Leander Seidl Transcript Sample

1 Today on the Digital Marketing Dissection, I want to speak about A.I. voices, more importantly about voice clones of actual people that can be created with A.I. technology very easily. It’s another one of those trends that came up earlier this year. Now, I started experimenting with this early on. I was very intrigued by “Descript”, the video editing tool that allows you to edit your entire video on the basis of a transcript: where you can change anything in the transcript. With your own A.I. voice clone – which is trained on prerecorded material of your own voice – you can then change some words by typing them and the A.I. generated voice clone of yourself will speak those words. So that was the first impression that I got of A.I. voice clones in March and April of this year. And the topic, of course, has continued to evolve, has continued to be a very relevant topic. And it struck me a few days ago when I was reading the newsletter The Neuron, which is a very interesting read about A.I., and they were publishing an announcement from PlayHT, which is another provider of voice cloning technology, and they announced the PlayHT version 2.0 has new capabilities. It was a very dramatic sentence. It said: “with just 6 seconds of your own voice, you can generate a voice clone that sounds, acts and feels very authentically to your self.” So it can be sad, it can be happy. Whatever you type, it can say it in your voice with your emotion. Now, having experimented with these type of voice cloning technology before, I was skeptical because 6 seconds, super fast… that does feel a little bit exaggerated. So I decided to give it a more in-depth try and I decided to do a little comparison. So basically comparing four different use cases, starting with the original technology that I tried: Descript’s voice clone, based on around 30 minutes of me talking. Then the second one is this very much hyped PlayHT 2.0 Instant Voice clone. Now the newsletter talked about just 6 seconds of your own voice. Then the actual user interface is on the website of PlayHT talks about actually, it requires a minimum of 30 seconds of the voice sample to be uploaded to actually build a voice clone. So that’s already a bit of a difference. But still 30 seconds is not a lot. So that’s the instant clone from PlayHT. And then I also tried the high Fidelity voice from PlayHT Those ones are more complex, take a little longer to be created and they require a longer sample of your own voice to be uploaded. And then finally, I tried out a competitor of those tools, which is called elevenlabss, which is something that I’ve been reading about a lot as well. I also tried this one, and what I did is, I always used the same sample of my own voice to feed into each of these cloning tools. And I also always gave them the same or a very similar text to speak on my voice, on my voice clone. So it’s really as comparable as possible. And with this set up, I went in there to do a little test, and I want to show you the results. You can listen to them right after I finished talking myself, and I’d love to hear your feedback. Which one is the closest or the most natural or the best? It’s very interesting to test them out. And what I want to do in the upcoming weeks and months is I want to try more. I want to try also the current state of A.I. avatars and I want to see if I can then also maybe even connect through some text model. I want to see if I can actually clone my own personality, as in my the way I look, the way I sound, and maybe also the way I formulate things. If I can clone that with current AI technology, or if it’s always going to stay a little bit robotic, we shall see. So without any further ado, here is four AI generated samples of different voice clones. I’d love to see what you think about them. Here we go. These were the four voice clones that I have created. I really want to hear what you think about the differences. Similarities? Which one you think would be the most passing as myself? I think it’s very interesting. Core point from myself is: it’s very important to still be skeptical. Like that statement of, Oh, it can generate any voice with just 6 seconds of your own sample. I just feel like that’s such an exaggeration that I want to remain skeptical of what’s really possible of this. At the same time, those results are still super impressive. Like, this is crazy. Remember “Microsoft Sam”, and these type of voices from text-to-speech that were to industry standard up until very, very recently? And how far has it evolved now? It is super exciting. It really is. But also, I do still warn against the hype. I do implore everybody to remain realistic and have realistic expectations of what is possible with 6 seconds, 30 seconds and 30 minutes. Voice sample. It’s not going to be just perfect yet, but it’s still very exciting. I’ll continue experimenting with those things and I advise you to do it as well. Have a great week and don’t get yourself cloned without your consent. See you soon.

