Meta's Creative AI Update And The Benefits Of Using AI Personas
Ready to stop guessing and start growing? Join Foxwell Digital Membership today, and plug into a network of expert advertisers who are just as committed to success as you are.
In this conversation, Will Sartorius, CEO of Selfmade, discusses the recent AI update by Meta along with how AI personas are generated and used for performance creatives on Meta and TikTok.
Will shares on how AI personas have improved their hit rate and provided guidance for their strategists. He also goes in on the prompts used to generate AI personas and how it's still very important to have a human strategists in the AI process.
Key Takeaways:
How Meta's AI update allows users to optimize images and test different headlines and descriptions.
Why AI personas have improved SelfMade's hit rate for generating effective creatives.
The specific prompts Will and his team use to generate AI personas without any outside data
Why AI personas can be so valuable in guiding the ad creation process
Why testing new personas is important, even if clients may be resistant to trying something new.
Will's top five AI tools for ad creation and content optimization
How AI can streamline the production process and empower offshore teams to become effective strategists.
To learn more about Will and the team at SelfMade head here:
Want to learn more about the Foxwell Founders Membership? Head here.
If you'd like to learn more about the Founders Community or want to become a member you can do so here.
Full Transcript:
Tris Dyer (00:00.21)
obviously.
Will Sartorius (00:01.678)
So, yeah, I live in New York now and spent my time between Connecticut and Vermont growing up.
Tris Dyer (00:07.73)
Nice, nice. Give me like the two minutes. What the hell is tick season?
Will Sartorius (00:13.07)
man, so it generally happens from, you know, April to June, it gets really quite bad. And that -
Tris Dyer (00:14.162)
Ha ha.
Tris Dyer (00:23.09)
Okay.
Edwin @ Snappic (00:23.513)
Wait, did you say tick? Like tick like the bug? Like tick season?
Tris Dyer (00:27.186)
like the really really really famous song
Edwin @ Snappic (00:31.897)
stick season.
Will Sartorius (00:34.957)
sorry, I thought you said tick.
Tris Dyer (00:38.034)
aha, yes, that's my thick Irish accent.
Will Sartorius (00:40.108)
Yeah, yeah. Yes, so stick season is pretty much October, November. We also call it, you know, more colloquially mud season. And then it's generally March and April on the back end. And it's just like, you know, we are in very, very northern Vermont. And you can imagine the dirt roads just get obliterated. So it's just like this massive holes in them. You know, you probably need better tires during.
Tris Dyer (00:51.378)
Okay. Okay.
Will Sartorius (01:09.9)
stick and mud season than you do in a ski season itself. So yeah, it's just like the time you don't want to be up there. Ski season is sort of over. It's just like a guaranteed toe at one point or another.
Tris Dyer (01:11.89)
Hey.
Tris Dyer (01:17.362)
Gotcha.
Tris Dyer (01:21.394)
area.
Tris Dyer (01:27.858)
Okay, okay.
Edwin @ Snappic (01:28.217)
guaranteed toe that is that is quite something I never encountered a guaranteed toe
Will Sartorius (01:33.099)
Yeah.
Tris Dyer (01:36.338)
Yeah, we're going to say, yeah, that's it. You're kind of stupid to go up there if that's the case. But there you go. And right. Look, we'll kick straight into it. And so that, like I said, you got the pre flight stuff. I'm just going to throw it myself just to make sure it. Yes. Just make sure our heads are on the right line. Yeah, that's fine. We had in the early days when we were doing this, we had someone that was like here and then someone was like up there and then down there. It just, it didn't look professional. So it's all good. So yeah, look, I'm going to introduce you.
Will Sartorius (01:42.795)
Totally.
Tris Dyer (02:06.29)
I'm going to say CEO of Selfmade. So you guys generate AI personas. That's going to be a massive one that I'd love to talk about. So AI personas and the performance credit. Want to hear all about that. That's going to be amazing. When you said that, I was like, we didn't actually have a huge amount of AI in there, but definitely want to hear more about that. Was there anything in particular that caught your eye?
Will Sartorius (02:14.089)
Yep, for sure.
Will Sartorius (02:24.68)
Awesome. And yeah, just for some context. Yeah, just for some context, we don't, I saw some things on TikTok and YouTube. you know, we're generally entirely focused on Meta. So like probably can't provide the best insights there. but you know, obviously Meta had a pretty big release yesterday regarding, you know, generative AI images and headlines. And like, I think that'll be a pretty big change for a lot of folks. Yeah.
