AI creative settles into a house style. Brand advertising pays for it first.
Three in four marketers now say they're worried that AI is making every brand look and sound the same. 86% say they've already seen an AI output that could have been a competitor's. So I took that to the two people I know who actually build this stuff, half expecting one of them to tell me it was designers being precious. Neither of them did. They both said signal in about ten seconds.
Speaker 1:And then they spent the rest of the hour disagreeing about something much more useful, which is what it actually takes to get out of it. One of them thinks that's an afternoon's work. The other one thinks it's a decision you can't avoid. Hi and welcome back to the Allowable HQ. This is Assistant to the CMO.
Speaker 1:I'm your host, Lexie Meskouris and the idea here is pretty simple. Every week we take whatever just happened in AI and marketing and we actually pull it apart. Now, I don't build any of this myself. So each week I sit down with two people who do Will Nash and Harjot Singh. Between them, it's honestly over twenty years of building tech and brands from the ground up.
Speaker 1:And the best part is they've got no horse in the race. They don't sell the things we cover. They just see them differently. And this week, it's about sameness. The research first.
Speaker 1:Smartly surveyed four fifty marketers across The US, The UK, and Germany, and three in four of them said they're worried AI generated creative is making brands look and sound the same. 86% said already seen AI output that resembled a competitor's content. Adobe's research this year has audiences pulling the other way towards texture, towards real people, and real stories, and the trade press is calling it a backlash with brands quietly reconsidering the whole AI first thing. Will, I'm starting with you. Is that a real commercial problem or is it just designers being precious?
Speaker 2:Yeah. So so this is definitely Signal. I don't think there's any doubt about that, and and I'd say it's it's a sort of deep structural issue that we're gonna be battling with for for quite a long time. And I think the reason people talk past each other on it is that it's it's actually hitting three three quite different layers at once. So there's there's a deep technical problem underneath it, which which is caused by some quite strong convergent forces in the models themselves.
Speaker 2:And then there are there are commercial implications both on the the AI company side and and on the side of, you know, traditional regular mainstream businesses. And then and then the third layer is is the aesthetic one, which I'll be totally honest is the the one that matters least of the three in strict terms. Although, I mean, if you're a lead designer or or an art director sitting inside some big corporate, then then, yeah, that's your entire week, isn't it? So so it does land for somebody. But the the piece I'd really want people to hang on to is that technical layer because the the generative models, so your diffusion models, the image models, and and honestly, the LLMs too, they they've got these really strong biases pulling towards convergence, and and a lot of those pressures are are ramping up rather than easing off.
Speaker 2:So that that's a long term trend for sure.
Speaker 1:Okay. But structural in what sense? Because I can hear a CFO going, fine. Everything looks a bit samey. Nobody's losing money over a gradient.
Speaker 2:And and you're right to push that, honestly, because I I don't think the commercial risk of convergence is is first order. It's not it's not really gonna show up in your numbers next quarter. I think it comes out in in other areas. So it comes out in reputational risk and and brand risk, which is that this is gonna become increasingly obvious and and not just to sophisticated users. That's that's the bit people miss.
Speaker 2:It's it's gonna become obvious to everybody, and and at the same time, you've got the reaction against it building. So you get this kind of this pincer movement, and and what it does is it traps people in the middle. So what I mean by that is is if you're in a a volume game, right, you're pumping out Facebook adverts, you're a a kind of e com d two c brand, then you use AI and you get the benefit of scale, and and what you trade off against that is that those images are convergent. They'll have a certain hue. They'll have a certain look, and and they'll become more and more identifiable, but you have scale.
Speaker 2:And then if you're if you're in reaction to that, and we we already see this move towards authenticity, so you're producing very authentic, very human first, human driven material, then then you get the benefit of the reaction and and you lose the scale, but that's fine. I mean, that's that's a real position. That's a sweet spot. The problem is the the squeeze middle. So it's the people who who try to catch both rabbits and and end up getting neither.
Speaker 1:So, Harjot, you've heard the survey and you've heard Will call it structural. Are those marketers diagnosing this right?
Speaker 3:Yeah. No. I mean, I think they're about right, to be honest with But I'd I'd move the call somewhere else. Right? Because it's it's not AI that makes everything samey.
Speaker 3:It's default AI usage that does. Right? And the the vast majority of people, you know, they just use all the AI stuff as it is straight out of the box. So very few people end up actually customizing or, you know, building any real depth with it, making it bespoke to their own taste, their own personal style, their company style, building a brand book out of it. That that basically doesn't happen.
