
Ines observes a trend of executive-led content in B2B marketing, citing examples like Adam Robinson and Stuart Machin (Marks & Spencer CEO). This suggests companies can learn from creator-led business models by putting authentic personalities at the forefront of their content strategies.
Ines explains how creators excel at treating content as a product itself — something B2B marketers can learn from. She highlights doola as a company applying this creator-inspired approach through high-quality educational resources.
Ines captures the essence of AI-generated content's strengths and limitations with this metaphor. While it can produce technically proficient content that follows the rules, it lacks the distinctive personality that makes content engaging.
Ines highlights a challenge many content teams face: AI promises automation but often misses the critical human elements that make content compelling. With short-form video, AI tools typically analyze transcripts without considering facial expressions, tone, or excitement — all essential factors in selecting engaging social clips.
Ines emphasizes that AI can't replace the critical thinking behind great content. She advises content creators to establish clear boundaries between human thinking and AI-supported production.
Hello... is there anybody out there creating real value with AI?
The AI conversation in content marketing has become deafening — skeptics shouting from one side, shallow tips from enthusiasts on the other. But somewhere in this noise, there must be pioneers who've actually figured something out, right?
We've gone on a search for the real pioneers — the ones who've ventured beyond the hype to succeed (or fail) spectacularly. Through their hard-won insights, we'll discover if there's actually something of value hiding in the noise, or if we're all just shouting into the void.
Check out other episodes in the season here
The Second Mountain (00:03:00): A book by David Brooks on meaning and purpose that Ines mentioned she's reread.
Emilia Perez (00:03:00): A film Ines recently watched and enjoyed.
The Substance (00:03:00): A film with Demi Moore that Ines and Ty found intense.
Ali Abdaal YouTube Channel (00:07:00): Ali's main content platform with over 6 million subscribers.
Part-Time YouTuber Academy (00:07:00): An info product by Ali that teaches creators how to grow their YouTube channels.
Productivity Lab (00:08:00): A 2023 membership program to help people double their productivity.
Fire Cut (00:11:00): An AI video editing tool that has cut Ali's team's editing time in half.
Grain (00:13:00): For call recordings/transcripts that feed into AI-driven writing workflows.
Adam Robinson (00:16:00): B2B founder (Retention.com) with strong LinkedIn presence.
Stuart Machin (00:17:00): Marks & Spencer CEO cited for personal brand-led content.
Dana Strong (00:17:00): Sky Group executive using personal content and social to boost brand presence.
doola (00:19:00): A startup whose founder uses educational content for growth.
Founder-Led Content: Blitz-Scaling Social with doola's Arjun Mahadevan (00:19:00): The Animalz interview with doola's founder mentioned by Tim.
VoicePal (00:22:00): Ali's new AI app that converts voice memos into content drafts by asking follow-up questions and formatting responses.
HeyGen and ElevenLabs (00:25:00): AI tools for creating avatars and voiceovers that Ines mentions when discussing current AI video generation.
Stay Strong: Never Let AI Fill Your Blank Page (00:30:00): Tim Metz's article on the Animalz blog that Ines cites as important to her thinking on "human vs. AI tasks" in content creation.
ChatGPT and Claude (00:32:00): Language models Ines and team use for brainstorming and drafting.
Spotter Studio and 1of10 (00:33:00): AI tools that optimize YouTube content by analyzing historical performance data.
Elicit (00:34:00): An AI research tool Ines' team uses to find and analyze scientific papers for evidence-based content.
Follow Ines Lee on LinkedIn and subscribe to her Substack, Side Road 38.
Ines Lee: [00:00:00] The thing that AI is really good at right now is giving us adequate stuff. By which I mean, you know, you look at it and the the structure is correct. The grammar's perfect, the flow is good. It's like a cover band that really hits all the notes perfectly, but you're never gonna feel that element of like personality, that element, or jazz.
Ty Magnin: Welcome to The Animals Podcast. I'm Ty Magnan, the CEO in Animals, and I'm Tim Metz, the Director of Marketing and Innovation at Animals. This season on The Animals Podcast, we're focused entirely on AI content use cases. You are coming on a search with us to meet the AI pioneers, those venturing beyond the hype to succeed and sometimes fail spectacularly.
Today's guest is truly fascinating. She went from teaching economics at Oxford to running a content machine behind one of YouTube's biggest educational creators. Ines Lee is the head of content for Ali Abdal, whose channel has exploded over the years to over 6 million YouTube [00:01:00] subscribers. In today's conversation, we cover what B2B businesses can learn from B2C creators, AI usage and video, AI usage and content creation, AI usage and ideation.
And Nina talks about exactly how they're using it to build ALI videos. We also cover some AI tools. That Ali and the team is actually building and putting out to market, so making a pivot into software, and we talk about how their AI usage has evolved. I think you're gonna enjoy this conversation with Inez Lee did you know Animals Now offers a podcast service?
