274. Can AI Create Packaging That Actually Sells?

0:01Welcome to Startup to Scale, a podcast by FoodBevy. I'm your host, Jordan Buckner. Join me as I talk to aspiring entrepreneurs, seasoned industry experts, and everyone in between as we unlock the keys to growing from startup to scale.

0:18Packaging and brand design is entering a completely new era. AI tools can now generate concepts in seconds, create endless variations, and speed up workflows that used to take weeks. But there's a huge difference between creating something that looks interesting and building out packaging that actually sells on shelf. Today, I'm joined by Mike Heplinger from Smashbrand to talk about the role kind of AI should play in packaging and brand design. And I really want to unpack where AI is genuinely useful and where it can fall short and get you into trouble. I talk with a lot of founders about like how to use these tools.

0:54And so I really want to break this down. Mike, welcome to the show again. Thanks for being back.

0:58Michael Keplinger:Yeah, Jordan, it's good to be here. Thanks for having me back again. I think I was looking up, we last spoke in November, and here we are in July. That's an eternity in the world of AI, isn't it? It is. It seems like everything has changed. Just to give a little context, can you just give a quick overview of the type of work that you do and SmashBrand, kind of how you think about these things? Yes, SmashBrand is a data-driven brand development agency for CPGs. We primarily work with fast-moving consumer goods, where consumers still make most of their decisions in split seconds at the shelf. And so my role at SmashBrand, as we do many of the things that traditional packaging design agencies do, but we are one of the few that has an integrated testing approach into there.

1:40Michael Keplinger:And so we're guiding all of the decision making from strategy to exploration of creative to final validation with consumer research along the way, consumer testing. And I lead our team on that side of the business. So tell me about how you're thinking about AI kind of more broadly as it turns to branding. Like, what are you excited about in the space? Well, I think, you know, in the space and you can abstract away to just in general, if you haven't been living in Iraq and using it, we've invented the most amazing tool ever to be created. And I do use the word tool very decisively because to me, it really is an extension of like human thought and taking away those pieces and being a tool, you just kind of put it into there.

2:22Michael Keplinger:And so as you translate that into how it affects specifically what brands are doing and putting together brand development and, of course, packaging design, there's many facets, of course. And on the creative generation side, we're really starting to dive in there and explore what you can do. That side of the creative side as a language model is a bit more novel and developing rather quickly. But I really think that there are a number of ways to speed up what you're doing. But more importantly, to actually explore a wider range of what we would call a kind of hypothesis of what's going to be successful and narrow down to focus your energy on things that are actually worthwhile, more than you could do before.

3:03Michael Keplinger:And that's probably the most, at a high level, the most exciting aspect of applying it to this industry. Yeah, I think that's totally fair. And then have you ran into any kind of problems as you're using some of these features or watchouts that you're like, I was excited about maybe seeing if AI could do X, Y, and Z, but feel like it falls short in these certain areas? Yeah, I'll start with another analogy because, you know, you use the AI tools and it is primarily a language model. And I'm working on something that I don't know a whole lot about. I'm like, oh, my God, this thing is so smart. It's like generating all this information.

3:39Michael Keplinger:And then you translate that to something that and I'm using it for something that I'm actually quite knowledgeable about. I'm an expert on. And it's so easy to see some of the gaps in its work. And I think everyone can relate to that because it translates to the creative process as well. And it's really easy to produce things that look really good. But there is really a huge difference between some concept that is very visually impactful, interesting, and something that consumers will actually choose over what they're already purchasing when they see it on the shelf. So that's probably one of the biggest things to kind of really stand out or look out for.

4:14Michael Keplinger:And there's also a fair amount of bias. Like these things are programmed to be helpful and honest and make you feel good about decisions. And so there's a bit of a bias that you have to be aware of that what comes out really needs some scrutiny. And I think those are probably two of the biggest things that I would call maybe gotchas and watchouts and using the tools. Yeah, I think that's key, especially as I've seen the tools develop where you talk about ChatGPT or CLAB is that even a year ago, there were some, call it more obvious hallucinations or like made up information where if you just look at it in site check, you're like, OK, I don't think that's actually right.

4:50I've kind of felt that over the past year that's gotten a lot less. But as you mentioned, there are more things like you're getting advice on very like basic or maybe a moderate level of understanding without the tool understanding like the nuance or how it applies in particular context or situation. I think it requires more scrutiny than to understand how you're using it to make decisions.