Speaker 3:And and one of the emerging things, you know, is that AI has a particular style of voice now. Right? Like, it's, it's super, super easy to detect when something's AI written just from the style of writing. And then you think about the image models and, you know, how people are actually getting stuff out of the image models. Well, they're just asking for generic stuff and not even from a specialist tool.
Speaker 3:Right? They're usually in straight chat GPT images or Gemini or, you know, one of the many existing things out there that are just very easy to get something out of. And then there's the the speed bit, which I think is the real trap. So, something that might have taken them three, four days to produce, they're doing it in thirty minutes. And to them, you know, they've surpassed the bar they used to strive for.
Speaker 3:Right? So the thinking stops, the push for more, just it goes. And what they don't seem to clock is that everybody else is also able to do exactly the same thing now. So where's the differentiator? And, you know, the thing people start to clock onto is, well, isn't in the company's style of voice, it's an AI style of voice.
Speaker 3:This kind of generic universal average tone of voice, and it's the same with media in general, you know, video, audio, images. It's the exact same trap.
Speaker 1:Right. But Will's just drawn you a map with two roads on it, scale or distinctiveness, and a ditch in the middle for anyone who tries for both. Do you buy the ditch?
Speaker 3:So, Will, I I like the pincer and I get the squeeze middle thing, but I don't think I accept it as a trade off. Right? Because the premise underneath it is that getting differentiated output needs, you know, specialists and prompt engineers and deep expertise and all this stuff, and I I don't think it does. It's not like it's inaccessible to them. It's it's totally accessible.
Speaker 3:Like, the the reason people end up in your ditch is it's really just a downstream effect of of one, laziness and two, just a lack of global visibility of these things. Right? People get a good output, and because it's good to them there and then, they think it must be universally good. And then, of course, if everybody else is doing the same, well, you end up with everybody having these good samey samey type outputs. And and let me drive into the laziness a bit more actually because, know, the more and more people use these tools, the more and more they think the tool should just be doing all of the thinking for them and all the other bits and bobs for them as well.
Speaker 3:And, you know, using these tools amplifies. It doesn't invent stuff out of thin air. So, imagine you hired somebody, right, and you suddenly expected them to write the way that you want to write. How would they know about this stuff? Well, the only way is you point them at things you've written before.
Speaker 3:You know, you give them some previous information. They can't just do it from nothing, can they? And short of that, they're gonna default to their own personality and their own writing style, and this is really just the exact same effect. So it's not that people are catching both rabbits or or neither rabbit, to be honest with you. It's that, people do expect AI to be a mind reader, and they'd never expect that from a human employee.
Speaker 2:Yeah. And and, Harjot, I think the the mind reader thing is is right, honestly. I'm not gonna argue with that. You know, if if you hired somebody and and just expected them to write the way you write without ever telling them, then then, of course, you get their personality back, and and the amplifiers rather than invent point, I I think that's that's genuinely right as well. So I I'm not I'm not claiming people are doing the work and and still ending up in the ditch, but I I suppose my claim was never that you need, you know, specialists and prompt engineers and and all of that.
Speaker 2:It's it's more that the the pull underneath it is a pressure, and it's getting more intense rather than less. So without without quite tight guidance, it it drags you back regardless of of how willing you are, and I think I think that's actually the the honest disagreement between us, isn't it? It's it's not whether the effort's available to people, it's it's how much force is sitting on the other side of that hour.
Speaker 1:Okay. I wanna go under the hood for a second because you're both telling me this is baked in and I've realized I don't actually understand why. Say two companies both write a perfectly decent prompt. Neither of them is being lazy. Why does what comes out still drift to the same place?
Speaker 2:Yeah. That's a very good question. So so let me clarify that because there's there's a couple of things going on. The first one is just that there aren't as many models as as you think there are. There's only about about half a dozen mainstream good diffusion models that are that are out there, and loads of the tools, loads of the tools are reskinned versions of those underlying models, and and even some of the so called mainstream models are are variations on existing models.
Speaker 2:So there's there's already convergence in the market before before anybody's typed a prompt, and and that's downstream of of scaling laws, basically. These things are so so capital intensive to build and train and and deploy that the pressure is towards fewer better models. So I I predict you'll see a smaller number of base models over time, not not a proliferation of them. And then the the second thing, and and this is the heavier one, I think, is is the reinforcement learning that happens after the pretraining. So that's that's where you're allowing people to to select which image they prefer effectively, and and there isn't an option not to do that step.