Tris Dyer (02:50.098)
Big tongue. Absolutely. Yeah, for sure. I think that's going to be completely different to what we had before, especially when it comes to actually using it because once it becomes available to everyone, everyone gets a go at it and then goes, I don't really know how to use this properly, like professionally.
Will Sartorius (02:54.471)
Will Sartorius (03:09.863)
Totally, totally. And there's been so many small AI startups that have tried to build this exact thing. You just put an image in, swap out the background, and you have an ad. And yeah, it'll definitely make some pretty big waves in the space.
Tris Dyer (03:17.266)
Yeah.
Tris Dyer (03:24.594)
Be different, different gravy, different gravy. Okay, cool. And just confirming on our side then, Edwin, I'll do the intro. Do you want to kick off with the first question? So have that ready? Will you our episode number 15? And we're going to talk about all things creative, starting with potentially AI or meta, which is really exciting. Okay, so what we do is we generally go for about, like I said, about 30 minutes. As I said in the intro, put your hand up if you want to re -record.
Edwin @ Snappic (03:47.897)
Tris Dyer (03:54.034)
What we do then at the end, and I haven't said this because I had to just say it on the call, at the end, I'll do a wrap or Edwin do a wrap of whichever will work that at the time. And then what we do is say, I'm Trace, I'm Edwin and you say, I'm Will. Thanks very much. We'll close off. All good.
Will Sartorius (04:08.452)
Okay, got it. Yep, sounds good.
Edwin @ Snappic (04:09.465)
well, one other thing, you're going to be chatting and then in the middle, we'll, we'll probably have like some reactions. Like I might go like, no way. Like, yeah, like, right. Don't, don't get thrown off by it. Like if it just like, just keep on plowing. and then I think, I think that's it. So I'm, I'm, I'm good. I'm good. Whenever you guys are good.
Tris Dyer (04:29.65)
Yeah.
For sure. And so, Edwin, just confirm me one of the things that Shane scolded me on the other day. It's Foxwell Forum, Foxwell Founders Forum. Or is it Foxwell Forum? I just want to confirm.
Edwin @ Snappic (04:38.809)
Yeah. Yeah. Fox, founders forum.
Tris Dyer (04:46.418)
Founders Forum, right. This is why I'm checking because he's right. Okay, cool. Mark and we'll go because we're sorry, we record everything just because, you know, we end up getting chat in the middle like shit. We never hit record. So, okay. Hi, guys. Welcome to the episode 14 of Foxwell Founders Forum. This is weird this week with Will Sartorius. He is the CEO and founder, I believe of
Edwin @ Snappic (04:47.833)
thunderstorm. Yeah.
Will Sartorius (04:48.898)
That was horrible.
Will Sartorius (04:59.811)
You
Tris Dyer (05:15.154)
shit, we'll start that again. You're not founder, okay, cool. Cool. Hi guys, welcome to episode 14 of Foxwell Founders Forum. We're here this week with Will Sartoris, who is the CEO of Selfmade. So they generate AI personas and performance creatives for meta and TikTok. And we want to hear everything there is to do with when it comes to creative and when it comes to just generally what's working out there right now. So welcome Will, thanks for coming along.
Will Sartorius (05:17.697)
Just sit here. Yeah, yeah. Yeah.
Tris Dyer (05:44.658)
Edwin, do you want to kick us off with your first question?
Edwin @ Snappic (05:47.961)
I this just hot off the presses. I want to hear from you all because you are with AI all the time. Meta just did this huge AI update. Tell us about it and tell us how are you using it?
Will Sartorius (06:02.014)
Yeah, for sure. So, you know, haven't had the opportunity to test it out just yet, but it's certainly an exciting change. And I think, you know, especially with catalog and dynamic product ads, like this is going to make a material difference. you know, just for context now, and I believe everyone should have it by the end of the year, but you can go in, upload a photo and then meta can optimize your background for that photo. So you can, if you had a water bottle or coffee cup, you can say,
I want this coffee cup in the rainforest, or I want this coffee cup on a ski slope, whatever it may be. So that's one big component that is going to definitely drastically change sort of the equation in terms of generating statics. And even in the past year, two years, we've seen a lot of GNI startups trying to do exactly this. So Meta is really getting ahead of the curve here. And having this additional functionality is really going, in my opinion, going to just...
flood the market with a lot of statics. Probably we'll see, you know, CPMs climb there, but in general, it's going to give brands much more of an opportunity to test a fair amount more. Yeah, and then I also think, you know, a big component of this is also going to be like sort of headline and description testing. The ability to...