5:13Michael Keplinger:Yeah, totally agree. I want to kind of also talk through the different processes as anyone or agency is going to be going through in the brand and packaging design process and understanding where or not like AI might be helpful. So if we kind of break it down into, at least how I see it, like brand strategy section, maybe starting out, then going to creative design and then actually designing the like brand architecture and actual the visuals of packaging and all the other elements, logo, et cetera, of understanding like where AI like may or may not be helpful in that process and where it still helps to work with an expert who's been doing this throughout their career.

5:51And so maybe starting at that early stage of strategy, how do you typically in the team have approached strategy and are there ways that AI may or may not be helpful there?

6:02Michael Keplinger:Yes, I think there absolutely are. And I'll just kind of walk through. I mean, I know that there's a way that you do it and you go through it and founders are approaching this, too, because we basically have the same kind of path that we put all clients on. And everybody knows this. You've got to get grounded in the category and really understand what the competitive space looks like. And that is an area certainly that AI can do a fantastic job. It knows pretty much everything about the products that are out there. I think an area to watch out for this, if you're working on some kind of novel category, subcategorization, it's probably going to fall short on that.

6:33Michael Keplinger:If you really have kind of something that's well grounded in there, getting a baseline of what consumer, what's really driving purchase, like what consumers are buying and why. And from our work, too, it's like, what are those decision drivers that really matter? And within there, too, like every on a food product, everybody has to show up for taste. Right. But then you start to get narrow in there and you define what really is your special sauce and that recipe of your differentiation. It's going to really understand that. So getting grounded in that a lot of founders actually built their products and things because they actually know this and they're scratching their own itch.

7:07Michael Keplinger:But it actually is also kind of coming to the other side of that. And that tends to lend itself well into into kind of early ideas of a brief, like how would you actually go about creatively designing that and then moving through. And of course, like in our process, too, and we're being very thoughtful about integrating AI into our process because we have, you know, longstanding ways of doing things. And it's not like, hey, it's this magical sauce that came out of some prompt. But instead, it's like feeding feeding kind of the grounded in science and success that we've seen. But translating that into on the creative side, and I've already touched on that, is moving that towards of like of that creativity of bringing forward just exploration of ideas.

7:48Michael Keplinger:We find that the tools and the creative side of the tools do really well of sharing and showing and pointing at specific other competing products and then identifying going from that other research. Like, what is it that your product can do well that's aligned with what consumers are looking for and pushing forward some of that idea? And I think that even at that stage, if you're engaging with an agency and showing up with that kind of work, I think we spend a lot of time in the discovery phase. And, you know, it's calendar time. It's hours of our strategist time. And so some of those steps, I think, can realistically be accelerated with using some of those AI type tools, I think, on that earlier side.

8:27Yeah, and I think one area that I'm excited about there too is, right, like as an agent, you have this longstanding process and then it's shortened because you've done it before. But then can you document that process in a way or talk through it that a AI system or agent can then use that same lens and those perspectives to help write out a full kind of creative brief or understanding it based on the research or different ideas versus using a more like large language model approach where it's pulling in outside information.

8:57Michael Keplinger:Yeah, I think there's a couple critical levers. And any AI tool is the it's garbage in, garbage out. And so what you feed it can be dramatic difference in what you get back from it. And so those critical thought steps, those strategic steps of getting those things right. Now, when you're talking about consumer products, especially, we all know that the AI has been trained on everything on the Internet. And we also know that consumers, we know, and it's well-documented research, that consumers often, especially for lower dollar amount items, they can't articulate why they chose something. Which means that the rationale for why they're making decisions is also not very well-documented and also not very well-embedded in the large language models.

9:40Michael Keplinger:And so that is the piece where you've got to be critical and understand that and feed it. So rather than relying on the large language model to just know this, it's going to fail and it's going to produce kind of things that don't work, in my opinion, is the thoughtful part that you keep to the human side and really focus on is really understanding that and feeding that in. And then you get much better results out from that. Yeah, I think that's where I just even remember when I was working with my own brand and doing a creative brief for the agency. Right. Like they send me over this template asking around, like, how do we talk about our brand?

10:13How what what kind of personality does it have? The tone of voice. And as a early founder, it's like I don't have the language to fully describe it. So I'm just kind of like mashing some words like, oh, I think maybe this or maybe that. And I have seen some benefits of being able to like write a couple of different prompts to understand. And like that's those are the words that I didn't have in my head, but can now be expressed through something like the AI conversational tool. So I get at the meaning that I feel, but I don't know how to express so that I can communicate to the design team what the brand actually could mean or say, like, no, it's not that it's actually this.

10:50Yep, exactly.