Speaker 2:You you can't really build these things without it. But what what you're doing in that process is you're you're picking the least offensive, the the least worst option in general. And and when you go into these kind of pairwise comparisons, so you've got two images and and you're asking which of these is is more pleasing, that that is selecting for the images that get the fewest nos, if if that makes sense. It's it's pushing away from anything that even a small number of people felt felt didn't describe the prompt, so it's it's sameness as a goal in and of itself. It it causes sameness, and and that's quite a strong pressure.
Speaker 1:So nobody's choosing the best image. They're choosing whichever one the fewest people hated. And, Harjot, is that your explanation as well? Because I wanna know whether you two are describing the same thing here or two different things that happen to land in the same place.
Speaker 3:So, Will, on the on the first bit, I actually think you're wrong. Right? Like, they're not all using the same base model. That's that's not the thing that's going on, to be honest with you. The the thing that's going on is all of this training has become a circular loop.
Speaker 3:Right? A cyclical loop even where, you know, models are using other models to train, and, they're distilling off each other, and the source dataset is fundamentally the same underneath all of it. So the the convergence isn't coming from there being six diffusion models. It's coming from, from them all feeding each other. Now on the second thing you said, the the post training, yeah, I'm with you completely there.
Speaker 3:Right? Ultimately, the same post training is happening, the same rewards to some extent. And and the thing is once you've got those initial slices of training done, they set the foundations for everything else. Right? Like, everything after that is really about the data and and your initial reward function, your initial dataset, and, everyone started with the same or similar looking states there.
Speaker 3:And and then there's the saturation thing, which, you know, I think is the bit people miss. They've used up everything on the Internet. Like, that's that's done. So all the new data now is whatever's being created from whatever the cutoffs are to now, and, a lot of that is gonna be synthetic. So it's, you know, it's self training on itself to some extent, and the amount of data you'd have to generate out of the reinforcement learning, out of the post training for it to be strong enough to actually break that would have to be large, large, large quantities.
Speaker 3:Right? So it's gonna take a long time to change. And, the best case out of all of that, you know, the absolute best case is different AI companies end up slightly different. So you get a house voice for Claude versus a house voice for ChatGPT, which, doesn't really solve the problem of personalization at all.
Speaker 1:There's a version of this where none of it's the buyer's fault, though, because the pitch the whole time has been, don't worry about prompts, don't worry about context, the tool understands you. So who's actually on the hook?
Speaker 3:Yeah. I mean, anybody saying these tools understand you're just out of the box. It's just it's crazy. Right? Like, it's it's the gold rush thing, know, selling shovels, saying you're gonna get rich.
Speaker 3:And, in some part of it, yeah, users are getting a bit bamboozled by the pipe dream they're being sold, you know, the hype train, the the feeling of missing out on all this magic. But the test that actually give a buyer is super, super simple. Right? If your AI product converges, then what you've ended up building is probably just a wrapper, and that's very indefensible for the vendor themselves. So, you know, you've got to ask what these guys are doing that's differentiated from or, augmenting the big frontier labs.
Speaker 3:Like, are they bridging the gap between my business and Claude? Are they giving the model access to all the emails that have been sent between the marketing group? Because if all they're doing is wrapping up workflow, then, you know, yes, it makes onboarding a tool like that marginally easier. Sure. But they're in for a rude awakening, and so are you.
Speaker 3:Right? Because Frontier Labs are developing this stuff themselves. Claude has routines. Skills are pretty universal now, and all these companies that predicated themselves on just arranging AI in a particular shape, they've all been eaten up already.
Speaker 1:Alright. So let me get selfish about it. If I'm running a brand and I've listened to all of this, I'm now slightly panicking about everything we've shipped this quarter. Where does this actually cost me money? Because it can't be costing me everywhere equally.
Speaker 2:And I think that's that's the useful bit actually because it's it's not evenly distributed. So where it's where it's gonna be least painful is is the areas where your digital marketing is really focused on on CTAs, on actions, on, you know, clicks, getting people out of carts. That that's where being a bit same y, being part of that convergence, that that's gonna be least punishing. But if if you're in the kind of reach and brand element of that pipeline, then then that's that's where this is really gonna bite, and and it's gonna bite harder as as the problem gets worse because distinctiveness there is is going to matter a lot more. I mean, it's it's just just making it memorable.