We're taking over your audience's earbuds, reaching them during their commutes, their workouts, or when they're doing chores around the house from show strategy to editing and distribution. Animals can handle your podcast for you. With that same originality and audience first approach that we bring to all of our content, every podcast episode can become fuel for your broader content program.
You can mine your podcast for ideas, for articles, social posts, and other kinds of content assets, [00:02:00] helping you create more high quality work. In less time. Ready to start a podcast worth listening to. Head over to animals.co. Book a call with us and we'll start talking about your podcasting
Tim Metz: goals. Enis, thanks so much for joining us on The Animals Podcast.
Before we get started, we always like to ask, uh, what content have you been consuming lately? Just as a kind of warmup. Anything that stood out for you, maybe professionally that you've, that you've read or watched?
Ines Lee: Ooh. Um, well firstly, thank you so much for having me on. Uh, in terms of things that I've been consuming lately, I've been rereading, um, the Second Mountain by, by David Brooks.
Um, which is, yeah, a book broadly about like meaning and purpose and how to find that through, through commitment in terms of watching. I think because it's award season, there's a lot of, a lot of eye catching films, um, on on Demand. So I've been, I've been going through a couple of those.
Ty Magnin: What have you liked lately?
Nice.
Ines Lee: So I recently watched Amelia Perez. I didn't know anything about it and switched on. I was like, wow, I haven't seen something like this, um, for a while. And [00:03:00] um, I also saw, I dunno if you guys seen the one with Demean Moore called The Substance and it was, yeah. I couldn't get through it. I like, I watched 30 minutes and I was like, okay, it's too much.
Ty Magnin: Yes, it's a
Tim Metz: lot. You know, I think we'd love to hear your introduction from you, and especially just like a short version, I guess, of your journey, which is super interesting of how you went from a PhD in economics at Oxford to becoming the head of content for Ali Abdel, one of the most famous YouTube creators, maybe with 6 million plus subscribers.
So. Yeah, tell us about it. How did you, how did you end up working with Ali?
Ines Lee: Yeah, no, um, it's, it's, it's been a, a somewhat of a zigzaggy career. So yeah, I came, came from an academic background and after finishing the PhD I did a couple of years, um, doing a postdoc in, in Cambridge. Um, and during that time. It was like covid it, I, it kind of spanned pre covid to, to post Covid and I was always interested in doing something a little bit more creative.
Um, and started, you know, doing a couple of freelance writing gigs. And at around that time, uh, one of [00:04:00] my sisters sent me an Instagram story of Ali, um, saying that he's hiring a couple of writers. You know, during that Covid period, I think that's when Ali's content, uh, really started to pop up on YouTube.
So I'd been following him for a while. And one of the trial tasks for that job was, um, writing a script in the style of one of Ali's videos. Um, so I thought, you know what? I've got a bit of spare time. Let me give it a go. Um, and submitted it. Yeah. I kind of started working with Ali October, 2021, and at the time I was still kind of straddling two worlds of, you know, teaching economics as my day job and doing economics research and then kind of doing this, uh, more creative writing thing on the side with Ali.
Um, so yeah, worked with Ali kind of. Part-time for two years-ish from 2021 till the end of 2023 when I decided, okay, I wanna go full into this. Um, and that's when I, uh, exited academia and then, um, jumped in, um, to, to the creative role completely. Um, and that's when I kind of stepped up into the head of content role [00:05:00] for Ali.
Ty Magnin: Interesting. That's awesome. Can you tell us how does like, or does economics and content creation play into each other in any interesting ways?
Ines Lee: Ooh. Yeah, I've been thinking about this because I think, you know, initially when I started sharing, okay, this is, uh, this is the next step, people were like, oh, that seems like a big U-turn.
I think there are, there are certain things. I think from a thinking perspective, a lot of Ali's content is quite, um, evidence-based and, um, I. In some ways, like can, can be rigorous. So I think that kind of ability to look at scientific information and then distilling it in a way that's accessible to, you know, a broader audience, that's one way in which it overlaps in terms of the tools and techniques.
And then I think like from an economic lens, I think a broad understanding of like how attention works and, you know, demand and supply for certain things, especially in this world of like, you know, way too much information and way too much content and. A limited time to consume that content. You know, that kind of broad framing of these basic concepts is also helpful.
Ty Magnin: Yeah, [00:06:00] yeah. Sort of like on a 1 0 1 level.
Ines Lee: Yeah.
Tim Metz: Nice. I mean, we, because we're talking this season and also this episode about, about ai, but before we really get into ai, I think it's probably helpful for people to, uh, 'cause I also didn't really understand that, like, what does it mean to run content for. A YouTube creator, like what does that look like in terms of what do you do?
What, what does the team look like and maybe what, what, what does the team behind him look like? Because even people know Ali, they just see him. So yeah. Can you speak to that?
Ines Lee: Yeah, for sure. Yeah. I think similarly to you, you know, when you look at a YouTube creator, I was very surprised initially to think, oh, there's like a team behind this one man person on a YouTube channel.
