The 2026 Meta Playbook: Creative Formats, Personas & Scaleable Structures

We're doing a lot of mashups now that are specific to certain personas and then we are doing lots of variations through like more video or saw a based hook variations. Yeah, where do we want to start? Do I want to start on creative? Let's just start on what's working right now, some just top level bits that are very actionable that people can take away. So one thing that this just seems very obvious, but I haven't really spoken about it much on the podcast, at least in Q4. I guess it kind of gets forgotten about sometimes now, but like mashups. So I know that sounds very, very obvious, but really starting to bucket creative assets by persona because we're doing a lot of mashups now that are specific to certain personas. And then we are doing lots of variations through like more video or saw a based hook variations. So again, just a more one that I wanted to raise because we've seen like a lot of top spend is coming through in multiple accounts that are feeding into that. And I think a lot of the time we're trying to reinvent the wheel and new kind of audacious big swings or using AI to recreate concepts when really like you're sitting on a gold mine with the assets that you have available to you. Especially now like for all those brands that have been deep in this creative volume game for six to eight months like their asset libraries are going to be huge. Like if we look at some of our longest standing clients, just thousands of videos in there. I think a lot of that is an exercise in like good file management is the boring stuff because it's hard if you've not done that to search through obviously you can get support from like frame AI and things like that. But yeah, I think this makes a lot of sense. Like you say, but booking by persona, how are you approaching the is it is it is it building a script first and then finding assets to piece to that or are you approaching it from a different lens? Yeah, so it's kind of modular in a way. So we the kids, it's like we're not a lot of the time this isn't just like sitting within your cloud storage or frame or whatever whatever you use is the in its simplest form of say you don't have that like back end asset management setup. So we're actually going into the add account and especially if you're using naming conventions by persona like what one that we did is we we looked at and for one of our accounts we literally just sorted by persona top spenders over the last 90 days downloaded all of those video ads. Obviously, all of them follow like a some form of kind of you know an iteration of of an aid of framework, let's say, and then from from say those 10 15 ads, there are there are hundreds of variations that you can do once that you file those into like a modular approach. So obviously hooks problem agitation solution, so et cetera, et cetera. So that's that's the way that we approached it from that through that lens. I will say that we are working on a lot right now in terms of that asset management side. Obviously, we've we've recently hired a director of creative opts who's working a lot on this in q1. So there'll be a lot more value that we'll be able to share there in terms of how we're structuring that going into into the the the second off of q1. No, it's yeah, I think that's great as well because it's looking in woods and like doing more with less and cost of creates it just like try to drive that cost of creative down it's obviously easy just to continuously add resource. It's like if you can make that stretch further through those through the through the asset library makes a lot of sense. Yeah, absolutely. And I think the key there is that we're we're we're utilizing new hooks to obviously get around the creative diversity piece or similarities. So that's just something to consider as well. I would say like the most consistently top performing founder framework that we're seeing working in across all of q4 and I've seen this from multiple of our accounts now is the is the founder podcast objection handling ad. So I knew you're going to say that I'm not close to the data but I've seen multiple of ours. I've also seen a number of other other brands doing the same way it's just direct call out of biggest objection in the hook right. Yeah, exactly all like it opens with a controversial statement linking to the problem. So I won't say the brand but the hook was. Would you let your kids start the day guzzling a kind of code yeah and but then it's like the skeptic comes in and starts pushing back with core objections and then the founder is is there is the the person to to combat those objections. Conflicts I guess which we've also we've also talked about how previously working well I think the founders concepts is again it's probably a number of brands watching the tried it and shelved it but it's just always works when you get it right. I would really recommend as well like if you're doing that with if you're the founder listening your working with the founder you someone on your team. Do a scripted approach to that but also do and do an ad hoc version where you are just throwing objections at the founder and just seeing what you get back as we did that with Dan from heights who's on the podcast a couple of weeks ago. And we got some like really nice moments from from just actually going a little bit more spontaneous with it. And layering in satire as part of that work quite well like he brought like a bit of a sarcastic edge to some of the the objections which which I think just helps drive that controversy as well. So another just another iteration format that we'll see starting to see really work is essentially like an X would you call an X post now is it I still say a tweet. Yeah, it's a tweet. But yeah, it's essentially where you have a headline which is a an X post or a tweet but then you have the video underneath it like someone had posted a video onto X. So again, it's like a if you have a best performer that's working within the account take that best performer put it into an X video with a new headline that sits above it. So it's still a video ad, but you're incorporating it into a slightly different variation of that format. It's interesting. I imagine there's other ways to do that as well with other other sites or the visual iterations of that. Exactly. I feel like there's just a there's just a forever evolution