Edwin @ Snappic (07:15.065)
All right.
Will Sartorius (07:27.003)
test numerous different headlines against top performers, descriptions against top performers. Generally speaking, a lot of performance creative agencies, ourselves included, aren't writing the headlines or the descriptions. So this will give brands a tool in their arsenal to come up with a plethora of headlines and descriptions to test out to determine, OK, we have our winner here. We should be using this winner with X, Y, and Z creatives.
Tris Dyer (07:54.258)
Absolutely. And so just on that then, so you're obviously you're creating creative from before and that's obviously, you know, being fed into meta. So you're talking about a flood of influx of statics and so on, but specifically talking about AI. So you guys have must be doing something similar to this for a while with mid journey chat, GBT, this kind of thing. How does that differ from those? Or is this something that meta has got in platform now that you can start to start to click one button and away you go.
Will Sartorius (08:21.496)
Yeah, so this will be in platform. And so it'll be at the ad creative level itself. So you're going to be able to go in, select this AI feature within ad creative. Assuming you have a great image of your product, all you have to do is upload that image, and then you prompt how you want this image to be backfilled. So it'll generally generate six to 12, I believe, variations of that. And so you can sort of...
Tris Dyer (08:43.154)
Nice.
Edwin @ Snappic (08:47.385)
Whoa. Wow, variations.
Will Sartorius (08:50.806)
Yeah, pick and choose the ones you actually want to test. And again, like there are a handful of companies that are doing this already. You know, we've seen Suna really lean into this sort of AI image generation with the photographs that they shoot in -house. But like having this as a tool in the toolkit will really enable.
a handful of brands that generate statics at a scale that they weren't able to do so previously.
Tris Dyer (09:21.714)
Mm -hmm.
Edwin @ Snappic (09:21.977)
And so you are you are working with AI personas on on meta. And so I have no idea what that means. So tell me, tell me, tell me what it means and how do I make money off of that?
Will Sartorius (09:26.453)
Correct.
Will Sartorius (09:37.109)
Yeah, it's a great question. And really anyone can do this. The data is out there. So what we do for any client is, well, we'll have a first meeting and the client will say, you know, these are our personas we like to focus on, like these are USPs, et cetera. And then we say, okay, that's great. But like, let us actually collect some data and sort of confirm your biases or sort of say we should be testing these new personas. So what we do is we collect, we pick,
Edwin @ Snappic (09:40.441)
Okay. Okay.
Will Sartorius (10:06.739)
three to five brands within the vicinity of your brand competitors. And what we'll do is we'll collect all the reviews from TrustPilot, the website. We use a tool called GigaBrain. You can search your brand on GigaBrain and that will allow you to pull all the subreddits that mention your brand or your competitors brands. And then for better or for worse, who knows how long until a reddit takes down GigaBrain, but you know, it's great.
Edwin @ Snappic (10:29.113)
better for worse.
Will Sartorius (10:36.786)
we recently started working with a company called brand wise and with brand wise, you can download a CSV of Facebook comments. So you have all this plethora of data for your brands and your competitors brands. So then what we'll do is we'll create a custom GPT for all of our clients and input all of this data. And through a series of prompting, and we just have, have like an automatic prompting system. We can determine like what your key motivators are, where your key values are and what your key,
Edwin @ Snappic (10:36.953)
Okay.
Will Sartorius (11:06.576)
pain points are, and then also what your customers' frustrations are, their irritations, and sort of their points of no return. And that's always great for our strategists to say, these are the components we should be focusing on the ad. But what the most important component of this is, are the AI personas themselves. So we have sort of two batches. We have the evergreen personas. So these are the folks that are buying from your brand again and again. So...
Edwin @ Snappic (11:07.993)
Okay.
Tris Dyer (11:12.274)
Mm -hmm.
Will Sartorius (11:36.335)
You know, for example, you could, you know, going back to coffee, this could be, you know, the, you know, the caffeine addict would be a persona, right? And like, that would be the angle to test. So we will generate four verticalized, evergreen personas. And then we'll also generate four sort of what we call like diamond in the rough personas. So what does that mean? These are folks that have used your product in a specific use case, but they haven't necessarily used it.
or more, I should sort of go back for a second. They have used your product in a use case that you haven't necessarily advertised towards. So, for example, we work with a company that sells meal kits and you can boil the water five minutes, the meal's ready. And generally we're speaking to folks that have no time between meetings, but there's a specific use case about camping.