Um, so I'll speak to what the team looks like for us. Um, so, and I think there are kind of generalizations of how, you know, creator businesses are run. Um, so broadly, um, in these creator businesses, there's like one main talent in our case, um, Ali, um, for our team, um, we're broadly split, split up into a content team and a commercial team.
I. The content team, um, are, uh, [00:07:00] focused on producing, you know, the best, you know, free content that we can. So that would be like the 99% of stuff that we put out. And that includes things like YouTube videos, newsletters, social media, content. You know, we used to also have a podcast. Um, so anything that falls under content would be under.
My, my kind of domain. And then the commercial team looks more after the product development side as well as, I guess what you would traditionally, uh, call kind of growth marketing. And that would include things like ads and paid spend and so on. Um, so in our case for products, the, the main thing that we've been focusing on are like info products, courses, membership subscriptions and things like that.
One of our flagship things is course called Part-Time YouTuber Academy. Where Ali kind of teaches people how to grow and scale their YouTube channels. So that's one big piece of our info products. The other thing is called Productivity lab, which we've just come up with last year, and it's a membership program to help people kind of double their productivity.
So I'm less in touch with that commercial side, but [00:08:00] yet, so my main job as head of content is to look after the, the free content that comes out. And the main revenue streams for that are, you know, AdSense partnerships and sponsorships. Um, whereas for the commercial side, it's. Uh, through, through the products.
Ty Magnin: What's the structure of your team?
Ines Lee: The structure, so yeah. Within, so it's broadly split into those, right? There's, there's Ali as the head, uh, Angus, our GM is kind of the, the integrator of, of the whole team. And he also sits as the head of the commercial team. Um, and then within the content team, I'll speak more about that, so that, that's me.
Uh, leading up a team with a YouTube producer, two people within social media. And then, uh, a few editors, video editors.
Ty Magnin: Nice. And do concepts come from in a team like this? Like, who's got, where does ideation come from?
Ines Lee: You know, it's been, it's been a real mix, you know, so let, let's talk about like, uh, YouTube videos, because that's kind of like what one of the biggest distribution platforms for us and biggest focus for us, that's really a mix of like, uh, the YouTube [00:09:00] producer coming up with a title, the Hook, which is like the core essence of a YouTube idea A lot of the times as well, Ali might ping an idea and that might be.
You know, from an interesting book that he's read, but half of the game with YouTube is, you know, getting people to click and then getting people to keep watching. Sure. And the part of getting people to click would mean. Taking that kind of half-baked idea of, oh, I read this in a book that's really interesting.
And then turning that into, you know, amazing titles and thumbnail concepts that we really feel confident people would click on. It's, it's a bit of a team effort. I would say. A lot of people pitch in with kind of ideas and concepts, but to really bring that to life and to really feel confident that this is something that will get out there.
The work really comes from drilling down into titles and thumbnails for the case of the YouTube channel.
Tim Metz: Let's talk about ai. Where, when did AI first come into the picture? Either for you personally or or in your work at Ali, or maybe that's the same thing.
Ines Lee: Yeah, I think it definitely did overlap when it came [00:10:00] into me and uh, when the business became aware of it.
I would say for me it was kind of summer 2023, so I think this must have been. 3.5 or 4.0 chat, GPT. So I was still doing, uh, work with Ali part-time and kind of half my foot was in the academic space and half my foot was in the creator space. Yeah. But I would say it's around that time that it came, it when I seriously started using it.
And I think the first time that I used it was I was writing, uh, a newsletter and was a little bit stuck on the subject lines. And my first experience was getting it to help me brainstorm some subject lines. And I kind of thought, oh wow, this is. You know, it kind of, you know, spat out 20, you know, pretty decent ones with minimal input.
And I was like, oh, there's something here.
Tim Metz: Nice. And how did it, how has it gone from there? Like, what does it look like now?
Ines Lee: So we do, we use it quite a lot and the use across the team, I would say is not, is uh, pretty, pretty individualistic. So, you know, people pick and choose to use it. To the extent that it supports the workflows.
So I would [00:11:00] say we kind of use it probably like th three buckets, and maybe this is a crew generalization, but these three come top of mind on the video editing side. We have certain AI softwares that help us really streamline creating the a cut of a video. So kind of pre, pre AI softwares, creating the, a cut of something would take, you know, a considerable amount of time.
And now with, with a tool called Fire Cut, which is like an AI powered editing tool that kind of cuts down on our editing time by one to two hours for the A cut, it would be at least like, you know, three, four hours. So it's, it's a significant like halfing of the time. Um, in order to get that. Because Ali tends to record, and this might vary from creator to creator, but at least for Ali, he like speaks to the camera and shields a little bit.
And part of the job of the editor is to cut that down into like a 20, you know, a really condensed and concise 20 minutes. So yeah, so that's kind of one part of it when it comes to video editing. And there are also like a bunch of tools that help us put, you know, captions on things. Whereas, [00:12:00] you know, pre ai that would be a consider, you know, also take a lot of time.