with it when it comes to those native formats. So a lot of those reworking well again over Q4 and into Jan actually we've done some of the classic Black Friday email statics what note from founders. If you're a wellness brand, you can reuse that format through Q1. So it doesn't need to just be used over the Black Friday. It's just again the copyrighted and the intention in the message needs to be needs to be right. Definitely seeing that work well at the moment. Yeah, absolutely. And then the final one is green screen reaction ad. So it's essentially a creator that is overlaid on top of another UGC ad that is reacting to the UGC behind them. And I think this was I haven't seen these for a while, but it's something we started doing again. I know we, you know, this is something that has been done a lot in the past, but not for I haven't seen it for a period of time. And that's working really well. So it's almost the individual that is watching the creative behind them, but then reacting to it in the form of a testimonial. So almost like this was the ad that I saw that took me on the journey to to solve X problem. Makes sense. Yeah, they've a lot of brands were doing them a lot of volume of that in early early 20s. It's just not seen as many of them. So again, it's quite easy to do makes assets won't go further makes assets work harder. If you can just get a couple of creators to turn those out, makes a lot of sense. Where do you think high prods going to, going to go next? We've obviously got the, we've got like these proven we've got the founder, we've got the podcast, we've got the ring doorbell if you cast that as high prod, but. Is there any like innovative, innovative, like bigger swings that we've seen there come through? Yeah, so again, I'm just so bullish on like taking YouTube formats and applying them to high prod. So there's a few that we're doing there right now. One is, have you seen, well, actually, it's more it's YouTube and TikTok, but have you seen the subway takes guy? I guess you're not. No, it's not. Yeah, it's a ghost. Not even a ghost. It's not there. So it's essentially a guy that like sits on the New York subway. And someone comes with like a controversial take and he says like 100% agree or 100% disagree. And they were kind of just like, it's, it's quite like a funny interaction about it. It's terrible. My mind went to sandwich when you said subway. No, it's in a subway man. That makes sense. So that's one that we're testing. Again, it's just like an evolution of like the agree to disagree concept, which was an evolution of like objection handling podcasts. And so we're, we're leaning into that as one option. I think another one is like thoughts from the bridge again of really famous or organic TikTok account, which essentially is like a person holding an old pay phone. And the camera is like from really far away and it's slowly zooming out and it's almost like flying the will fly fly on the wall style. Yeah, where they are kind of quite authentically and vulnerably opening up about something. We're seeing that working for like more emotionally driven brands where there's like a clear problem solution. And the individual is like talking through their problem and then into the solution. And I would say like, yeah, generally there to the stuff, two of the things that we're testing right now. But I think that the dialogue, heavy, I know this, your thoughts kind of going down that, how, how do we, how do we break through AI almost in that high prod style? Yeah, yeah. What's the, what's the, what's the next evolution of it? Like, where does it go? I think, I saw, take the holiday, put a tweet out about this, I think, about the concept of like live streaming and where how big he thinks that's going to be as like a medium of creative because that's impossible to well hard, very, very hard to AI. And you see like people like that speed who is like the pioneer of live stream content and how that becomes like millions of clips. And like he believes he was saying that he believes there's going to be a brand built through that lens. Like that's going to be their primary means of content distribution and then that's going to be called their strategy down. So is that the end state of like the furthest possible you can get away from AI live in the moment content? Yeah. It gives you to be creative as a business, not just in a process, which I thought was quite an interesting thing to think about. But I'm sure there'll be other big, big like, what's the next podcast? There'll be something this year that's the next podcast. I think it's just when we find that. Yeah. I think like it's really difficult to say, but I do just I still believe like how if we can through the lens of like emotional customer testimonials, I still just see that being really hard to like replicate. Yeah. I also think the one example of that is that that Perez ad that did the rounds of them interviewing customers on the phone. Like we've reproduced, we've iterated on that a few times a client. And that's the format that again, if every brand got it right, it would probably work for most businesses. Yeah, exactly. It's especially like the solution first brands. Yeah, I do think it's like you want it. You need to be careful not to take too many big swings and try and do just be throwing out crazy concepts and focus on those fundamentals. Like we always say, but the interest is to see where it goes. Yeah, I still think we're like I still think we're I think UGC is getting really close now. But my product is although from a visually, it's is very close. I still I still think there's a disconnect there from it being like dominating and driving performance over high production as a mechanism. There's like elements of those high products that will be driven through AI such as B-roll and things like that. But like our CAD 2.0 is a great example of like they're really getting close now and it's I'm quite impressed with it. Yeah, I think it's getting I've had a very this is probably a whole episode. We should probably do a whole episode on AI tools that we could we've tested and did we did that once and it did really well. Yeah, we should do it again. Yeah, I think it's close for anything that's like quite still quite formulaic. I think it's not it's not close to like TikTok shop long form narrative storytelling. But I think that is hard to replicate, but also it's going