Edwin @ Snappic (12:32.537)
Yeah.
Will Sartorius (12:34.988)
And they're like, you know, I brought this on camping and like, it made my life so much easier. This was like sort of, again, like a diamond in the rough. Only one person mentioned it. You're like, so maybe we should be testing top of funnel ads for this sort of persona. And so generally we'll come up with, you know, four of these verticalized diamonds in the rough persona to test top of funnel ads. And then with the evergreen personas, we're doing full funnel approaches because we know that these folks have purchased from your brand in the past, but continue purchasing from your brand.
Edwin @ Snappic (12:37.465)
Okay. Yeah.
Will Sartorius (13:04.715)
So it allows us to have handful of evergreen concepts as well as plenty of new top -of -final concepts to test on an ongoing basis.
Edwin @ Snappic (13:12.441)
So does this make, so I'm gonna ask you the very blunt question. Does this make money? Like does it, have you seen the needle move past what, you know, past the regular stuff when you're doing the regular exercises and you build the personas, you know, back like the Flintstones versus now with AI and this wild stuff, does it move the needle?
Will Sartorius (13:36.361)
Definitely, yeah, it has improved our hit rate immensely. And I think with the Evergreen personas, like candidly like that has been great because like that allows us to, you know, first month or two to generate concepts that we are pretty confident are gonna work. These are, these are.
Edwin @ Snappic (13:37.689)
Yes!
Edwin @ Snappic (13:53.113)
Give me numbers. Give me numbers here. Like what are we talking about? When you say move the needle, like what like ballpark? What are we talking about here?
Will Sartorius (14:00.583)
Yeah, I mean, we're targeting, obviously, above a 1 % CTR, above a 2 ROAS, and above a 20 % thumb stop. So if we're not getting that, we're iterating. And so the evergreen persona's off the bat. But then also, something we found in our early days was that we're delivering a set number of creators every month, right? And our strategists sort of need some sort of guidance. They, you know,
Tris Dyer (14:12.146)
Mm -hmm.
Edwin @ Snappic (14:23.769)
Yeah.
Will Sartorius (14:30.342)
Obviously they're checking motion, they're checking Facebook ad manager, but that's going to not contain data about new concepts or like bigger swings. So having these sort of guard rails for our strategist to say, no, this is a persona we want to test this month. Like these are the USPs to utilize and these are the angles to take with them allows them to sort of not get overwhelmed with the paradox of choice and actually really refine.
concepts on a monthly basis.
Tris Dyer (15:01.106)
So that's a good point that you make there actually just around like these are the guardrails that your strategist can look at and go well, yeah, you could literally endless amounts. So what are the data points that they're using? What's like, you know, with AI we've seen this, right? And you can see this across your chat DBT. If you just ask it a dumb question, you're gonna get a dumb answer. The same with, I'm sure with this Facebook AI, but then when you're creating personas, same idea. If you're, you know, that old adage of shit in, shit out.
Well, let's have a look at it in this one. What information are you putting in to guarantee or to at least give a good probability of getting a good persona that is actually going to start moving the needle?
Will Sartorius (15:42.562)
Yeah. So we have refined our prompts a million times, but the biggest thing is just to reiterate to the model to not input any data from outside of the data that we're inputting. Right. So like it is relying entirely on those reviews. It's relying entirely on those subreddit comments. And, you know, moreover with our motivators, I use in pain points. We.
say like, this is how many reviews speak to each motivation. And then we provide an example of an actual review, right? So like, we are just synthesizing the data. And this is something we used to do pre -AI as well. It was just much more of a manual process, right? We'd collect reviews, we tried to find themes in it. But this has just allowed us to scale this portion of the business, you know, at a much faster pace. And doing it, we...
Edwin @ Snappic (16:22.713)
Yeah.
Will Sartorius (16:34.687)
done sort of the side -by -side tests, the manual tests versus the AI tests, and always have landed on a very similar place. And what's often great is that a lot of the personas that we find align with what the brand is anticipating. Like, we know this is the persona we've wanted to test, which also builds trust. It anchors them and says, OK, these guys know what they're doing. So then they can say, OK, the remaining personas.
clearly are ones that we didn't know about and we should be focusing on. So having that overlap makes a big difference.