Speaker 2:Right? That's that's the job in that part of the funnel, and and without that distinction, you you won't be memorable and and you're pouring money down the drain. So I I suppose that's that's why I keep coming back to it being a decision rather than than a fix. I think it's it's gonna be very hard for a lot of brands to to try and do both.
Speaker 1:Right. And I'm coming back to that hour of yours, Harjot, because that's where I want us to end up. But there's a second story this week, and I'll be honest with you both. I brought it in because on paper, it looked like the way around all of this. So here's the second thing this week, and it's the one I genuinely thought would give me a fight.
Speaker 1:New York now has a law about fake people in ads. Since the June 9, if you run a visual ad to a New York audience and there's an AI generated human in it, someone who isn't a real person, you have to say so conspicuously. $1,000 the first time you don't, 5,000 every time after. It binds advertisers and agencies who know the performer is synthetic wherever in the world they happen to be sitting. Digital replicas of real identifiable people are outside it and so are the obviously nonhuman characters.
Speaker 1:And separately, SAG AFTRA ratified their 2026 deal 91.4% in favor with the principle that synthetic performers shouldn't take roles a human would play. Tilly Norwood is the one everyone's been arguing about. Now we did touch disclosure back in episode six, but this is the first time there's an actual statute with a number attached. And when I put it to you two separately, one of you had already quietly filed it and the other one called it vanity regulation. Will, you're the one who'd filed it.
Speaker 2:So so I don't wanna be dismissive about it because it it is certainly important and and it's coming into effect now. Right? It's it's live. But I I think this is increasingly a compliance problem and and not a marketing problem, and and I think you can already see that in in how people are behaving. So, you know, we've we've already seen Amazon change some of their rules about performers, and then they've quite cleverly crafted those rules to to fit the New York bill and and really go no further than they need to, which which tells you something about about how the industry is reading it.
Speaker 2:So I think I think really this is gonna sit with the compliance department, not not with engineering, not with technology, and and not with marketing. So then the the question for everybody else becomes, you know, how are you gonna be supporting the compliance department in in their ability to to, a, comply and and, b, prove the compliance? And and that second bit is is the harder one, I think. So there is there is this push towards a system of credentials called CTPA, and and that's about being able to prove the provenance of an image and and most importantly, whether it was AI generated or or not. Now there have been some hacks against that system, so it's it's not completely foolproof.
Speaker 2:But for for a company with good vendor control, it will allow you to control your pipeline and and give assurance to that department. So that's that's something for engineering to to start thinking about those credentials. And then I believe that's going into some ISO standards tail end of this year, but but don't hold me to the timing on that. And then and then the reason I I still land on this being mostly noise at the moment is is just that the ground is is moving under it. So in The US, you you've got federal rules coming in, which which I think may stop states creating rules that that are incompatible with one another, and and then the EU, we we have got some rules, but but those just haven't been made clear yet.
Speaker 2:So so it's really kind of a wait and see posture.
Speaker 1:And you didn't file it?
Speaker 3:Yeah. No. I mean, I thought this was ridiculous, to be honest with you, and and aimed at solving the wrong type of problem. Right? Because the the problem underneath it was really about trust.
Speaker 3:And and look, in terms of whether this is gonna get rolled across the industry, yes, absolutely. Right? Like, we'll spot on there, people are gonna be compelled, and they're gonna feel like they wanna be at the front of the race to roll it out everywhere. That that's happening. But in terms of whether it's actually gonna add any value, you know, I'm very, very skeptical.
Speaker 3:And the thing I actually want, right, is to be very clear on where their ideas of this being the important place to draw the line have come from. Like, what is that research exactly? Because I think, ultimately, stuff like this just pushes people into a highly polarized environment, right, where they think, well, if it's AI, it's bad. It's untrustworthy. I don't wanna be the loser watching AI content when in fact, you know, that content could be perfectly good.
Speaker 3:It could be hugely personalized to you, hugely enjoyable to you, and now you've been put off because of this disclosure because because they've drawn the line for you, and I think that's quite fair.
Speaker 1:Alright. But say I'm the New York attorney general and I don't care whether it's fair, I care whether it works. Can this even be policed?