to be an interesting path this year. See where we get to. Yeah, I you make a really good point there like those native TikTok shop startups. I think again like that that stuff is the way it's produced is is very difficult difficult to replicate from an AI perspective. It's not that like single creator one shot where they're in front of camera and usually it's like you know we're we're camouflaging it through more overlays and things like that. It's almost the rawity of it makes it more authentic. Yeah, I'm definitely seeing all the demo. I see that that whole AI creator been I've already seen it in a couple of accounts. I've audited whether if they're selling to over 45s they're just they're using a lot of AI generated you GC because they just that's all the difference. Yeah, it's good but as you get younger it becomes becomes more obvious. Any more creative bits or no, let's jump into some AI stuff. It's a good segue. I wanted to just touch on creative testing first. I think that's a segue from from this. I don't want to go into I feel like we could go into a complete black hole around what's the right creative testing set up in accounts. And to be honest we do it differently in accounts depending on where's our constraint. Do we is our constraint we don't have that many assets and we're trying to learn and we're trying to drive better feedback to create better output or is the constraint that we have so many ads that we need to be careful not to spend too much percentage of our budget in testing and therefore inflate the overall CPA of the account. So you can do that a bio you can do that CBO you can do it as see there's so many ways to just to cut number one question that we can. I'm a big fan of a bio but I think CBO is very useful when when you have to like when you want to be very intentional with like I'm only going to spend this percentage of budget. I want everything to take a minimum spend to give me a signal. But we prefer a bio by concept and but one of the things I'm definitely noticing is is. Promotion of assets to core campaigns been a challenge at times like something can perform amazingly well in an ABO testing and then you'll spin it into a ASC or whatever that core. Driver of scale and performances and it will either not take spend or just won't pick up performance like like it was previously so I think it's really important right now to. Be still be trying that and obviously hopefully that that doesn't happen in the account but still scale up anything that works in testing like we we used to this a lot on tick to scale individual concepts. You want to be maximizing that horizontal scale as well as the vertical scale and your core campaign because it's just it's proving to be quite limiting. And it may not be a problem like right now because you might have an asset that is like just taking all the spend in that ASC but I think you can get into position where you've got spend centralized so much that incremental reach starts to become a problem. Just before we move on because I know there'll be some people the next question off the back of that is the feeds into what you just said is before we're like horizontally or vertically scaling what how are you deeming a scalar like what what is that I know that again differs brand to brand but. You know if you were to like generalize that to some degree what what should brands be thinking about there when they're promoting winners. It's significant spec you want to get significance so it needs to be on or within a certain percentage of your CPA target after significant spend but that significant spend is relative to your AOV. So we'll usually we'll usually set a daily budget and let things run for a specific like three day period but the budget will be set relative to the AOV. And then we can make a decision on whether it's it should be promoted or it should be scaled up or whether it should be just left like maybe that performance needs to maybe the performance is tending towards the right CPA on insights and therefore we may leave it for a few more days see if it reaches that point. But yeah that's the approach we take I think also as you. There's a lot of nuance in account decision making like you did I think once you've spent a lot of time in another counter you understand the brand you can start to look at leading indicators of performance and make quite informed decisions around whether you need to spend money on that set anymore. It's like if you put something in after like a day or two it's spent a couple hundred quid and it's meaningfully lower CTR and like or there's a couple of correlating metrics that like are so off. What you know to be good in that account you can you can make faster sharper decisions but see an element of intuition. Yeah I think there's a bit of intuition in other counts certainly as you get more volume. I think what what that shows though is that like add set wise and campaign wise that there's an element of like delivery by persona and that whole Andromeda piece. Either either meta still isn't able to effectively deliver in an AC across multiple assets consistently which is sometimes the issue. Hypothesis there maybe like it's centralizing all of the spend to the absolute best performer. But it would actually be but there's actually like multiple in there that may reach multiple audiences but if that makes sense. So your ceiling would be higher if it was sharing spend across because you were reaching to completely different audiences in that campaign. So what this is given more and more rise to for others is just ASC by persona scaling campaigns by persona that back down into like you testing strategy. If those personas are super clear and you've got consistent winners in both. Right so persona on a campaign level I think yeah I think that's once you starting to push we are obviously working with a lot of brands in the UK that are pushing some significant spend for this market. And when you look at the other count it's like if if another agency ordered it there'd be like what the hell is going on in here like it's an absolute mess. There's no best practice like but we need to do it because it we need to there's metrics we need to ensure happen today to drive performance two three weeks down the line. And a lot of that is around new audience reach frequency CPM incremental reach. And I think Facebook isn't matter isn't great right now when you at scale