Edwin @ Snappic (17:11.353)
And so what are these prompts? Like no gatekeeping, what are these magical words to whisper to the machine?
Will Sartorius (17:19.358)
Yeah, for sure. Again, let me just pull them up real quick. They should have had them. So.
Tris Dyer (17:24.338)
All right, take your time. We can cut that so you can take your time. It's all right. Let me hit the power of showbiz.
Will Sartorius (17:28.894)
Okay, cool.
Will Sartorius (17:33.724)
Okay, so we upload all of this information and we subscribe to a team, GBT, so we can upload as much as we want. And we'll say, based solely on the data I provided earlier and using no outside information, come up with three motivators and key values. We define motivators as desired outcomes or purchase prompts. Values are defined as unique benefits and advantages, delightful features or deal breakers.
Pain points are defined as issues that other competitors are likely facing that the clients are solving for. They should be positive, but they should not involve customer service or delivery. They should be succinct points about the brand. And then we reiterate again, do not use any outside information from the reviews provided for. Insert brand here. There were X reviews in total. And then in your output, include how many reviews speak to that specific theme and include one review that best exemplifies the theme.
There is no need for any additional analysis, just those two elements. Also, for the review provided, do not make something up. Only use the reviews I provided for beta, as in pull one of the reviews I listed above, and then I provide an example of a pass prompt, or like a pass output.
Tris Dyer (18:42.226)
Hmm.
Edwin @ Snappic (18:43.801)
Wait, well, tell me, how much do you charge for this? Like, how much roughly do you charge for this?
Will Sartorius (18:49.881)
Yeah, so we white label this to a few agencies that just like uses on a monthly basis to get new personas. But we for all of our brands that we cover, it's just covered as our, you know, within our monthly retainer. So even like our low end brands that are paying us 1500 a month for four creatives are receiving this analysis.
Tris Dyer (19:11.954)
Okay.
Edwin @ Snappic (19:11.961)
You're being too modest about this guys. Will just gave you the formula for like, I could totally see this being packaged for like $10 ,000. Like, Tris, have you ever charged? Like, how much have you charged for this kind of work?
Tris Dyer (19:23.922)
No, not for stuff like this, but this, I mean, realistically, this is where I'm trying to dig into, I'm trying to understand with these personas, it's not just a persona, because brands have this idea where there's like, I have a mad persona, this is our ICP, and we're going after, sorry, perfect customer, and they're going after that. But they forget that they have to make new and unique ways of actually coming up with new angles and hooks and everything else, and they're different personas that you're going after. So.
This really makes it a real life situation. It's like any company that has values that stick it up on the wall is like hungry and things like that. It's like, that doesn't mean anything. This is the same idea on personas, which is really interesting. I've never tried it like this. I'm really excited.
Edwin @ Snappic (20:02.105)
Bro.
I've seen people charge $20 ,000 for this kind of stuff. Like, like debt, like I've seen an invoice for like $20 ,000 for this kind of stuff. Not even like people saying it. I've seen like an actual invoice.
Tris Dyer (20:14.29)
haha
Tris Dyer (20:17.874)
Nice, nice, nice. So tell me this. yeah. Okay. I was gonna say just getting away from the charging for this. What's really interesting on this is, you know, how, so how do you balance between results that are happening in, in the, like in the live kind of platform versus how your actual, your new idea is coming up, right? Cause you've got to be iterative, but when do you become, when does revolution become evolution and vice versa?
Will Sartorius (20:19.572)
crazy and like in a sort of sorry, go ahead.
Will Sartorius (20:44.211)
Great question. So we like to sort of think of those as two separate pillars, right? And in our medium package, which is 12 craters a month, we've actually built a model that we just call it the iteration tracker. And so our strategists are really focused on new concepts, right? They're looking at the new personas that they receive every month for top of funnel. And I should reiterate that if we start seeing a persona doing well, for example, like that camping angle, then we'll build the full funnel.
We want to start top of funnel and start getting spend. And then if that's working, we'll do the middle of funnel, social proof ad, and then we'll go bottom of funnel, statics, gifts, whatever. So that's sort of one side of the business. And then we have what we call this iteration tracker. And so what that does is it plugs into your Facebook ad manager and it finds the wounded birds, if you will. So these are ads that maybe have great ROAS.