10:51Michael Keplinger:And I thought of actually translating what I was saying into something a bit more tangible. So imagine you're working on a pizza product, and this kind of comes from some research that we did in the past. And you're using, just to kind of early concepts, you're using a large language model. Maybe it's Nano Banana or Chai Chibi Tea to generate some early concept ideas. And you're like, I want some imagery that makes this pizza taste delicious, amazing. And it's going to make something. It always makes something, right? But then you compare that with something that's grounded in some consumer research that you know that cheese and the texture of the crust are important to establish taste in the pizza category.

11:29Michael Keplinger:And that oftentimes, usually the cheese is represented by gooey, stretchy stuff. And the crust is harder to show. And you kind of have to kind of highlight it and then also have like a message call out. So this is insight that we learned. And then you feed that as a part of your very detailed brief. And then you get concepts that come out that are very actionable and that resonate and actually deliver on those decision drivers like we were talking about that lead to choice over competitors. Yeah, I love the good information in Eagle's better information out because a lot of times, too, you have to approach in a step-by-step way of figuring out the strategy, all those components of what actually makes for a tasty-looking pizza and feed that into the image to actually give you something better that you like.

12:14I love that. And then I've also seen some founders, right, saying like, oh, let's see, maybe let me like skip some of the mood boarding process and do it myself by just generating some images of like what the packaging or the product could look like. And I've often seen that some of the results might look like interesting on first look, but then there's just like fundamental flaws in either like the imagery that's using or like the bottle is actually one that's not really readily produced or like just really interesting things. But I'm curious about how you've seen kind of that transition, like once you have the strategy and a creative brief around the product, then actually doing some of the mood boarding, the visual creative design to understand like the direction of the product and have you experiment with like some of the AI features and it like work well and then maybe again where it's kind of falling short.

13:02Michael Keplinger:Yeah. So I think that's where one of the biggest values you can get out of it. And the fewer different designers that are working on a creative project, the more this is beneficial because we as individuals have kind of a we have a scope of creativity. Right. And those creative juices are here and they can. And so we find we almost always in the creative phase are bringing in more than one creative, one designer for that diversity that you can get. And so this kind of gets to the point of your question is what AI can really do is that early kind of look, especially using multiple tools, because it's kind of like different people doing it or even like very different prompting and alterations to the brief is that ideation.

13:43Michael Keplinger:It's like it actually creates like, ooh, that's interesting. And then and so the creative space, which, you know, what's the point of mood boards, right? It's like, hey, put the color. Here's some examples. It's because people have a really hard time, especially when you're presenting and you're a client and you're trying to say, ooh, I like this. I don't like this. A lot of I can't even tell you how much creative energy is spent on creating that just to know that 90 percent of it is going in the trash. Right. So this is from a creative perspective. This is the most valuable part of AI. Now, granted, everything I said before applies better inputs, better outputs.

14:18Michael Keplinger:But to just explore so quickly a creative space that resonates because it's, you know, it's not just what works, but you got to kind of like it. It's your brand. If it's ugly and you're like, no, unless it just like beats Coke, then you're like, maybe. But other than that, you're kind of like, no, I don't like that. Get rid of that. Don't show me that again. And so that's where it's just really invaluable to get that space going. And then even creatives to kind of just pick up from that and then actually apply their art and their kind of expertise to really transcend that into communicating about a product and a value proposition is invaluable.

14:53Mike, I think that's really special because I also know, like, do you design in a different sphere? But a lot of times as designers, you have to go down a creative rabbit hole that's not going to work out. But just to understand where the real tension is and where the real interest is, like, in something. And if you're manually creating that stuff, it's not that it's a loss. But as you say, like, 90 % you just won't use. You still learn from it. You got smarter, but it's never actually going to show up. And so I think that creative ideation stage is kind of really important. And then as you go from there, though, into more specificity, is that where you see that it really helps to have a trained human designer actually putting their creative touch on those initial concepts?

15:33Or kind of was that transition then into the actual packaging design process?

15:37Michael Keplinger:The short answer is yes. The maybe a little bit longer answer is that, and this is actually like we're seeing so much now. The conversation about AI is even coming up with clients. We're almost towards the end of our work. and they're like, hey, I made a concept because we test also, right? And they're like, hey, what a concept. Can you stick it in the test? And you can tell when they just kind of pulled it out of thin air and they kind of did their own thing because it tests very poorly. But sometimes they take the work that we've already done and we've already done one round of testing. So they actually have the proof points of what makes a good concept.