if you have everything centralized to one campaign because it can often just for spend to one persona. So there's just just something to test out if you're hitting that incremental reach. Limitation you're pushing five multi five figures in spend you're struggling to reach our next layer I think really think about your campaign self through the lens of persona. I think it can help you get to that if you've got a good creative strategy that backs that backs into that can help you get to that next level of volume yeah that makes a lot of sense I think the key there as you said is like. Take a step back to the route though which is actually ensuring that there's there's real clarity around that person those personas yeah always speaking about those personas and how that feeds into the strategy. Yeah that has to be I think there has to be enough differentiation. And that drives the campaign self I do think there's a lot of probably an opportunity in some accounts to even test it test it for like the same persona but just different like very different vehicles and hypothesis would be that that may work as well. But I do think there's still leverage in that account set up especially as you start to spend more your volumes increasing quite a lot to worth experimenting with another quick hack. Shout to Ben who actually put us onto this who we work with on a few accounts. Big cup he gave it the name the big cup graveyard where you put all of your past top performers big cap relative to your CPA and just leave it. It's not very consistent but what it'll do is in periods of peak demand a lot of the time it will pick up incremental volume and it'll spend and it may spend a lot. So we can pay days it can just capture a quite a fixed cost fixed rate low risk extra spending volume that you just wouldn't have captured otherwise works really quite well in January for a couple of our brand since we rolled out and implemented it so it's definitely worth testing. So you see that working better in what peak periods for the yeah because you set a CPA CPA a CPA bid and metal only spend if it's if it can achieve that target for you. I only bid and the auction if it believes it can reach that target. And therefore I as demand in market increases and that's it that has an element of cyclicality evenings weekend pay days. You'll see that campaign spend but it'll often not spend anything outside of those periods I'll spend very marginal amounts so. Just just just seems to be working and seems to be a bit of a quick hack for a bit of extra volume so it's worth worth trying to roll out and some accounts yeah no watching is that like new customer. Acquisition yeah it would all be it's quite it's very proven to be quite incremental as long as you set up exclusions right I'd say the exclusions is a good topic actually that's one of the things that. I see when we're auditing accounts just people aren't intentional enough with exclusions matter has become so much better over the last 12 years of 12 years 12 months. At delivering and finding new customer but only if you set it up to do so so like make sure your event much qualities really high make sure you've got all your capy set up. Make sure you've got engaged customers defined as broad as possible. So Instagram Facebook website make sure you've got the same with customer list unpurchase all the data flowing in because. As soon as you start to let existing customers leak into campaigns like it's an easy way for Facebook to hit a CPA target so just spend a lot of budget towards that yeah. So we've seen that be a big thing that we people need to tighten up so yeah but if you set that up correctly on big cap I thought we're also seeing cost caps. In accounts alongside a car a C cell be quite incremental and non competitive as well like we're having both campaign setups run alongside each other haven't been too kind of realizing so I think that's also worth testing if you can if you've got the volume. I was going to say one more thing that just came to my mind. It will come back it will come back and I guess just why you're what you're thinking on that is good opportunity maybe to just talk. Off the back of what you were saying about personas just if we're taking it back to the actual creative strategy part of like how I'm just maybe thinking slightly differently about persona mapping and taxonomy yeah and because I think especially when you're driving like real scale one thing I. Mistake I see with brands that are probably spending like north of 500 care months a and still like isn't hugely significant at that level but I would say that that's where we need to start thinking about. Lairing personas in different ways so we're using like this I'm starting to think through the lens of persona strategy in like this three layered taxonomy which is the macro persona which is the. The who so the type of person or the state that we're speaking to so that's where you're like more standard persona table will come into that's that's. Helistically sits across multiple micro personas which is the middle there which is the why so this only exists conceptually essentially it's the dominant motivator of attention inside each of those personas so an example of that is like you have your macro persona let's say it's like a protein. Supplement so one of them is the you know one of the personas could be the the protein Jim go that's and it's the male protein. Jim go that's the macro persona but then the microbes would be going into slightly different layers so for example one of them might be the taste really so you're breaking that into primary desire so enjoyment without compromise the core fear. Protein usually taste grim and then the belief which is and protein powder usually tastes like crap as an example and but then another micro persona might be like the identity keeper so they the primary desires the core fear and the belief are different so even though they all feed into the same macro persona. It's about how how do we split off the different belief systems core fears and desires for those those almost subsets of that macro persona so we can. When we are labeling our creative by that we can speak to it slightly differently and as a result it's going to help with incrementality because we are speaking more specifically to that persona in different ways versus having like a blanket way of speaking through that messaging. So that's just