To put the captions on things, typing them
Ty Magnin: up, listening to them. Sure. Yeah. Yeah,
Ines Lee: a hundred percent. Yeah, exactly. And then when it comes to like more gen AI stuff, uh, and like word-based stuff, um, I think there are two ways we use that. One is turning transcripts into various type of written content and that I find this.
To be particularly helpful when we're trying to, you know, scale Ali's voice, because I think his written style in some ways follows quite closely the way he speaks. And so having a transcript, whether that be, you know, a recording of a Zoom call or like a, you know, a grain transcript or whatever, um, that's very helpful for us, like to capture the essence of the thing he's thinking about.
Um, so yeah, transcript to written format, that's one thing. Um, and then. There are some workflows now where we're trying to, um, also use, uh, get it to [00:13:00] analyze writing samples and really using that to help us build out a brand voice guide.
Ty Magnin: That's interesting. So it's working backwards from existing samples to produce a voice and tone guide for Ali, or Yes.
You know. Nice. I like that. Right? 'cause it's not sort of the conventional way you think about it. You think like, oh, I have to analyze this, come up with a prompt, and then use that prompt to produce other content. It sounds like you are actually feeding into AI examples in order to kind of create a prompt and it works
Tim Metz: well.
Yeah. Just one step back to the process you just described, like how much work is then still involved if you take a grain transcript and you wanna get to a finished piece of content?
Ines Lee: I mean, it really depends on what the work is. Um, so let's say. I don't know. Let's say that the piece of content we're trying to come up with is a, you know, a personal newsletter and Ali had a really interesting conversation with like a coach or something and we have the grain transcript.
You know, I think there's still a lot of like human decisions that need to be made between the transcript to like the newsletter output. And the key things for me are really [00:14:00] like the human's taste in like, oh, what are the interesting and counterintuitive points? I think we're still not at the stage where like the AI can pick those things up because a lot of this is subjective and it comes with like an understanding of, you know, what's happening in Ali's life at that moment that makes this point interesting.
What stage is our audience at that would make that point resonate? And so there's still a lot of like selection that still needs to happen before the output is at the stage that we're completely happy with.
Ty Magnin: It's really good, right? Like, uh, thinking, uh, that AI is missing a bunch of context that can make, especially if it's a sort of personality driven asset like the YouTube show that you do.
Uh, it doesn't have context as much on his personality, like his personal life, other things that he's read potentially, uh, at least not the way that AI performs today. Yes. Um, so I like that. Then you're thinking of, okay, well we have to add that, we have to select one that actually aligns to that. Larger context.
Uh, I think that's smart.
Ines Lee: Yeah, no, a hundred percent. [00:15:00] I, I do feel like, yeah. You know, we, we have this discussion quite, uh, quite a lot at work where, you know, where it becomes cheaper and cheaper to use AI and it becomes cheaper to and cheaper to produce content. What is the thing that will make you stand out?
And a lot of our conversations kind of boil back down to, you know, that element of personality and that element of like, you know, experience where. It's difficult for, for an AI to replicate and, and come up with.
Ty Magnin: Yeah, totally. Right. That idea of like personality could translate more maybe into B2B type businesses because like, it seems really clear how it works on YouTube today with these bloggers or you know, creators.
Ines Lee: Mm-hmm.
Ty Magnin: Yeah. When you look at your friends in B2B or poor souls like animals are, are you like, oh man, you guys are missing it, you should be doing this thing? Or are you kind of like, do you have an opinion on. How it can and, and could be used more in B2B.
Ines Lee: Yeah, no, this, this is a great question and I feel like there are increasingly examples.
Of [00:16:00] founders or co-founders or even execs of P two B companies who are stepping into that more personal brand space? Yeah, I, I, it's become a bit of a cliche of, you know, that saying that people follow people rather than people following brands. And I think that underlies a lot of that movement that we're seeing now in terms of people who come to mind, Adam Robinson, I think that's his name of RB two B.
Um, he's somebody who's popping up quite a lot on my LinkedIn feed, both in terms of like written and video format, both like building in public stuff as well as. Talking about the product he's creating. Um, so he's been, he's been somebody that's of interest to me. There's a lot of synergy between the thing he's building and the type of thing he's writing about.
And I think one of the things, you know, that stands out to me with that example is that there's a very strong founder product fit that that works particularly well for founders and companies who have pro who, who have that kind of fit. Two other examples that come to mind that are kind of more UK context.
I. [00:17:00] These aren't B2B companies. Um, but I think the people doing this, um, kind of embody that spirit of like personal branding that I just spoke of. There's a, there's a big retail chain, I dunno, which we call it retail chain, but have you guys heard of Marks and Spencer's in the uk? It's kind of like a.
Ty Magnin: Trader
Ines Lee: Joe's
Ty Magnin: not me. Yeah, yeah. No, I know what, yeah, yeah, yeah. Okay.