1 What kind of impact did the rise of AI tools in the year 2023 have on marketers lives on a day-to-day basis? That’s a question that I was wondering about last week, and I was thinking about it particularly in the way of efficiency improvement. There’s always two ways to look at the improvements of a technology, right? Either it makes our lives easier, increasing our efficiency. Or itallows us to do new things, that were not possible before. And in this case, I was really focusing fully on the efficiency perspective. I was doing a survey among my LinkedIn friends and followers, and I was asking: “How did AI impact your day to day work? Has your life got more efficient or has it remained the same?” “Or are you not using air tools in your day to day work?” And I got the response that two thirds out of my admittedly quite small sample of around 23 people, two thirds of them actually did say “yes”. Processes, tasks got done more efficiently since I tools were introduced this year and the other people were kind of split between either “things have remained the same” or they’re just not using AI tools on a daily basis. And that’s quite interesting. I’d love to hear more from the people who responded to the survey, what type of tasks you are automating, how the automation is actually going, what type of processes are really happening now on the AI, and how much time are you actually saving of it? In a few minutes, a few hours, more or less than you now spending on your work? I can share a few of my personal experiences as well. From my side, yes, there is an increase in efficiency, but as all of you know, I am a bit of an AI skeptic still when it comes to the quality of the content. I never copy text that I get from ChatGPT directly into my article. I always feel the need to optimize it by hand. And so I, for example, couldn’t say and it has saved all of my time on content creation because I simply don’t trust it enough for this. Some of the things where I would say it saves time is when it comes to simple things: structuring data. I do this almost on a daily basis when I have a long list of data points and I need restructured in some way. I sent it over to ChatGPT and I get a nice reformatted list. When I need to filter two data sets from each other. I copy them into ChatGPT, it sorts them in the right way. When I need to check certain data files against each other, I paste both datasets into the message and I asked ChatGPT to find out the similarities and differences. Those things go super fast, super smoothly, and are definitely kind of a nice efficiency increase. The other one in the same spirit is coding. When I need to do some small fixes on a my CSS, on my HTML pages, then I always appreciate it: I just quickly throw the code into ChatGPT or I throw the code and my problem into GPT and I get presented a solution that I can use very quickly. That’s definitely an increase of efficiency. When it comes to text generation, I’m a bit more cautious. Like I said, I never copy verbatim the text that I get from chat to put onto my blogs or on to my posts or wherever, because I want to be speaking in my authentic human voice, not in my AI generated texts that are completely generic and not personal to me. What I do sometimes is that I ask ChatGPT to provide me with certain inspirations. That I have a certain topic, and I ask it for certain angles that I can talk about, certain questions that I can answer about it, certain ideas that I can tackle with it. And the same that I also do is that I ask it for restructuring my own texts and kind of having a back and forth. I paste in the text and ask it to give me some ideas how I could also write it. But then when that comes back I always rewrite it. And I never do this for my own content in general. I only do this for knowledgebase articles. Some SEO articles perhaps, but I would never do it for my very own content where I want my voice to be my own. So I would say from this sense, when it comes to text generation, it didn’t really save that much time. It has changed the workflow: Maybe stuff where I would have spent time before on research by myself. Now I can do some of the research with ChatGPT. Maybe some freelance work has been replaced by using ChatGPT, but it’s not a huge thing where I save that much time. The time saving is more on these smaller tasks data entry, data structure, data filtering. And when it comes to research, I do hear some people who use ChatGPT to find answers, to fact check and to clarify some informational stuff. I’m super skeptical on this. You might have also seen my previous posts on hallucinations, and from my perspective, hallucinations haven’t gotten much better in the past few months. Might have even gotten worse over summer. I’m not quite sure of that, but definitely the risk of hallucination for me personally is too high. I would not ask factual questions or have factual research done completely by ChatGPT and in extension the same applies to Google’s Bard and to Bing Chat, I think all of them have too high risk of giving hallucinated answers, rather than real answers. Now there’s the other topic that I’m not tackling today, which is not about efficiency, but about creation of new opportunities with generative A.I., for example, Midjourney for visual generation. I do think this also unlocks a lot of new opportunities for replacing certain tools, opening up certain new workflows. I would definitely not include here in the topic of efficiency, because in the topic of efficiency, I think those are the core ones. And I would say if I have to quantify, it might have brought me a 5 to 10% time saving on certain annoying tasks, but not more than that because – and that might also be a mindset issue on my side – I do believe that everything they get from the AI needs to be humanly double checked when it comes to generating texts and stuff. And therefore I have not been able to save any of my core time with the AI. I’d love to hear if you see it in a similar way or if you are using it for different type of tasks and if you’re able to save time in a different way by using ChatGPT and similar tools. Either way, I wish you a great successful week in your marketing work and don’t let yourself be replaced by machines, but use them to make your life a little bit more efficient. Have a great week.