something that I'm starting to think more through the lens of when we're when we're driving more significant scale. That makes a lot of sense I think intentional persona mapping is definitely is it's going to be really big over this this year I think we were obviously rolling it into our process through that work how are you thinking about prioritization across persona. So I think that's a really important part of prioritization really important part of any strategy but especially creative strategy when we're building a building it for a new brand that's maybe coming on board and we might have and we've got four personas how are you. How are you prioritizing across those is it just like data points volume of data across the research is yeah it's a mixture really so we obviously have break our research into. Market audience. Compressed an organic yeah and off the back of that we we then that funnels into strategic opportunities found within each of those sections so almost filters out all the crap and then build that into prioritization so that is a off the back of that we will build our persona. Personas and then all but all of those strategic opportunities then we can use that as a data source for AI and and that by the way feeds into customer reviews it feeds into you know market analysis et cetera et cetera all of the we we use AI to like scrape red it as well that will be specific to different points personas and motivators and and AI does a really good job now of being able to consolidate which which are the priorities and which aren't I mean it's absolute simplest form if you weren't doing all of that you could literally if you've got significant data in terms of customer reviews like we we ran this for heights and they had thousands of reviews so we're able to like literally take those reviews and it would give it and then cross compare based on the personas that we gave it and it was actually able to just say it essentially labeled and the volume of reviews that related to each of the personas so even through that and it's simplest form you can get to a good level of prioritization I think it's that but then there's also as you said from a media by perspective like a level of intuition involved as well. Yeah yeah from the context that we've experienced and from the brand side exactly made sense and there's two interesting new sources of customer research that I was chatting to a couple of people about that maybe a quite quick wins well ones are relatively quick win for brands that aren't already doing it ones a little bit more involved we've touched on it briefly in previous episodes but I think both it could provide a new way to uncover a lot of insights I was speaking to someone the day who had implemented a post purchase survey on the check on the checkout completed page that was like an open ended question with popped up. Immediately as you landed and just asked them what triggered you to purchase today and they'd used that to validate a lot they'd use that to just dig into well the triggers firstly but it uncovered a lot a lot of insights. Is there like an incentive attack? No it's just just I actually didn't ask that question thinking about it I think you probably could but I think it's because it's so like direct and so quick after purchase I imagine you just get quite a decent completion right at scale yeah yeah they they'd use that to rebuild an idea of what's the reason why someone to just just change a perception on the trigger of the purchase like they had an idea about the person who bought their products and it's quite high price point it was often an upgrade but the trigger was often experiencing specific faults or problem with the. Like the alternative cheaper products that was coming up so regularly that they were then we're able to kind of call out those really niche nuanced examples rather than just the bigger overall problem was just a really interesting like simple quick hack to get new fresh data points I think triggers are quite important the quite like it's a bit different to like pain point yeah you can have a pain point but then there's something that almost triggers it to be. It's greater than the cost of the product that you're selling if in some ways from like an impulse perspective. So I think triggers are quite intentional word I think you probably get quite a lot of learnings there if you if you're implemented that most runs yeah so true I think yeah we don't we don't speak about triggers enough but as you say it's like there's always some form of trigger whether that's emotional whether whether that is based off of some form of action failed solution like and capturing that in the moment is really valuable because as you say like a lot of the time that's where. You know the the longer that you extend that feedback window like the less accurate I guess that yeah it's because it's like yeah it's like something's driven to that website in that moment to buy example being I was. procrastinating on by the water ball for like four months because I just kept by and plastic balls and gym and then I went one day in the unknown plastic water balls I bought a bought a wall ball that's a stupid like example top of mind. You know I mean it's like people other people will see something procrastinate around purchasing it and then something old trigger them to go and make that action that day and they got a lot of data points around that and it shifted the strategy quite a bit. The good worth doing second one is customer interviews. I actually saw I shared it in our chat today that 11 labs have released customer interviews it scale through AI which I think is a very interesting. Great idea really good idea yeah I mean it was expected to come because if you think about how it's being used as like a sales mechanism now for like even just sales calls and things like that yeah and we haven't tested this literally came out yesterday but. We've seen the impacts of getting customers on on the phone and we probably don't internally do it as much as we ideally should because it's quite time consuming it's. And we're moving at a lot of pace for a lot of clients and I think that opens the door to maybe being able to just script those out and just. Doing like and you could even then think about try to do it across different segments of customers you could do it across different personas if you can get that detailed and how you segment