AI For CPG Businesses

Jordan Buckner: [00:00:00] Building a CPG brand is fraught with challenges, especially in today's market. There are a lot of industry pressures from both the consumer side and the supply chain industry side, which is making it really difficult for brands to succeed. That said, there's a lot of areas of opportunities and ways that you can, one, better understand how the industry works to succeed, and then also using tools like AI to better understand your customers and run a more effective and efficient business.

So I want to really jump into, for this conversation, what AI is. And the, the fact it can have on you as a CPG business, as a way of thinking about ideas of running a more successful company and have that chance to, to actually grow. So for this conversation, I invited on Andreas Duess, who is the CEO of 6 Seeds, a company that deals with AI and data for CPG brands. Andreas welcome. 

Andreas Duess: Hey, thanks for having [00:01:00] me. It's a pleasure to be here.

Jordan Buckner: So I know we've known each other for a while now and a couple of different perspectives on the industry, but I would love for you to tell our audience a little bit about yourself and then we'll jump into this conversation on AI.

Andreas Duess: Yeah, sure. I'll give you a little background. So I I've been in, I grew up on a farm, really, really interested in food for a long time. And then I lived in London in UK where I went to university. And while I was there, I worked as a line cook. I made some money working in commercial kitchens, really loved cooking, but you know, didn't Think that was a long term career.

And then I ended up working with early technology companies while I was there. I worked with early Google, early Facebook in the UK, Sony, those kinds of things. And got exposed to also to early AI at the time of the company called autonomy was the first one of the first programs where a computer could get sense from context and written document.

Really, really clever stuff from there. I moved to. Canada to Toronto and there I sort of [00:02:00] accidentally started one of the largest the most successful independent marketing agencies specializing on food and drink. I grew from a freelance job for a food company and I. Kind of figured everybody has to eat.

So that sounds like a great market to be in. It was a market that I feel passionate about myself and we just grew the business. And then about five years ago, when we were at about 42 people, offices in Toronto, offices in Montreal, a friend of mine pinged me, texted me, messaged me. And she said, I want to show you something, but you can't talk about it.

And I said, sure, what's it going to be? And it was an early large language model. And that large language model was two things. It was absolutely terrible. And it was very clearly the future. And I'm old enough to remember when digital took over from film and you shoot commercials and the same thing happened there.

It was awful. It was terrible. Everybody laughed about it. Then all of a sudden it wasn't terrible anymore. And then six months later, everybody was shooting digital. Like the, the switch job was [00:03:00] rapid. And so I went home. And immediately had a panic attack about the future of my business, you know, I was thinking, oh my God, a lot of the things I can see a lot of the things that we charge money for a computer will be able to.

Do in the not too distant future. And so I spent about a year worrying about this and trying to figure out ways, how to implement this into the business as technology, when it was finally there, really dove deep into it. And then somebody offered to buy my company for me or buy my shares in the company because I owned 50 percent of it at that time, somebody made me a very good offer for my shares in the company.

And. I decided to take that offer now. It wasn't quite enough to sail into the sunset, but it was enough for me to take a step back and thinking, you know, which problem can this technology really, really solve? So one of the problems that we always knew existed, but. We have never been able to move the needle on was the 82 percent failure rate in CPG and you know, the number fluctuates between like [00:04:00] low seventies and the high eighties, depending on whether you are a big international brand with all the resources behind you or whether you are starting a new business where , you know, bring your grandmother's pierogies to market But be that as may, the failure rate in CPG, as you know, is absolutely astronomical.

And so we kind of figured out, we kind of said, well, Is there a way where we can use data and AI together to move that marker? And so what we did, we went out and we talked to partners and we pulled together an awful lot of data. So we're pulling data from Retail, so we have access to retail data that's not Nielsen data, but it's coming from credit card companies.

It's coming from the Instacards of this world. And then there's data from food service. So I think there's about four and a half million food service data points now in the database. That's constantly being updated. So we know what are people eating in restaurants? How often do they eat it? When they eat it?

Home consumption data. If you have a smart fridge, do not be surprised if the smart fridge company shares the [00:05:00] data of what's in your smart fridge, anonymized and all that kind of stuff. And then finally, the whole data is compared and cross referenced against the public conversation and this immediately, so this is about a trillion data points and we have this for the US, Canada, Mexico, Brazil, most European countries, India and Australia has just come online as well.

And what this allows us to do, it allows us to close the say do gap, because when you talk to people about food, when you ask people what's important to you in food, people will tell you a beautiful story about how they care about organic food or natural food and all these kinds of things. And then when you look at the numbers, when you look at the drivers in food and drink, there are three that stand out.