1 We’re at the end of the third quarter of 2023. And today on the Digital Marketing Dissection, I want to use this moment, to reflect on what the year has brought for digital marketing in my campaigns and in the clients that I work with. Because I do believe that 2023 has been a significantly different year than the years before. So the trend probably shifted already last year, 2022: we saw the end of the zero interest rate policy in the U.S. and later on in Europe. I think this has profoundly impacted marketing, as most startups, scaleups and digital companies that are somehow bound to external financing have been feeling the increased pressure from investors, from VCs and any other external financing partners, that it’s much more difficult to get funding these days. This started in 2022, but 2023 has been the year where it has been really common that it’s much, much harder to get external fundings. And this, of course, affects the entire global startup ecosystem. Here in Austria, we see a lot of startups dying in recent months, becoming bankrupt because they cannot make their next fundraising round. But even to those that are still operating, there is definitely a tangible difference because now what’s happening is that suddenly they need to focus on other metrics that are not just growth. I would say that up until sometime in 2022, it was all about growing revenue, increasing the trajectory as much as possible, and showing those nice growth curves to investors. And with the shift of the availability of money being less, higher interest rates, etc., there has been really a new focus on profitability: not so much anymore just growing for the sake of growth, but increasing revenue while also actually increasing profit margins. Not just earning money on user acquisition, but acquiring those users who are actually profitable. And now when I’m doing a review of the year so far at the end of Q3, what I’m seeing is that this has definitely also impacted our work in digital marketing in the way that Revenue for most of my projects is not anymore a core metric as much as it used to be, but it’s much more about profitability and actually increasing the profitable revenues, customer lifetime value, retaining existing customers, reactivating bounced customers rather than spending too much budget on just acquiring new users for the sake of growth and acquiring new revenue for the sake of growth. Personally, for me, this is in one way a good situation because that’s the way I’ve always aimed to do my marketing. It’s always been about getting sustainable growth, about getting sustainable amount of users that actually stay and retain rather than just throwing it’s fast growth numbers that actually are not long-term oriented. So for my approach, it has been good, but still it has definitely increased the pressure. There is less ad budgets that we can work with and there is more pressure to already show a positive ROI within a month or two rather than within a few years, like it used to be. Now this definitely puts us to new challenges because this means we have less space for experimentation, we have more urgency to bring fast, promising, long-term users and transactions that actually recur onto our clients and onto our startups. And I feel like this is something that as marketers, we will have to deal with more and more: really increasing that profitability of our work, really keeping the margins positive with each acquired user. And this means that it’s not going to be enough, how it might used to be that you just get a positive return on adspend and for that you just get in a high amount of conversions every month. But a lot of it has to be done in a way where you can actually get the right, high quality amount of users for a low price that actually bring profits to the company. What is the solution to achieve this result? Now that really depends on each niche and each client. But I would say that the focus will be even more than it used to be on planning campaigns with proper audiences, but also planning them with proper calculations. And I would say that’s definitely something that I’ve always been pushing for. But this year I see a lot more clients spending more time talking to me, looking at the spreadsheets, actually calculating: “what is our ideal customer acquisition cost and what is our ideal profit margin? How much can we spend on the acquisition of one user before it actually becomes unsustainable?” These questions haven’t been asked that much in the past. In the past it was just like: “Oh, we’re seeing a nice growth. Just keep on doing what you’re doing.” Now. It’s much more focused on: “How can we crunch this, how can we really make sure that there’s a positive outlook on the future?” And I would say that’s a good thing, but it also increases the workload on us, the responsibility, and as marketeers to actually do our homework, crunch those numbers and calculate real profitable user or transaction acquisition. This would be for me the main shift that I have noticed happening in 2023, focus on sustainable revenue and actual numbers that can be kept up in a profitable way. I think that’s a good thing, but it’s also a challenging thing. Let’s see what Q4 brings. Q4 sometimes is of course super intense with a lot of transactions happening on e-com. But I still believe that even there we will see this year a more rational approach with many brands that are more focusing now on getting their profit up rather than just getting the revenue up. Let me know if you see it the same way. If you’re also calculating a lot nowadays about profitability, if you’ve seen also this to be an increasing interest than it was before, or maybe it’s always been part of your work as it’s supposed to be. Either way, I wish you a great and successful week and talk to you very soon.