them in the back end of the business yeah. And you could tailor the script towards that so I think that's a really interesting. It's a great idea round of research that we're going to see come to the to the forefront yeah we'll definitely test that and we'll let everyone know how we get on with it. I remember what I was going to say earlier that I forgot and. Just back reverse back into the account like one thing that we. Big piece of advice like even if you don't turn off winning ads like I just don't believe that you should ever turn off an ad that has been a significant winner. And that even even in times when CPA may start to go above target we've had a couple of cases recently where we've seen that this behavior and accounts and. You come in without context and you turn off those winning performers because they are above CPA because we're going for a period of. Where we need to do work we need to improve the performance the account as a whole and it can be tempting without that context just to turn the ad off because it's it objectively looks poor but a lot of the time they can be supporting the whole. Add accounts and it's it's always a view of let's beat that let's succeed that let's we may need to take a spend off whilst we do that to make sure CPA comes comes down and we have that under control. And I just don't believe in turning off significant winning ads it's always a bit beat them beat them process and. Back to so any any more bits on persona so I know I think we will definitely do like a full episode going much deeper into this it's just some stuff that I'm working on at the moment so we can do a full full full episode break down of that so we should we get on to AI yeah let's get on to AI so yeah you want to kick us off here with I know you've been spending a lot of the Christmas break and just experimenting playing what have you been doing. Well I feel like we've gone through as I said another acceleration of just like narrative around a eyes pace of progression and I'm not very technical but over like December. Mid-December late December there was a lot of narrative on Twitter from. Big influencers the cloud team themselves around the the quality of output people were getting from cloud code and how. It almost reaches this inflection point where the is so good that you just do not need any technical background to build really significant pieces of software so. Just went out of rabbit hole sorry I saw a tweet actually that can reference that I think really crystallizes like what we're going to experience over the next 12 months and it was it said. Those people who experiment with an a curious about the front frontier models of AI are going to start to enter a fundamentally different reality to those who don't use them because they're going to be able to produce seven. Seven days worth of work in one day and then it's going to be 30 days work in one day and then that's like the that's the sort of cycle we're in I've also been reading strike presses. Oral history of AI scaling and like how that's been going which is quite I recommend people going and kind of read some of the fundamentals about how the models work how how scaling. Model training and like the what what that means and some of the security stuff because that's mental like some of the. You don't you don't have any idea how like a model actually work yeah so that's crazy which is which is bonkers and then that leads to like a lot of. Questions around security and yeah anyway that's very philosophical at that word and yeah I thought that was just and so that led me to just doing more more research and just spending more and more time trying to just interface and build ways of working with with the tools and. Just think called right now is just so so good like what they're putting out feature wise see them is almost like the apple of the I like I enjoy using the tool way more than the others even though maybe Gemini's better at. Certain things that they've got it's got like a taste aspect to it which I think is really enjoyable I agree the UX for me is is is definitely the most enjoyable and. I feel like they also are doing that whole like Steve jobs talks about that intersection of humanity's in technology that's like one of his big principles and how you humanize tech and I feel like the way that they're talking about skills co worker. Rather than gem projects it's like is really it really humanizes the experience and makes it really relatable and makes it really easy to understand which I think makes you want to go and explore it more exactly it's that whole thing of jobs as well simplification yeah and I agree like especially with called skills like. You know you can you can technically do all the same stuff across a lot of these different and models but the way how accessible Claude make Claude skills and how you can literally have a conversation within it will direct you in the right yeah in the right way to upload the right context the right point and then help you actually layer. It's logic flow is is like just just so beautifully done and it makes it so much easier yeah and I think the context window for Claude is is just so so much great so much so strong relative to others so. I think one thing that a lot of people on our team and probably in your businesses is like you want to centralize everyone to a plan because and then you need to make sure everyone's got the connectors in their setup on like max so notion for us. Google Google calendar all of these like foundational things I think it's easy to think that everyone is doing that but often they aren't. And you can go and it's retrieval system is amazing so if you're working on a problem and you and you ask it to go and reference anything that it thinks is relevant to that problem in notion it will go and retrieve those files and download into its context so. You need to do that foundational piece but then yeah I think it's been been experimenting with all of the different areas I'll let you talk about skills I think I'm a massive fan of deep research still I used to start to use deep research a lot more for. Just most problems because you got one you can also like have about seven running and just disappear and it'll come back and it'll just be a loads of information it doesn't always need to be on them. On like massive problems of research in its in its truest form it can often just be like anything really