And number one is affordability. Number two is availability. And number three is, does it taste good? Those are the three main drivers and so there is this massive gap between what people say and what they do and so the first thing we did with our data we figured out exactly what they do and because we have an AI layer sitting on top of this data and because the data that [00:06:00] comes in is so. good We can now tell you what are the flavor preferences for bread for a 35 year old woman living in Austin, Texas, compared to a 60 year old man living in New York City. We can really drill into the data and pull specific items out and that allows us to work with our clients to help them create products that people want to buy because we have that observed data and be to create messaging that really resonates with people.

70 percent is Nielsen data from last year, 70 percent of all marketing campaigns and CPG cost more than they earn. So. Seven out of 10 times when you have a brand spending money on marketing, they're spending more than that campaign actually brings back into the coffers. 

Jordan Buckner: I love the look at like how you're able to capture all these different data points and then help brands make decisions based on them.

And I think a lot of people, when I talk to founders about. They're trying to figure out this to a couple of different layers. One, like, how is it going to affect the [00:07:00] industry overall? And then at the micro layer, how can I use some of these mostly conversational AI tools to run my business? And there's a couple of different layers in between.

And so I'd love to dive into three topic areas. One is kind of understanding. The industry structure and like how things are done in CBG and then how brands can better understand that to have a better chance of success to is what you're talking about a little bit is how I can help you effectively can deal and manage the data that you have.

And then 3. What are brands doing and companies doing to use AI on an operational level while at the same time, understanding that it can't replace, you know, a lot of the emotional humanity elements of a company. 

Andreas Duess: Yeah, , perhaps we start with this. Let me start somewhere else. Right now we're at a stage with AI where a lot of the conversations are about AI.

But what's going to happen in the next five years [00:08:00] is that AI will become the equivalent of electricity. You know, when we're having a conversation here, Why a fiber networks and laptops and webcams and microphones it's electricity that's powering out of this but neither of us is excited about. The name and the technology of our power provider.

At least I'm not right. It's just something that is there. And so the same thing is going to happen. In my opinion, it's gonna happen to a I that a I will be a thing that is there and it's allowing us to do certain things better. Like this is a lot better than us having this conversation with a written letter and then you publishing an article about it in six months from now.

Right? So that's where I is going. And as a result of this there is a, let me go back 300 years to explain this. There is a thing called, I think it's Javens, Javens Paradox. Don't write angry emails if that name is slightly wrong. But this happened when steam engines first appeared. So when steam engines first appeared, they were so inefficient that the only place you could [00:09:00] run a steam engine was close to a coal mine, because it needed such enormous amounts of coal.

And then James Watt, famous engineer, appeared and he made steam engines more efficient. So within three years, it took only about 20 percent of the coal you needed three years ago to get the same power out of a steam engine. \ And two things happened. Shares in coal mines completely collapsed because people thought that you needed less coal. Then there was less value in coal mining. But the exact opposite happened. Because all of a sudden it was possible to run steam engines away from coal mines, away from areas where there was naturally coal.

People put steam engines into Absolutely everything in the most ridiculous places where they made no sense whatsoever. But they also build a train network and they also put steam engines in agriculture where it mechanized agriculture and they built factories and they built power stations and the industrial revolution gained power on the back of this.

That's where AI is right now. Right now. [00:10:00] We're at the stage of AI where Steam was when you could only use a steam engine really close to a coal mine because it was expensive to run, right? But we've just seen the Chinese model coming in, the R1 model that is far more efficient than anything else we've seen before.

And that is the James Watt moment where all of a sudden I can open my phone and I have an LLM running on my phone. Right, so that's where we are. So this is what's going to happen in the future. AI will be embedded in most things and we won't talk about it anymore because we're not interested in the AI anymore.

We're interested in what the AI helps us to do. And that gets me to your question, Jordan. There are two things that are really important in AI. That you need to remember. Two words. One is called meaning making, and the other one is called sense making. And meaning making is, I'm going to go out and I'm going to celebrate my grandmother's heritage by bringing her secret pierogi recipe to market.

That's born out of my [00:11:00] love for my grandmother, respect for her heritage, all of these things. Or, I want to make sure that the gluten free cookies I bake for my kids are available for other kids. It's all about humanity. And you and I, we have brought products to market that were based on that kind of thinking, right?

Then, you use AI to help you make sense out of that meaning. Talk to me. About the pierogi market. Tell me how many pierogies are being sold in the US each year. Tell me what people are eating in 2025 and which of these ingredients can I use in a modern take on my grandmother's pierogi. And what I mean by this is what's AI.

Is that we as the human being set the intention and then we go to the eye and say, help me get there when things go wrong. And, you know, as somebody who creates content yourself, I suspect, you know, this from firsthand experience, when things [00:12:00] go wrong is when you try and outsource when you say, do this for me and where things typically do a lot better is when you say, do this with me.