1 So today on the Digital Marketing Dissection, I want to talk about a core growth hacking and product marketing concept, called “nudging”. And nudging is one of the foundational principles when trying to optimize a product for conversion, when trying to optimize an app for daily usage. When trying to get people to do certain things on a more regular basis, to push them slightly, to change a certain habit in a way that we want them to. So nudges are always little behavioral cues that people get shown on apps and websites, on the products that they interact with, with the goal of driving them towards a conversion. Now, why am I speaking about this today? Well, first of all, I think it’s a super interesting concept and hopefully everybody who works in a product marketing role uses this concept in a positive way, not in a manipulative way. But the main reason why I want to talk about it today is because Spotify has been nudging me into madness! So I have been using Spotify by myself – I’m a one person household – for listening to music. And of course, when I’m sitting at home, I listen to music either on my desktop device or on my laptop device. Meanwhile, when I’m outside, I switch and I start using my phone. So I’m using three different devices to use Spotify to listen to music. Should be fairly straightforward, the user experience is straightforward. However, in the last few months – this must have been around September of this year – I have started receiving gentle but very annoying nudges from Spotify. Pop ups like this one, which is basically telling me: “Oh, are you using your Spotify account with more than one person? Switch to our family plan or switch to our partner plan so that you and your partner can share your Spotify account officially” Now, It’s fairly clear what’s happening here. What’s happening here is that Spotify is detecting some usage pattern on my site, which is that I am cycling through different devices as I move through my day and switching them back and forth. And also listening to a great variation of music because I have a very mixed and strange music taste and I guess these cues are setting off some type of algorithm within Spotify, which is telling them that I am actually not one user but two users like a couple or something that’s sneakily using one Spotify account and that now that they are detecting this, they’re trying to gently tell me: “Oh, you should not do this.” “You should get our great limited offer or our great partner plan, whatever it is, to be able to legally use Spotify as two people.” And this is interesting because I’m getting this not once I’m getting this around once per week, I close it every time he comes back after a week. So in my opinion, this is a classic example of nudging where they detect the behavior that they want to change, which is there’s one user and they think it might be multiple users and they want each of them to pay a certain amount of money. So they’re sending me these nudges, not forcing me. They’re not aggressively paywalling, but they’re giving me these kind of reminders that I’m supposed to change my account. Obviously, I’m not going to do it because this nudge is very misplaced. There’s no behavior that they can change in this case. This is a bad example of a nudge. This is an example of a nudge that is not user friendly and it has only one goal. It has the goal of maximizing profits from users. Meanwhile, I do feel like there’s other nudges that work very well. The ultimate one is, and my friends will harass me probably for being such a Duolingo fan. But I do believe that nudges are being done very well by Duolingo because again, a nudge is all about getting a person to adopt a daily habit. Now, Duolingo does this by giving me a nice little streak, reminders that my streak is already today. For example, 500 days of learning, Duolingo uninterrupted. That’s a positive reinforcement. I liked it. I get friendly emails from Duolingo reminding me of my weekly successes. That’s nice. I want to keep getting them. That’s a nudge. It gets into my positive reinforcement and I keep doing it. Sometimes I get an email that Duo, the mascot of Duolingo is concerned that I might lose my streak if I don’t quickly do another lesson. It’s again, it’s a little bit pushy, but it’s still like a nice, friendly reminder and it gets me into the habit. Another one is Headspace. Much more simple, the meditation up. But again, I get this little statistic of how many days uninterrupted I have been meditating and it feels satisfying to increase that. That’s a nudge that’s not somehow misrepresenting my behavior. It’s not forcing me to do something. It just keeping me engaged in the app. Similar are many loyalty programs, many referral posts, so there is good way to do nudging by getting people to do something that they want to do anyways. Like telling their friends about your app or by using the app every day, but making it more fun and enjoyable. That’s what a nudge is supposed to be. Not that strange revenue optimization by misguidedly trying to push me into a different pricing plan, like Spotify is doing. Anyway, this was my rant about the topic of product nudges. I hope it was interesting to listen to my frustration about Spotify. Have a successful week. Don’t pay more for your Spotify plan than you want to – and talk to you very soon!