Ines Lee: Yeah. So, so the, the CEO of Mars and Spencers Stuart Machen, I feel like he's been like ahead of the game for quite a while with his LinkedIn stuff. He, he kind of like, you know, speaks about the brand in a way that feels quite personal. And another person is, uh, Dana Strong of Sky Group a, a British media and telecoms, uh, company.
And the Sky, sky group is kind of like putting spend behind her. Social media content. So I think that's a really good example of a world where the founders and the execs of these. Companies are championing the brand and then the brand using employees, uh, or execs as you know, distribution channels.
Ty Magnin: Yes. [00:18:00] I love this.
I'll have to look into both those examples because those are, um, to me unexpected, uh, uh, industries or kinds of companies Yes. That have their exec be so public facing. You know, and again, I'm, we play so much in B2B software, right. That like. That's most of my world. Yeah. It's interesting to see it happening outside of tech.
Ines Lee: No, that that is, that is a good point. I think it's inter, I think it probably comes more naturally. I. To do it in a B2C space because it, it's a thing that you would use day to day. It's easier to speak about a product that feels like B2C, so it's not surprising that, I think a lot of the examples probably come more from that space.
But yeah, I think the pr, the underlying principles are like similar. So yeah, I would say definitely personal brand is kind of one lesson from the creator space. Another lesson potentially is kind of. Treating co content as a product itself. Creators really excel. You know, the content is the product. That is where a lot of them start off in terms of like making money from, you know, whether that be AdSense or partnerships, you know, [00:19:00] both B2C companies and B2B companies doing this, you know, increasingly well.
I. One that's on my, been on my radar a little bit is a, is a company called Doula. Um, and again, he, the, the founder of Doula is a good example of somebody who does founder led content marketing very well. Nice. Um, but if you go on Doula's website, there's like an amazing set of resources on company incorporation in the us which is what, uh, doula specializes in helping people.
But in terms of like the guidebooks that they produce, the, the webinar series and so on, those really feel like educational resources. And I feel like that's a space where like creators really excel and, uh, B2B businesses can also learn to, to, to do more of.
Tim Metz: Awesome. We have an interview with Ian actually.
We, we interviewed him for the blog, so we have a Amazing, we will put it in the show notes. Yeah. Yeah. It is really, really, really interesting guy. I'm also thinking, as I was hearing you talk, like, it's almost like the B2B brands have it backwards. Like, Ali first built his personal brand and he created, so they, they first put all the effort behind.
It's [00:20:00] naturally as a creator, but they built their personal brand and then from that they can launch products. Right. Because you're also, you, you mentioned to me that you're developing an AI driven tool. So I'm curious, like, is that more just purely based on that Ali has such a big audience now, or is it also because of ai?
What's, what's the role of AI in being able to now launch such products?
Ines Lee: So, yeah, I guess taking a step back, I, I agree with you. Like, I think any, any business is. Easier with a bit of a personal brand because part of your distribution is, uh, distribution problem is reduced in terms of Ali stepping into the software space.
I think there are kind of two things. One is just personal interest of, oh, you know, it'll be interesting to, um, step into the software space, but we also see kind of that space as being the space where there's potentially, you know, exponential growth. Whereas with info products, at some point there's gonna like be diminishing returns to.
To what you can get from a course and so on. Um, whereas I think with [00:21:00] the software space, that's, that's really the, the area where we see potential to scale. So that's one thing, kind of like personal interest and seeing where are the opportunities for the business in the next couple of years and potentially even decade.
The second thing I think is, is exactly what, what you were saying, you know, with, with creator led businesses. What a lot of creators now are doing are like partnering up with people who can either, who either have expertise, you know, in operating some kind of service led thing or have expertise in building, you know, software.
And the creator kind of brings the distribution, you know, through say, talking about it in a YouTube video or doing a demo or a, a purely dedicated YouTube video to that product. And the, the person with the expertise is behind the scenes building the thing and making the product as good as possible.
Ty Magnin: Zooming out a little bit. I'm curious, with this software that you're building, I mean, you could tell us a little bit more about it from, but from what I understand, it's something you can use on your own in your content creation process. Is that right?
Ines Lee: So [00:22:00] I'll, I'll speak a little bit. Sorry to skip that step of what, what the product's about.
Um, so yeah, Al Ali's been, um, working on an app called Voice Pal. Um, and the idea of Voice Pal is you speak into it, um, and, uh, it's like a voice memo and you just, you know, ramble a couple of thoughts. And it will then ask you a couple follow-up follow up questions, and then it will create a first draft of whatever piece of content that you want to create.
And the idea of Voice Pal is that this is the ghost writer in your pocket, and it will allow you to transform your thoughts into a piece of written content, you know, as quickly as possible. And for us internally as a team, it's been really helpful having Ali use Voice Pal because he can. Say a couple of things to this app and we will get, you know, a transcript or a couple of notes from that, from which we can then use to create content.