These are ads that have great CTR, maybe okay watch time, but have a very mediocre hook. So the model ranks everything with a letter grade. So it'll say row as a, you know, CTR, a watch time B and then it'll say thumbs up D. So the model will flag this and say, Hey, you know, this ad needs a new hook. So it'll automatically create a task within our backend and it'll say to the strategists, you know, here's the ad, here's the issue and like, here's how to fix it.
And so that's just on an ongoing basis that we're constantly iterating on these like sort of wounded birds. Additionally, like if we see performance creative start facing creative fatigue, then we'll, the model also sort of flag that and say, this ad has been performing really well, but you know, ROAS has been dripping. Clearly it needs an iteration. You know, here's the ad, here's the recommendation. So it's really two sides of the same coin, right?
Tris Dyer (22:33.298)
Mm -hmm.
Will Sartorius (22:35.791)
The ads you're receiving on a monthly basis are either going to be these persona top of funnel ads, or they're going to be generally the first few months we're going to focus on evergreen, sort of those top of funnel, middle of funnel, bottom of funnel approaches.
Tris Dyer (22:51.186)
Gotcha, gotcha.
Edwin @ Snappic (22:53.209)
And so you're using, obviously you're using the AI personas to help guide the ad creation process. Let's talk about the actual production process. Are you using AI in the actual production process as well?
Will Sartorius (23:09.005)
Negative. So all of, yeah, so we have strategists and we have editors and so AI is just our guide. I don't know if y 'all have tried like, you know, add creative .ai, but it's just not there yet. I think.
Edwin @ Snappic (23:10.073)
Okay.
Edwin @ Snappic (23:14.297)
Okay.
Okay.
Edwin @ Snappic (23:22.361)
thank god. I thought I was the only one. I was like, I'm behind, I'm behind. Okay.
Will Sartorius (23:27.533)
No, yeah, it's just, you know, we're definitely keeping our finger on the pulse in terms of ad creatives. I think, you know, generating B -roll will be very helpful in the future. You know, we obviously, we subscribe to Storyblocks and artlist .io, but being able to create more customizable B -roll for ads will be great. We've also started experimenting with UGC AI. We'll utilize Arcads and Creatify. Arcads is definitely a bit better than Creatify.
Edwin @ Snappic (23:33.369)
Okay?
Edwin @ Snappic (23:38.457)
Yeah.
Edwin @ Snappic (23:46.105)
Okay.
Will Sartorius (23:57.324)
and those are good, not great, but honestly having, and we've also found, I know we've talked, you know, want to talk about creative performance as well, but like we've found that opening up on a talking head generally stops, you know, the thumb, you know, so having the UGC at least for the hook has been very helpful. also, you know, the green screen educational sort of top of funnel ads perform exceedingly well, especially with a brand with a lot of nuance, you know, for, and then.
Tris Dyer (24:11.858)
Hmm.
Will Sartorius (24:27.115)
or brands with not a lot of nuance. We work with OpenStore, which is an aggregator. They have a handful of brands, and then one of them is Jack Archer. And this is a company that is very similar to True Classic, but it really leans into humor. And so we did this educational green screen ad with the UGC AI talking about how tight pants will lower your fertility. And we pull up studies and we have the UGC AI talking about, this is a sperm you want, this is a sperm you don't want.
And like just going over how wearing loose or more comfortable pants will just change the game for, you know, your relationship and your future offspring. So like, you know, that's where I see like, did you see AI playing a big part is like these more in a green screen educational sort of top of funnel ads.
Edwin @ Snappic (25:08.633)
Hey, okay.
Tris Dyer (25:18.962)
Yeah. Yeah. Yeah. Who said this, this podcast wasn't educational? Wear looser pants. So, I mean, guess, like this sounds, it sounds amazing, right? It sounds like next level. Sounds like things are really kind of taking that next level, not just through AI, but obviously the approach that you're taking, you know, with your company in terms of the, in terms of showing how you not just creating creative to say, this is a new idea. This is new. It's creating creative purposefully for unique people.
Will Sartorius (25:23.673)
Yeah.
Tris Dyer (25:48.722)
I guess, you know, what are the common challenges that you see? Like, where does it go wrong?
Will Sartorius (25:55.174)
Yeah, it's a great question. And I guess I should sort of zoom out for a moment and sort of talk about like how this really solidified is like we were at, or I was at a conference, you know, past year, I don't remember exactly what it was. And there's this panel and they were talking about how important personalization is in retention. And in retention, you have a lot of personalized metrics. You know, where folks are clicking, you can create, you know, custom.
profile settings for your folks. And like, if you had a million people working for you, you could send a customized email to every person. Right. And then I, you know, raise my hand and ask the question, like, how are you coming up with personas for paid? And they're just saying that their response was, well, you just have like weekly brainstorming sessions. And like, that was just like such a disconnect for me because like he was talking and really validating the concept that like personalization is absolutely key for retention.