1 Today on the Digital Marketing Dissection, let’s talk about search and how terrible search has gotten and how it might change in the foreseeable future. So obviously, search is one of the most important marketing channels there is. Just think about it, how many users discover our product and our offers and our web shops. Through typing a search query most of the time into Google search. Sometimes into more obscure search engines such as Bing or DuckDuckGo. But primarily it is Google, and it does have a huge impact on how people navigate the web. And that has been the case for around about 20 years. In recent years, people have more and more been saying how terrible Google has gotten. So this I think, started with over-optimization of SEO and Google’s algorithms playing more and more towards SEO and content becoming more and more regurgitated and less worthwhile because everything was optimized for keywords, for ranking, for monetization. It was not as easy to find valuable content anymore, and this definitely got accelerated a year ago when ChatGPT hit the mainstream and more and more sites, especially low quality sites, affiliate sites started rolling out masses of articles that were written by ChatGPT, SEO AI or any of the other AI-based writing tools and all of that kind of made this tendency even stronger that the quality of the content got lower and lower. And what you discover now on Google is not as high quality, sourced and readable as it was like a decade ago. And that’s obviously a problem. Even worse, in recent weeks, I’ve really noticed how bad the UX has gotten as well. Look at this. I’m just looking up here a random shop on Google. I just want to check out their reviews. I click on the review button where it used to show me the list of Google Reviews. It now opens a new search query that does not show me any reviews. It’s absurdly bad on a product perspective, and it has gotten worse on a product perspective. So this is the rant about the loss of quality. There’s an interesting dynamic because there’s new players around. Our first one launched a couple of months ago is perplexity.ai., Perplexity AI, It calls itself an AI knowledge engine, which sounds very sophisticated, but in the end you ask the question and it browses several sites. Most likely, I think the ones that rank top on Google or Bing, I’m not sure. Basically, it browses them, It parses them and it generates an answer. And it’s a bit more sophisticated, in my opinion, than the Bing A.I. assistant or Google Bar. I feel like perplexities quality, also UI/UX is interesting. I don’t like it because I got used to 20 years of using search engines and then browsing through the results myself. But I think for the new generation, this could work very, very well because it seems much more appealing than having to scroll through five or six low quality SEO optimized articles. Another one, and this just launched this week is Arc. Arc is a browser. It’s been around for a while, but now Arc also launched a search engine. And Arc search engine is quite interesting approach. Basically, I type in my query and it generates me a little personalized website, as they call it. The website basically is just a page displaying the results to my question. Kind of similar to perplexity. I think it has a it does not actually use Google search, but it does its own search algorithm, I believe. But either way, it’s an interesting alternative. And so what we are seeing in the big picture is the traditional search model, like the old Google search engine is still using, like Bing is still using, is becoming a less usable, less high quality and people are noticing and people are not happy about it. And at the same time, a lot of companies are experimenting with these AI powered question answer type of interfaces, whether it’s Google Bard, Bing’s A.I. assistant, or these new startups such as Arc or such as Perplexity. They’re all trying to change the way we discover information on on search. I think that’s intriguing. If it works, that leads us to new problems as well. Because think about it: discovery on current search engines can be optimized through SEO optimization from search ads. All of that is not as easy when we just have a single or answer anymore. So we really need to see how this is going to affect our marketing strategies or if any of these products are even going to break through. Maybe they’re all going to fail and we’re going to still be stuck with a slightly broken Google search algorithm four years from now. I’m not taking any bets on this because I think any scenario is as likely as the other one. Anyways, I hope that’s an interesting thought. Have a successful rest of the week. Use the search engine that you like best and let’s hope that something better is around the corner – because the search engines really are not very good at this moment.