So yeah, so that's the, that's the kind of 62nd spiel of what that is.
Ty Magnin: Yeah, it's very cool. Zooming out a little bit, like, how do you see AI impacting the way that content is created? Like if we take it to that level, like what do you think [00:23:00] the current state is and what does it look like down the road?
Ines Lee: If I had to distill it, they're kind of like two sides.
One is. AI is gonna allow any creator out there to turn, as I was saying, you know, thought into output at a much bigger scale and a much quicker speed. That's one side of it. On the flip side, however, I think it's also going to lead to the proliferation of a lot more generic and adequate content, and the thing that AI is really good at right now is giving us adequate stuff.
By which I mean, you know, you look at it and the the structure is correct, the grammar's perfect, the flow is good. And it's almost like, it's sort of like a, it's like a cover band that really hits all the notes perfectly, but you're never gonna feel that element of like personality, that element of jazz.
So AI is very good at that, and it makes a lot of sense, right? Because these models are trained to all the stuff that's being inputted into these large language models. So as a result of kind of [00:24:00] content becoming more adequate and like generic, I think this is where as well, like content creators. Need to be very, very good at like keeping their tastes and building teams who understand their taste and are able to like protect that.
Because, you know, I think, uh, one concern is like as generic content becomes more and more the norm, that's also our sense of what is good. Protecting that taste and understanding what it is that makes something stand out, I think is gonna be even more important.
Ty Magnin: I love the metaphor of a cover band. Uh, yeah.
Tim Metz: I'm also thinking about the flip side though. Like, are you concerned about a world where AI can easily replicate Ali in a way? Well, either for your own benefit or that other people would do that. Like there's so much content of him out there, right? What if I train my AI model and then I can just say, here's the kind of fake Ali or whatever.
Are you, are you thinking about that? Like how to protect your ip or that it gets.
Ines Lee: Yeah. You know, right now, like as in right now, [00:25:00] meaning pro potentially for 2025 and 2026, not, not a big concern because, you know, even within the team, I think we feel like constrained by the workflow. Like we feel like we're not producing stuff at enough, at, at a quick enough speed at the same time.
However, I think it will, that would potentially be more of a concern as these systems get more, more and more sophisticated. The reason why it's not a concern in the immediate. Future is because we're such a video heavy. Uh, content platform. I feel like video-based AI softwares are not quite at the stage where it's so authentic that it's hard to catch.
So right now in the video space, like the state of the art is probably combining two tools, uh, one of which is called Hey Jen. Uh, I dunno if you guys heard it. It create basically creates an ai, uh, avatar view. Mm-hmm. Combining Hagen and, um, a software called 11 Labs, which is really good at creating AI voiceovers.
Using those two things to create video content, [00:26:00] but the skill that the editor needs to edit those two things seamlessly, um, in order for it to feel really authentic and real is, is, is pretty, let's just say I, I feel like very few people are able to reach that stage right now.
Tim Metz: So it's still a few years off.
At least. We have a few more years, hopefully,
Ines Lee: hopefully. But maybe I'm being overly optimistic.
Tim Metz: Another thing you mentioned related to video and AI is that you, you, you, you were working with an agency to build ai, a agents for your content workflows. Can you explain what you're doing there and what you're trying to, what challenge are you trying to solve with them?
Ines Lee: Yeah, for sure. You know, one of our challenges is trying to click pick from our long form YouTube videos into shorts. And it, yeah, CLIPP picking from both long form videos and potentially podcasts that Ali has been on currently, like we've explored several options. One is like the manual labor of like somebody watching through that, you know, two hour podcast and then finding these 92nd chunks.
That's just like time, time consuming, and we don't quite have the manpower to do that every week. The second [00:27:00] challenge is that when we do try to explore AI tools, of which there are a couple out there, the performance on those clips just aren't. As great as we, we would like it to be. And so we haven't been able to replicate the performance that we have had with like manual clip picking.
And one of the, our hypothesis for why the AI system, the current existing tool out there don't work as well, is that I think they kind of click pick in a very linear fashion. So it's kind of, they take the clip, turn it into a transcript, and then it's like words word base, and then you kind of. Select the words and, and, and the sentences that you feel like would make a good 90 minute clip.
But that again, kind of neglects the context of like, oh, how was he saying it precisely at one minute, 34, you know, what was the facial expression? Because when you're kind of like condensing from a long form clip into a transcript, you're missing that. Even things like to nation, how excited the voice is and so on, which is so important in a short form clip if you've only got those five seconds at the start to hook somebody.
So what we're trying to do with this AI agency, and I wouldn't be able to speak to the [00:28:00] nitty gritty of what they're doing, is basically finding ways that we can combine those first two methods that we've tried the manual clip picking as well as the AI thing, and finding automations and building a system that would allow us.
To feed in, say, a 20 minute YouTube video or a two hour podcast and output 30 to 92nd clips that we could then use for social media contents.