16:07Michael Keplinger:And then some of those concepts can really do well. But at the end of the day, if you subscribe to the idea that AI is a tool, therefore you are responsible for the work product. It's your work. whether you used AI or not. And so from a founder, if you're generating this idea and you're like, I made this. I used AI, but I made this. And this is good enough to go to market with. And I trust it. Then by all means. But if you're exploring, if you want someone else to do the work, which is why typically people go and hire an agency, then the AI doesn't replace that. That's my opinion. It can accelerate that.

16:47Michael Keplinger:And an acceleration to often, you know, I think that there's a whole nother discussion about where this leads as far as price and cost of things. The way that we're embracing AI is it actually it allows our people to focus more on the pieces that really the critical thought. And like I said, we're exploring more creative to get to what matters as opposed to just doing the same thing and making doing it twice as fast and just being more. So we're able to produce better work because of it. But certainly it does open those doors. But specifically, I think to answer your question is like, if you didn't trust yourself to do it on your own, I would be leery about trusting AI to replace the agency.

17:27And then I know a big part of what SmashBrand focuses on is the data side of how well a packaging might perform. And so I'm curious, either whether using it or just theorizing, like once you have a few packaging concepts, what are some ways that AI might allow you to test the feasibility of the packaging? Yeah.

17:48Michael Keplinger:So some of the most reliable aspects are basically because it's well studied is about how things and colors and shapes grab people's attention. And so AI can do a really good job of like knowing which kind of concept might grab more attention on the shelf, where my eyes might gaze, because you do have kind of a sequence of messaging hierarchy to deliver your message. And then it's a whole nother topic that I think is going to become highly debated about synthetic respondents. So synthetic consumers. And this all goes my whole point of view on this. I'll kind of start there is I do not believe that AI can produce all synthetic data and replace consumer intent.

18:27Michael Keplinger:And it all is. And the reason behind that goes from what I said earlier about we don't make in this decision making. We don't make like the most rational, perfect decisions like AI is the most rational, like computer algorithm. Right. Not only that, but its knowledge comes from human knowledge, from the Internet, and it's been seeded with that knowledge base. And it doesn't have an understanding of why we make decisions. And that's because so much of them have so much emotional nuance in it. And we just, it's like this system too, this gut feeling. And I don't think AI can do a real good job of that.

19:00Michael Keplinger:And I've actually piloted and tested some of the tools that are out there now from big research agencies. They get a lot of kind of fanfare. But unlike a lot of founders, we happen to have 10 years of testing data. So I just give them assets I've tested before. I know what consumers think. And it's just I'm not seeing results that are even remotely close and reliable. So our approach is to leverage AI because I can't rely on the large language model, but I can train it off real human data. And so early using AI to validate ideas and seeding it with some kind of consumer data and then leveraging that to learn, but then retaining the most important testing at the end, our purchase intent, which is really like, which of these will I choose in a given in like a buying situation is always going to be real target consumers for us.

19:47Michael Keplinger:But there's a lot of promise on those earlier stages because you're learning and iterating. You're not producing something that is going to be final yet. And there is value there. I love that. And especially, you know, you mentioned something around like your gut feeling is how a lot of consumers make decisions. And even as founders, as creatives, sometimes you have to go with that. And actually, there's another founder, David Hennemeyer Hansen. He runs a company called 37 Signals, which produces like Basecamp, a bunch of software. And he talks a lot about how the like human gut feeling is one of the best AI processing type computers because all this information you just absorb from the world just as a person.

20:25And you don't know the logic of why you feel a certain way, but your brain just kind of outputs like, OK, I actually feel this is a good decision or I want to buy this product without understanding all the full steps within. And it's just really interesting. I think a lot of consumers obviously like act that way and it's hard to replicate or really tap into how other people make decisions. They'll say one thing, but then actually do something else, as we've often found.

20:49Michael Keplinger:Yeah, it's not completely just like unexplainable. It's just very hard to explain. So like what we're working on, we brought in a PhD neuroscientist that really understands decision making specifically through consumer decisions. And so our bigger kind of plan and what we're working on is building our own proprietary model based on the data we already have and how consumers to kind of pat and then use AI to help guide towards that decision making, but then still seeding it always with some real consumers to kind of focus in. Because just looking at the data from ourselves, it is crazy how different the same person will just think differently based on the category.

21:31Michael Keplinger:like you're in there buying a tech product like a new router for your house and you're very focused on a rational thing like what are these specs that it has how does it do this and then you're the same day you stop at the convenience store because you're hungry and you want a snack and it's just like what flavor of the day right so it's so so so different and and you have to kind of unpack what's really at play in that particular case yeah i think a big part is you mentioned right what level of data and what amount of data can you get or can you need to make that prediction more accurate, right?