I don't I wouldn't I wouldn't think of it as just a research agent I just think of it as something that. He's able to spin up loads of sequential agents and and come back with a lot more depth around the topic and I'll put a we're putting the in the links below a guide really in depth guide on how to leverage deep research to its maximum which includes like how to prompt properly. How to get context in but one of the big things that I loved about deep research is it always asks for clarifying questions. And you notice it does it there but then I was like well that's you should pull that thought process into every track so at the end of every promise I just put ask me as many clarifying questions you need to ask until you are confident you have absolute max context on the problem. And that that was often spit out like 16 to 18 questions and then I'll have to whisper flow or type them through but the quality of output is so much greater than when you don't and I think for many. I've kind of the prompts engineering like a really big topic but I think prompts engineering is often about minimizing differentiation of output that minimizing variance of output in a system so if you're putting in a prompts and you need the output to be consistent every time. Prompt engineering like narrows variance but you're tweaking prompts I think when you're doing a lot of qualitative work it's more you better focus on context I actually think the prompt has very little impact I don't even like the phrase prompt engineering I think it should be context engineering is gathering the right context to be able to create a better output yeah I think I think the actual prompt obviously I'm not saying it doesn't have the best practices the guy. We'll share in the comments has a lot of those but you better spend it you bet eight out of ten context three out of ten prompt is going to be the opposite every day and I think you only need to really obsess around prompt if you need if it's like data if it's like you really need certainty that every time you run that like a question yeah but I think the prompt is the vehicle for context yeah yeah and yeah I fully agree I think deep research is is is gray I've been playing a lot with Claude skills which as I said is really like in in any it's really true the point you mentioned just about like curiosity and like how the the gap it's going to create with those that are like ten steps ahead versus not and I think one of those is that I see just create creating a huge gap in inefficiency and also quality of output is is in a vehicle to get there is Claude skills and I think there's there's so much that can be done utilizing this tool and thinking through the lens of like what are the skills required within your business to get to to what your output is and how that we isolate those different skills you know for example creative strategy is two broad but then if we distill that down to script writing there's there's so many opportunities that we can build there so I we built a Claude skill around script writing and we have this logic flow where essentially the it starts with the the skill the skill logic which essentially includes core script writing principles so it's more thinking universally around script writing funnel position logic format detection and routing logic so essentially it knows through that prompt it can detect OK what is the right format to push us into and why so it's almost like a it becomes like this skill that becomes a master for and for almost prompting other skills within it so we have a format specific skills set we have a funnel position skill set and it's able to retrieve the right level of context of each of these to be able to get the right output and then once you want to once we get through the the initial skill layer we then go into smart framework selection so it's able to actually pull out once it has that macro understanding of of the actual skill itself being script writing it can then give us more specific context on the areas that we are discussing so for example if we're creating a high production ad it's able to pull the right context from high production theory and also high production best performers and this is where what you mentioned there about like the connectors which are great with Claude so because we've got notion connected and we've got a whole database of best performing frameworks that are split between higher productivity and static. Claude is able to then understand when to go into that best performing framework database pull the context of what is working right now with high production and use that as leverage to drive a better output of the script that it's going to it's about to create nice and yeah you can start to stack these can you you can start to have it so it pulls in multiple skills throughout workflow exact and agent network but but slightly different. So think it becomes just again leveraging that context that unfair advantage which is so powerful yeah and it's thinking of it in layers so you know you have your score your core script writing DNA which is the same across everything that you're doing so thinking about when you're building skills what that means from from a layer perspective and then how that feeds into like specific pieces of context and that kind of may just shows like it's why it's so important that every every skill that you have as an individual and you have within the business like the same as like S O P's like we document a process for everything that we do but I think we're in a world now where we need to be documenting every single skill set and subset of those skills as well. 100% agree I will using it for non acts like tasks that I'm just execution either so presentations management approach like you can codify a business approach to management in a skill and that can become a soundboard for for difficult conversations or managing people. I think you can do it for general general communication. There's so many different ways to ladder this up. There's a lot is a huge skills library on. Anthropic which you can which you can pull a lot of pre built skills from and I think that will only expand but. I agree it's just such a such a powerful tool yeah need to be leveraging the one thing I would just to to cap this off from a script writing Claude skill perspective like the context documents that you will want to build is a master script writing