Tim Metz: That's cool. That's your second voice pal. Your clip pal. Then like, if it's ready, we also also wanna have it.
Ines Lee: Yeah. And is it, or is it
Tim Metz: specifically trained? I don't know, like if you know the details on that, like is it specifically trained on Ali or it's just it, it could potentially do that for others as well.
Ines Lee: Yeah, I think we're building it very much in a way that is like con uh, creator, agnostic. A Clipp Pal is a great name, by the way, so you should take that. Um, yeah, so very much creator, agnostic, so hopefully it's something that, you know, if it works, then there's no reason why you couldn't take somebody else's podcast and get it to, you know, find those 30 to 90 seconds.
Tim Metz: [00:29:00] That's very nice. One more in the practical AI part. Is there any specific workflow or prompt that you find super useful that you use all the time in your daily work?
Ines Lee: Yeah, I think my main thing actually is feeding it an example. Of something that I want it to mimic the style of or the quality of getting it to analyze that thing before I get it to produce the first draft.
So the prompt would be, I'm now going to give you a 500 word example of what I would like you to write. The first draft in the sound of, or in the style of. You know, please help me analyze this and, you know, break it down into these categories. And these categories tend to be things like tone, sentence length, et cetera, et cetera.
And then once it gives me that analysis, only then do I say I. Okay, now help me write the first draft following the rules that you've just set for yourself.
Tim Metz: So you first give it, you first, let it kind of think that through before you then move forward.
Ines Lee: Yep. A hundred percent. Yeah. I think another thing for me, you know, this [00:30:00] is definitely a year where I've been thinking more about like, Ooh, what are the things that I do wanna outsource to ai?
And what are the things that I don't, one thing that I've been coming back to a lot, whenever I speak to a team member who isn't quite getting the result that they want from ai. I feel like it's usually because we mistake in kind of writing for thinking and kind of, I, I guess 75% of good writing in quotations is, is thinking and it's only like the final 10% when you're literally typing on, on, you know, on a keyboard.
And the mistake a lot of the times. And the thing I wanna get better at is just really making sure what are the boundaries of like. Doing the heavy lifting of the thinking. And what are the, at what point do you wanna hand off and outsource to the ai Uh, Tim, your like, article on, on the Animals blog.
Actually, yeah, that really shaped my thinking as well in terms of like, oh, what is the point when you hand off and what is the point that you wanna preserve? And I think a lot of the things that I've decided, oh, it's really important, both as a team and you know, for me individually to [00:31:00] preserve is like that thinking.
And by that I mean kind of like I need to know exactly what, what is what it is. I'm saying, you know, what are the key points? Why do I think it's like null one counterintuitive, and what do I think the rough structure of that piece is? Before I even go near, you know, chat, GPD or Claude,
Tim Metz: I, I think that's gonna be a real challenge.
I, I, you know, even though that article was helpful, but I still struggle with it, even though I'm aware of it in a way. But I, I noticed again now recently, like I'm starting to use AI more again, and then like, it kind of gets sucked into it and your brain just kind of becomes lazy. Or it's like you need to find a way to really pull yourself out of it.
It's almost like you need a distraction blocker for social media. You need some kind of thing that cuts you out of AI for a while, so you can. Be more creative. That's such a good
Ty Magnin: point. Yeah. It's easy to be lazy with it. Right? And just kind of dump something and take a mediocre idea and run with it. Uh, but the better ones are gonna be more, you know, they come from deeper or they take longer before you bring it towards some [00:32:00] AI assisted workflow.
Can you walk us through a typical AI assisted day in your content process?
Ines Lee: Yeah. Okay. So let's, let's say we're putting together a YouTube video, and we want it to be an AI assisted YouTube video. So the process would start with ideation and brainstorming. So the first step is like coming up with a really great title, thumbnail and hook that we feel confident about.
So if that process, were AI empowered and typically, you know, we, we begin with ideas that come from us and ideas that, you know, either we've seen on YouTube and we feel like, ooh, it can apply to us if it were AI powered. Then I think the ideation process would take two steps. Number one is like, oh, I have a rough idea.
I have a sense of this title might work well. I would then go to CLA and chat to BT and feed it at the CSV file of our best performing titles and some numbers on like impressions and view counts and say, this has worked well for us and I have this idea. Combine these two things, the CSV file and this idea, and help me, you know, brainstorm 20 more ideas.
So that would be kind of [00:33:00] process number one for ideation. And then kind of the second bucket would be using a lot of these wrapper tools that are currently out on the market. Um, the two, the two that we use the most is, um, something called spotter studio. And, uh, one of 10. So one of 10 basically. Tries to get your video there.
There's a ranking system on YouTube Studio where you try to, you know, be, be the top, which would be one of 10. Spotter Studio is an AI powered thing where it links up to your YouTube channel, it understands the context and the data and so on, and helps you brainstorm titles, either informed by your historical data or YouTube channels in a similar space.