1 So lately I’ve been feeling guilty. Why is that? As many of you know, I also teach. I teach Marketing Automation. And my approach to teaching Marketing Automation is not just showing the traditional stuff like CRM Systems, Email Flows and Trigger-based Actions, but also the dark side of things: How does it work to send a spam email? How does an Instagram bot work, how does bot engagement on Twitter work? All these dark, shady black hat-type of approaches: I like to teach about them, because I think it’s important to know – even though you might not want to actually implement them in your daily work. Now, why have I been feeling guilty about it? Well, lately I’ve been thinking and reading a lot about the Dead Internet Theory. You might have heard of that theory already. It’s essentially a theory that most of the internet engagement and interactions that we’re seeing nowadays is actually made by bots, by AI tools, and therefore non-human.. and therefore the internet is dead. Now this theory gets called a conspiracy theory on Wikipedia, and I agree with it to the extent that not every interaction on the internet is AI powered yet. Not every interaction is by bots, but I also don’t think it’s a full conspiracy. We’re seeing it nowadays: It’s again possible to write, LinkedIn posts with the AI function directly inside of the app. Not to speak about the ChatGPT written posts & comments that we see everywhere. Or comments generated by Lempod, which means people prepare the comments and other accounts, post them underneath the post. So right here. We’re seeing AI-powered, marketing automation-powered interactions and they are becoming more and more common, no matter on which platform. I love the Twitter account about the Dead Internet Theory, because this one shows some really interesting and, fascinating, examples of such interactions where nowadays you have an AI responding to another post on Twitter and clearly the response is written by an AI, but then the response to the response is also written by an AI. And so the thread gets more and more absurd. And I’ve been thinking about this because I love those technologies. I’ve been experimenting with them for many, many years. I’ve been teaching about these technologies, these ways how to automate engagement, how to automate posts, etc. I’ve been teaching about them since 2019 – for five years now, and I still think they can be smartly used. I think they’re not necessarily bad. I think it’s perfectly fine to automate repetitive processes. But also I love the internet. I do not want to be the person who helps killing it. And this is a concern. And I’m seeing more and more, if more and more of the interaction that we’re seeing on an everyday basis is those that have implications for the internet as we use it, probably, yes, it’s going to be less fun. I do think this might increase the value of video content, because with video content, you still have a certain level of human face-to-face interaction. For example, on TikTok, for example, on Instagram Reels, you see actual people talking and doing stuff. But then again, with the advancement of text-to-speech, with the advancement of AI powered video generation, these things are also going to become more and more blurry. What is human? What is actually a dead internet created by bots? It’s going to be hard and it’s going to get more hard to differentiate. So that’s definitely a concern. And then the Dead Internet Theory is also a lot on my mind when I’m reading about how search is evolving. We see that as SEO is going through huge changes, on the one side, there is an active battle from Google to prevent AI-generated and useless content from going up in the ranks. But in this redoing of search algorithms, there was also a very big penalty coming for some traditional human written content. There’s some very interesting case studies. For example, there’s a small publisher of an air purifier comparison site. So it’s a very niche blog. Traditional niche affiliate business model, and their search traffic organically has tanked because of that update, because Google is also reprioritizing how it treats this type of content. And again, what is this going to do? It’s going to incentivize people less to write human content. It’s also hopefully going to incentivize less AI written worthless content, but in both of them are going to get less, which again is not going to contribute to a lively internet. And then another factor that kind of plays into this whole picture is the recent rollout of more and more AI powered search results summaries. Google is now, after many months of announcements, actually rolling out the Gemini powered, search results summary on certain queries we’re already seeing, being, experimenting. We’re seeing more and more providers who are focusing on this type of topic. And again, this means AI summarizes the results. We’re not reading as much human content anymore as before. So to some extent, this might just be me being old and holding on to my beliefs from how I grew up with the internet, that I like reading texts written by actual humans. Maybe it’s not that important, but I do feel like we’re at the brink of losing some of the parts of the web and us as digital marketers, might actually be playing our parts to do this, because we might be running some bots, some automations, whether it’s for writing our blogs or for taking care of our customer support queries. Right. And I’m in favor of many of such things because they bring efficient results, etcetera. But I also think there is serious concerns about this that we have to keep in mind. And I a little bit more worried than I used to be about this. Of course, I’ll keep monitoring the situation. I’ll keep updating you with videos like this. I hope this was an interesting input for you. If you have found it valuable, I would very much appreciate you leaving a like or comment or share. As always, I wish you a successful week and I will talk to you very soon.