principles context document now a little hack and if you remember my story telling in scripting mastermind that I did on dc diaries and take that take the transcripts and create a context of the back of that because a lot of the inputs that we use in ours are from that video and the other that then you just need specific to the different formats right so. Take take all the transcripts of your best performing high product you do see ads and static ads from the last 90 days and put them into individual documents and do the same with a hook theory principles document and then really it has everything it needs to looking at script writing more holistically and then from a context specific point of view. The only addition that we have is like a creative strategy one on one context document that again is almost like the layer up from the script writing principles document and which is more just about like fundamental creative strategy principles that then feed into script writing as an element of that nice very comprehensive. That's like three years of work right into into the context so it is but I think if you're listening to this you don't have that context yourself within the business like go to the dara denies of the world. There's so much content out there that you can literally just go to YouTube show transcript download transcript put put them all into a chatbot and ask them to consolidate that information and put it into a context document like that in its simplest form will still do a great job. 100% that's how that built his he built a creative strategy agent he said it just downloaded all of your all of our episodes and I've heard it into into a bot and one of the thing on like prompting that I've found to one thing I'm prompting that I have found to deliver like a lot better results if people aren't doing this already is like chain of thought and asking the AI to like go through step by step is it's reasoning and in its process of getting to a decision. I think it often allows especially when you're doing probably like data or problem solving bit itself correct. If it finds an issue when you watch it's watch it's thought process and which can minimum which can lower the error error rate or hallucination rate which is interesting. Yeah obviously got that like Claude from that perspective and you've got Claude code and like vibe coding and Claude released co work which is like the more user friendly version of Claude code at this time just a couple of days before we recorded this. Which they said they built solely using Claude code so they didn't know engineer touch there they just vibe code the full thing well. She's just like the cost of software is trending to rapidly towards zero and the ability to I built a tick tock shop forecasting tool in probably about four hours and it was fully usable and it was that process was taking us. Our tick tock shop lead is probably taking him 45 minutes per forecast to build the model and now it's just input really nice UX UI output with an export function. It's quite exciting because that opens the door for like a lot of like niche micro tools that can have a lot of impact on business or even personal life I was speaking to someone the other day who are built who is building like their own operating system for themselves. On their phone as an app and they're just doing it off through Claude code hosting it on the phone and then that's going to plug into like it's going to handle a lot of like to do list communication. Like house like all those all those shitty little admin things that you don't really have somewhere to write them down to thought was really cool. But yeah, it's just it's again, it's it's a new it's it's it's very fun. It's it's like very very dope and it's like I was doing if that's the word but. Just to build shit like in a way that it's possibly do before because it just like it completely removes barrier to entry doesn't it so like you can really stretch your legs through creativity that you never thought was possible. Yeah, you need everything 100%. And especially in that that field of like tech and apps and web apps. So I think I think software enabled service providers is where we're heading that's definitely where we're heading it's like how do we. Build a software that makes our team way more efficient and provides the platform for innovation and growth for more businesses that we can work with. It makes you think doesn't it like is there a world where SaaS becomes something that is way more customizable so you have like a core SaaS product that solves a problem. But everyone wants to you know like there's there's I mean it is simplest form like there's there's so many journals and planners that I use that set templates but actually there's like a there's a very specific way I would like that to be. Formatted and same from a SaaS perspective I'm sure there's like another layer from we have the core product that solves a problem but then through no code we're able to actually customize that to our own. Desire yeah that's what I think that's that's kind of where I think you'll go because you look at HubSpot take HubSpot as an example massive CRM platform CRM for every business is quite different especially based on category. But there's common strands that run between that so you can see in a world where software cost is 10 into our zero you're able to produce. Very complex or relatively complex systems without that much technical knowledge. If they don't they need to focus on providing fundamental best in class building blocks because otherwise they'll get someone who comes in and makes the ecom agency. CRM that's perfect for that use case or the business is just going to build it themselves notions a good example of those a product that. Yeah the building blocks as well I guess yeah notions good. But yeah I think I'll be interested to see where that cuts industry category goes really over the next 12 months but certainly. You know another four weeks of rapid progression anything else or should we wrap it up there think we should wrap it there and yes. If you if you enjoyed this make sure you give us a like comment and subscribe we we always forget to do that sooner and the. We did an intro but yeah no but it was good because it was an authentic and yeah we were just chatting yeah true like like friends we do like each other yeah yeah yeah. But yeah we see you on the next episode.