And it's allowed him to scale his business and it's company, you know, his clients businesses. But just say you're doing a weekly brainstorming for pay. It just like felt like, how can you sort of have both going on at the same time? And so the idea behind sort of the AF personas is to, you know, for our agency partners to just catalyze that conversation with creative strategists. And so again, like giving them guardrails.
Tris Dyer (27:09.81)
Mm -hmm.
Will Sartorius (27:22.499)
And so I know that's a bit of a sort of a tangent. And I don't know, Tris, can you sorry to repeat your question if you wouldn't mind.
Tris Dyer (27:28.338)
Yes, I mean, where does it go wrong? Like how does, you know, with AI, I'm sure there's loads of opportunities to go off in this way, and not just with AI, with making personas and making creative, where do you mess up? Well, not just you, but where does it go wrong?
Will Sartorius (27:31.668)
Yes.
Will Sartorius (27:42.529)
Totally. I would say where it can go wrong is if you're testing new personas and they don't work. And that's when clients are like, we know it works. Like we have these evergreen campaigns that you've created for us. And why don't we just create more of those? And so that can be challenging because especially with clients with lower budgets, they just want
Tris Dyer (28:01.746)
Hmm.
Will Sartorius (28:12.127)
the same ad a thousand times just told a different way. because they're like, this is what lower CPAs. This is what drives conversion. Why aren't we doing this? And so to say, we want to come in and test these four new personas entirely that you may not even think are relevant to your brand can be challenging. Cause the client may say, well,
Edwin @ Snappic (28:16.249)
Right.
Edwin @ Snappic (28:23.065)
Okay.
Edwin @ Snappic (28:33.177)
Okay.
Will Sartorius (28:37.15)
I don't think that this is our target persona, or like, I don't think that this person makes sense for our brand. But regardless, you know, we always implore them to at least give it a shot, give it some spend and, you know, some of the time it works, some of the times it doesn't.
Tris Dyer (28:51.217)
Nice.
Edwin @ Snappic (28:52.409)
Wait, so what do you do? Like what do you do? You just tell them like, like, trust me? Like, trust me, I'm a lawyer. Like, trust me, I'm a doctor. What do you do?
Tris Dyer (29:01.33)
the
Will Sartorius (29:03.453)
Yeah, it's, it's, it's generally a conversation of saying like, listen, we're just testing, you know, this top of funnel concept, like there's not much budget going to it. There's limited risk. and if it works, then we build out a full funnel approach and then you're going to start making a lot of money. Right. But like the risk is limited. And also, you know, I should preface that I'm not a media buyer, but we recommend.
Edwin @ Snappic (29:11.225)
Yeah.
Edwin @ Snappic (29:14.937)
Okay.
Will Sartorius (29:31.42)
doing a, you know, a CBO and an ASC and to test all new creatives within a CBO, right? And so like, if you're doing a CBO test, you're not forcing any budget to any particular ad set. So like, if a new persona isn't working within the ad set within the campaign, then you're not wasting any money, right? And so like what, what's, what's the risk of the test in the first place?
Edwin @ Snappic (29:41.593)
Yeah?
Tris Dyer (29:57.522)
Yeah, true, true. It's like testing of any description.
Edwin @ Snappic (29:58.873)
And so, for the sound bite, I wanna know top five AI tools that you're using at the moment, excluding chat GPT. Tell me what are your top five go -to AI tools at the moment.
Will Sartorius (30:14.746)
Yeah, I mean, 11 Labs, obviously, I think everyone is sort of all about that. It's essential. You know, we've cloned founders voices. We've been able to create so many more founder ads, you know, using 11 Labs. Like that's been excellent. Descript has been really helpful as well for our editors and just peeling a green screen from the background and also just transcribing and picking up sound bites for ads.
We'll actually use sometimes to script in conjunction with GPT. We'll get the transcription and then we'll optimize the add -in GPT. And then more recently, the ArcHeads Create -a -Fy combo for UGC AI, that's been very, very helpful. It allows us to create UGC AI sort of like that. And then GigaBrain, I mentioned earlier, that's sort of the web scraper for Reddit.