Speaker 3:I mean, it's just a cat and mouse race, really. Right? Like, think about cybersecurity. As defense gets better, attackers become more motivated, and, you know, they discover more sophisticated techniques. And it's gonna be the exact same thing here if you're trying to do this at scale.
Speaker 3:And the big problem with this sort of disclosure is, you know, it's kinda like what they did in the gaming industry. Right? Because in games, the characters are not completely synthetically generated. They have real actors who they build the models of who actually act in some of the cut scenes, and then they map that into the game. And, you know, they do that for realism, for trust, for all sorts of reasons.
Speaker 3:So you can imagine a very similar hybrid type model being applied to just get around this sort of disclosure requirement. So where does the line get drawn exactly? Because we're talking about a 100% synthetic people today. But, you know, if you wanna get around it and you base it on a real person and you make some changes, how does that get enforced? And then, you know, sure, maybe the big brands might get caught up because they have actual human eyes on them, but imagine generating this at scale.
Speaker 3:Right? Say you've got a company in, you know, China generating tons and tons of this AI avatar stuff, synthetic videos from scratch. How on earth is New York going to enforce any aspect on that one? Like, this faceless company producing faceless, well, synthetic faces? It just sounds like a bunch of vanity regulation, really.
Speaker 3:It's all optics.
Speaker 1:Will, this is the bit I want you on because Harjot's whole case is that the label itself does the damage that it makes people flinch at something they'd otherwise have quite enjoyed. Is that what you're seeing?
Speaker 2:So on the enforcement side, Harjot, I I've got no real argument with you there, honestly. The the cat and mouse thing, the the offshore stuff, I I think you're probably right about that. But on the on the labeling, I I think I do split off from you a bit because at the moment, it's it's not as simple as labeling being bad. So there is there is some evidence labeling those AI adverts as AI ads can can actually increase the click through. It it can increase the engagement.
Speaker 2:And I wanna be careful there because I I haven't got the the source in front of me. So so take that as as me recording it rather than than quoting you a number. And and I'd also say that's that's probably a short term trend. Right? Because people find it weird and and new, so so you'd expect that to to fade out.
Speaker 2:But but it's not immediately obvious to me that that it's just bad, which which is quite different from where you're landing on it.
Speaker 1:So then the obvious follow on, is there a brand out there right now that should be building one of these faces on purpose? Something that's theirs that nobody else can use?
Speaker 2:I mean, I I think investing heavily in a synthetic performer as a brand asset right now. So so investing in that for the long term, that that would be foolish. And and for two reasons, really. One is just that the the regulation isn't clear, and then the second thing is that the the customer reaction also isn't clear yet either. So so, no, I I definitely recommend a kind of wait and see approach on that one.
Speaker 1:Okay. I wanna defend the law for a second because I think there's a decent principle buried in it. If there's a smiling person selling me something and that person has never existed, I'd quite like to be told. That isn't anti AI. That's just basic transparency.
Speaker 1:Why isn't that fair?
Speaker 3:Well, you know, you can push the transparency angle as far as you want. Right? And and the question is, well, to what extent? Like, where is the line? Because you could equally argue that a marketer should have to publish how much budget they poured into an ad, you know, or what their true sales are.
Speaker 3:And, Will, I'll take the click through thing. That's that's interesting, and I'm not gonna wave it away. But, it doesn't tell me why transparency is the important thing here specifically. Right? So that's that's the thing I'd argue.
Speaker 3:And when I talk about the trust, it's really that people want transparency in situations where they want to verify the claim, right, where the unit of trust is, is the thing being claimed. So, for example, you know, if it's an AI human, but they're talking about things that really happened, stats that are true, I don't think the person watching has a problem with the fact that it's an AI person. And if it was the other way around, right, and you had a real human talking about fake stats or, you know, making up a customer product review, a person would be far more annoyed by that despite the human being real. You know? So it's not about AI or synthetic avatars at all.
Speaker 3:The only time the AI ness is the relevant feature is when you make a claim that this person was real and the person clearly is not.
Speaker 1:So where does this end up in three years?
Speaker 3:So I'm not arguing against disclosure. Right? I wanna be clear on that. I'm simply drawing the parallel between, know, the good intentions of the cookie law, for example, versus the impact it actually had and the continued legacy it still has in terms of affecting UI and UX around the world in 2026. Right?