Um, so you can kind of like click on that YouTube video and it will brainstorm some titles, um, and thumbnail concepts as well. Then we'll go onto the writing and research process. If it were AI powered, I think this would mean us giving it a series of prompts that are quite clear in terms of like, okay, I roughly want the hook to look like this.
You know, give me 10 [00:34:00] versions of this 32nd hook. And then in terms of the main, main body, I would want us to go in with a very clear sense of like, these are the three talking points. And then getting AI to like flesh that out potentially, if it is a. Evidence backed research video, like some of our health related ones or even some of the more neuroscience ones.
There's a AI tool called Elicit, or I guess it's more of a website, but basically it helps you search for like the latest scientific information and it does a really good job of, um, finding like the latest papers, helping you analyze it and get a sense of what is the evidence out there. So, uh, we will be using that as well.
And then the final bit is the editing bit. And I think that, um, for, for that process, we'll use some of the tools that we spoke about earlier on our conversation, uh, like fire cut to help us streamline the process of creating the A cut. I think those are the core three steps. So ideation, writing and research, and then that beginning bit of video editing.
Ty Magnin: That's great. I love that there are like steps within the steps too. There's like [00:35:00] workflows within this larger workflow and like, you're not just basically going into clot and being like, you know, write me a script and make me a video on X, Y, z, like, boom. Like that's the road to mediocrity at best. That's probably the right hundred percent, right?
Like you are using it to support, you know, specific steps within a larger workflow and, and you know, you're asking for 20 ideas, not just like, gimme a title. It's like, gimme 20 titles and then there might be two rounds of that, I'm guessing. Yeah. But you're, you're sort of still using. We'll call it a human in the loop, right?
Like to pick and, and run with, um, selections.
Ines Lee: That's a really good way of describing it, like I think. Having an understanding of the workflow, which only kind of you have and the team have, and then being able to break down those steps and being like, Ooh, it's only from steps number one to five that I want to outsource and this, this is the way in which it will support me.
That's, yeah, that's really key.
Ty Magnin: On your team, have you like shared prompts? Like do you have a standardization of this workflow or is it kind of like from scratch every time? [00:36:00]
Ines Lee: Yeah, so it's, it's not, not exactly from scratch, you know, we do have some shared prompts, especially the ones where we're trying to, you know, get it to sound like Ali.
We have some shared prompts across the team, but I would say in terms of workflows, it's pretty individualistic right now. So we just kind of pick and choose what works best for us.
Tim Metz: I guess final question, where should we follow you, Inis, because of course I think people can find Ali, but I know you also have a Substack and so Yeah.
What, where do people go to learn more about you and follow your own writing, which you're also doing. I know.
Ines Lee: Love that. Thank you for asking. Yes. I recently, very recently started Substack called Side Road 38. Um, so I will be on there. Um, I think I'm at in on Substack and I am also on LinkedIn, so there's two places.
Tim Metz: Awesome. Thanks so much for joining us.
Ty Magnin: Yes, thank Youas. This is fascinating. Appreciate all your time.
Ines Lee: Thank you for having me. Great speaking to you both.
Ty Magnin: Alright, Tim, so what'd you think? What'd you learn today?
Tim Metz: Well, it's fascinating to just get a picture on the content team behind a creator. 'cause a creator is essentially creating content and [00:37:00] I had no idea what it, you know, there's a big team actually behind like 10, 12 people.
Yeah. It's interesting to see how they use AI tools, both in terms of. What's already possible and what's not possible. And
Ty Magnin: yeah, I liked her perspective of simply like AI has helped them move faster to create more content. Right. It's been an efficiency gain, it sounds like sometimes, you know, it's saving a couple hours on editing a video, uh, which is significant, you know, if it's like almost a 50% reduction in, in effort or, or in time involved.
And she talked about how AI can help people create a whole bunch of mediocre content. Right. What was her metaphor
Tim Metz: again? That was
Ty Magnin: good. AI is like a cover band. Yeah, yeah. A cover band. Yeah. Right. It's like, you know, they can play the notes and they can play 'em pretty well, but like, it's not the same as hearing it from, you know, the original author of that song That one Hits.
I, I like that metaphor. I might steal it and use it. The other thing
Tim Metz: I had to think about, she mentioned a tool, now I forgot the name, but that looks at the headlines they have for the YouTube videos and the metrics, and then knows which ones that [00:38:00] works well. And then based on that creates new variations for new videos if I understood it correctly.
And I, and that's probably gonna be a very powerful thing to have these tools that get. Really specifically tuned on metrics of what works well, what doesn't work well, and then use that to, to help you create the right kind of content.
Ty Magnin: I love that too. I mean, that's the tip of the day, right? Double down on the wins.
You know, I think she even mentioned loading in A CSV with recent YouTube data, along with then an idea. To help her refine it. So agreed tools are probably gonna integrate that more. That Lex use case sounds amazing. Yeah. But also like try throwing into GBT or Claude and, and just asking it to be smart for you.
See you next time.