Edwin @ Snappic (31:02.489)
Mm -hmm.
Will Sartorius (31:13.082)
And then finally, I think this is five, Browse AI is our web scraping tool for collecting reviews and collecting Facebook comments. and I guess I started this one more. BrandWise, I mentioned them earlier, and they're great for scraping Facebook comments.
Tris Dyer (31:31.41)
Nice, nice. So it sounds to me like through all of that stuff there, original content is, it's a thing of a past really. You've got AI creating all this stuff, iterating on all this stuff. A lot of the content actually being created sounds like it's down the toilet.
Edwin @ Snappic (31:31.449)
Okay, okay.
Edwin @ Snappic (31:42.841)
Will Sartorius (31:49.242)
Yeah. So yes and no. Like, again, all of this is just helping us dictate strategy. Like we re the only content we're creating with AI is are the UGC AI personas. But like, at the end of the day, AI is helping us guide our strategists, but the strategists are still the ones that are really coming up with the scripts. They're the ones that are saying, you know, this is a really funny hook. We should use this. So, you know,
Tris Dyer (32:13.234)
Hmm.
Will Sartorius (32:19.38)
It's a tool, it's a stepping stone. And we have a very large offshore team and it's empowered our offshore team to become strategists at the same caliber to any US based strategists in a matter of months. And that's been a massive unlock. But yeah, in terms of the content itself, our strategists are still the ones that are going through Lucid Link or Sweet.
studios or Google drive. And they're still the ones that are picking out the clips that make the most sense for the ads. They're the ones that are maybe going to story blocks or list on IO and finding the B roll for the clips. And like really just trying to take all this information, make it succinct and fit within the thesis that the AI is helping drive.
Tris Dyer (33:08.338)
Gotcha. Gotcha. That's yeah, that makes gives me a lot of heart to really make sure that we're not just throwing it into AI and that, you know, that's what it is. Like there's actually some there's some brains behind it as well, which is really nice to see. That's it's always encouraging. Edwin, do you have any other questions?
Will Sartorius (33:26.579)
I think, yeah, Tracy may be losing your hair.
Edwin @ Snappic (33:26.713)
Sorry, I think you're cutting out a lot Tris, can you tell me, say it again?
Tris Dyer (33:32.594)
so I'll just respond to.
Will Sartorius (33:41.875)
Amen.
Edwin @ Snappic (33:43.449)
Yeah, it's a well, can you stick on afterwards? I want to show you something actually, and I want to get your thoughts on it because it's actually something that I've been testing and maybe it'll help. Maybe it'll help you, too, but I want to hear your take on it. I'm going to I'll pull it up.
Will Sartorius (33:49.011)
Yes.
Will Sartorius (34:04.563)
Love you, guys.
Tris Dyer (34:15.259)
Hey, I'm back again.
Edwin @ Snappic (34:17.529)
All right, cool.
Tris Dyer (34:19.347)
Are you hearing me?
Will Sartorius (34:19.347)
Yeah, he seemed to be crystal clear, yeah.
Tris Dyer (34:22.419)
Course and crystal clear. Damn it. OK. Shit. OK. I think I'm not going to be able to put any marks on this now because I'm back in what seems like a separate room. OK. Sorry. I'll just respond to that again. I was my my response was just I was going to be. Thank God there's some brains behind it. It's not just all chucking it into a. I mean, that was a scary moment right there that we're going to start handing things over to the bots. But it's good to hear there's a good process behind it.
And then my next question was going to be, Edwin, do you have any other questions?
Edwin @ Snappic (34:57.017)
No, I'm good. I'm good. Should I outro?
Tris Dyer (35:02.451)
Yeah, yeah, do you want an outro? Will, that was fantastic. That's really, really good.
Edwin @ Snappic (35:04.569)
Yeah. Yeah. So well, the way we will outro is I'll do I'll do a little spiel. I'll say until next time, my name is Edwin. Then Trish will say my name is Trish. And then you'll say my name's Will. And then I'll be like, we'll see you on the next one. OK, cool. All right, guys, we had Will on the founders forum today. We learned a ton of stuff. AI personas off the wall.
the tools that you gotta use, that you gotta be using now. It was a lot, guys. My name's Edwin.
Tris Dyer (35:39.411)
I'm Tris.
Will Sartorius (35:41.324)
and I will.
Edwin @ Snappic (35:42.745)
We'll see you on the next one.
Tris Dyer (35:44.147)
See you later.