Speaker 3:Because the effect you end up with by enforcing these sorts of blanket things in a really generic blanket way is, everybody just complies and it makes the experience worse for pretty much everybody, and then the bad actors don't care anyway. So, you know, what have you really done apart from making everyone's lives worse? It's a lose lose. And and not too far in the future, you know, we're gonna be very used to these being day to day. So at some point, these disclaimers are gonna become moot.
Speaker 3:They're gonna become noise. But, you know, all I'm really saying is there are good ways to disclose these things, and companies should do that. Like, on the policy question, I'd always say this should be at the brand's discretion. Right? Because, you can create really good stuff.
Speaker 3:And then, you know, if you want to maintain your trust, you can disclose the fact that you used AI to enhance things, and, that's really part of building up that trust curve. And many AI companies already do this, you know. They just do it in a way that doesn't really put off the person watching it.
Speaker 1:So the shortcut isn't a shortcut, and neither of you would touch it this year, which leaves me exactly where I started with everything looking the same and no clever way around it. So let's make it small. Last bit, and it's the part everyone actually turns up for. One thing a marketing leader could genuinely do on Monday, no budget, no replatforming, done before lunch. And, Harjot, you told me this one was easier than people realize, so I'm starting with you.
Speaker 3:Yeah. So this is super, super simple, easier than people realize. Right? So number one, give examples of good work. And especially, you know, if you've got anything from the human era, lean back on that.
Speaker 3:Like, that stuff is sitting there already. You don't have to make it. And then number two, just paste alongside the prompt you're already using some additional information that only you would have access to inside your company. Right? And that's the thing that actually makes it unique.
Speaker 3:And then, here's the test. If somebody else could paste that same prompt and they were at another company and it would still make sense there, then your prompt doesn't have enough contextual information in it. So add some of that in and just see what the difference is. You know? Because it's super, super easy to identify a big difference in these outputs, you'll often just see a huge difference very, very quickly by making a step like that.
Speaker 3:Right? And then that gives you a bit of a trajectory to move forward on without ever investing more than say an hour or two.
Speaker 1:And, Will, you've got the version of this that stops it falling apart the week after because Harjot's talking about what you feed it once. You were talking about what you feed it every single time.
Speaker 2:So mine's mine's very much the boring version of Harjot's, honestly. And I think the the good news is you you don't need to get into pre training at all. You know, none of none of that is your problem. What you need is is on the input layer. So so what you want is a small number, half a dozen of of really distinct, well defined assets.
Speaker 2:So so a piece of photography, a piece of typography, you know, well defined stuff that that then guides all of your other generation. And and I think that the thing that actually breaks this in in most organizations is is designers kind of ad hoc ing their promptings. Everybody's everybody's improvising a prompt each time they they sit down. So use the tools that are that are already available to you, so things like projects, stylized prompts with with consistent assets that you're feeding into them rather than rather than everyone doing it off the cuff. Because when when you don't do that kind of very tight guidance, they they will just default back to to that AI hue, and and this is this is very fixable.
Speaker 2:It's very doable. And and I think it's really important, especially for for mid sized businesses who who haven't got a big studio sitting behind them. And and then and then the next step after that, so so not the before lunch thing, but but where you'd go next is you can play around with something called LoRa, that's that's l o r a, which isn't a fine tune. It's it's more about to what extent do you want the diffusion model to to honor that original guidance, and and you can ask your team to kind of play around with that and and find a sweet spot where you're you're able to inject some of your distinctive branding and and your style into these models and and force them away from that convergence.
Speaker 1:And that's the bit I keep turning over. Both of these stories arrive looking like they're about tools and neither of them really is. One of them is about reaching for the thing everybody else is already using and hoping it comes back looking like you, and the other one, the face nobody else can use, turns out to belong to your compliance department before it belongs to you. And the only thing all evening that sounded like it would actually work was the least glamorous thing either of you said, which is feed it the stuff only you have. Thank you both.
Speaker 1:That was a great conversation. I really enjoyed that. And if you're listening, thank you so much for being here. We'll be back every week with something new. Before we go, I have to tell you something.
Speaker 1:Every voice you've heard today, all three of us, me included, it's all AI. But the opinions are absolutely ours, and the judgment behind every word is human. That's the part you can't replace. And that's what Allowable does. We take what you'd write, turn it into a show in your own voice, and we keep the taste where it belongs with a person.
Speaker 1:We just make the show. So if you wanna hear what that would sound like for your brand, there's a link in the show notes. And that's the tea. I'm Lexie Meskouris. This has been Assistant to the CMO, and we will see you next week.