And that is one of the biggest and most important differences being that that you, that people need to keep in mind. 

Jordan Buckner: No, I like that. And I think that gets into a couple specific use cases as well. So one thing is you're mentioning is almost like product development, but I'm like the concept development side.

So you have that initial idea. Help me test. Is this a good idea? How to make it better? What's the potential market sizes? It currently is based on historical data and purchase behavior. Are more people purchasing pierogies now than they did before? Or is it downhill trend? Right? And not that any 1 of those prevents you from moving forward with the business, but gives you a lot more insight on how you need to approach the market.

And differentiate yourself so i think that's really interesting what other use cases are you seeing companies using a for every every day 

Andreas Duess: [00:13:00] yeah so i mean one of the things that we can't leave projects for currently involved in the rest company. And I'm sure you, sure a lot of your listeners are familiar with Ozempic, GLP 1 drugs that are appetite suppressants, basically.

So, there is, the data that's currently out there basically says that 30 percent of Americans, US citizens, will be or will have been on Ozempic. in about two years from now. It's a massive number. Walmart published data where they're saying people who are Ozempic patients spend 30 percent less when they come and buy food at Walmart.

Now that is a significant reduction in spending. So one of the things we are currently doing, we're working with this rice company and we're figuring out what is important to Ozempic patients. What are they looking for as far as nutrition is concerned? What are they looking for as far as flavors are concerned?

What are they looking for as far as experiences are concerned? And then how does this fit in with a high protein, high fiber diet that is [00:14:00] recommended by medical professionals when you're on Ozempic and. How can we create a rice product that actually fulfills these needs, right? That is still easy and fast to prepare, but it fulfills the special needs of this group.

And for there, AI is incredibly useful because it allows us to dive deep into all of these data points and pull those relevant answers out. 

Jordan Buckner: I mean, I know you work for a project, but I'm curious, role that AI plays in kind of solving a problem like that beyond what a team of skilled professionals can do is, is it kind of.

Accessing more data quickly and being able to pull those insights out faster or being able to take like a more comprehensive look at it instead of just a limited look. Where's that differentiator for 

Andreas Duess: all of the above right and like none of the things that I'm saying and none of the examples here will in any way diminish. The expertise and the inside of a [00:15:00] skilled professional. This goes back to meaning making and sense making what it helps the professional to do is work with a larger data set is work quicker is being able to say, well, Let's find out if the needs of nozempic patients in Louisiana are the same of those of nozempic patients in upper New York state, right?

It allows us to go into the detail of the data a lot more and then figure out how to bring the right product to market, you know, whether, for example, for this is a genuine conversation we're having with our clients right now and. Currently, the decision is to bring different markets for the southern US and for the northern US because the conversation and the personal needs of the consumer are markedly different, whereas the nutritional elements are exactly the same, so the product will be the same, but the packaging will be very different.

The branding will be very different. So those are the things you can do, and then that allows you to figure out exactly How to communicate and how to bring this product to market. And [00:16:00] again, that solves the 82 percent failure rate problem. 

Jordan Buckner: Interesting I like that. You know, I am a big proponent of AI tools.

I use chat GPT pretty much every day to answer questions at this stage is both like personal and professional for the business. One thing I find is really effective at. Are fine being able to ask questions based on your own personal information in a way that Google allows you to do kind of for the Internet, but even more stream right and saying, Hey, I have all this information.

That's too much to process or will Take a lot of time to look through, how can you help me surface this information and insights about it, but also the next level is how can you then help me decisions based on that, whether things to consider, how are the best ways to structure plans around this, as you mentioned, how do you work, use the ad to work with it, ask questions, get answers, responses I really like the, you know, using it conversationally to get at, Answers that you otherwise may not have been able [00:17:00] to get to.

Andreas Duess: Yeah, 100%. So there's a number of different things. So, you know, there are there are sort of different tools here that we're talking about. And you have specialist tools, like, for example, nutritionist database and that nutrition database contains 15, 000 ingredients, and it now allows a product developer to explore new combinations and then work with them.

And then you have generalist tools like large language models, like the chat GPT's and the clothes and the Google Gemini's of these worlds, which have been trained on absolutely everything. They could possibly be trained on and for those as a user to make them successful, we need to set guardrails.

And that's exactly what you said right here is my data work with that data. And get me the best possible information out of it. So one of the things, for example, that in my own company, from a behavior point of view has been changed is we document absolutely everything. And that includes that we are recording as many internal conversations as we can possibly record.

We record every [00:18:00] single client conversation. We have that transcripted and it goes into our database that we have the ability to. work with that data and make it useful for us. So for example, our process very often for putting proposals together nowadays is to have a long in depth conversation with the client.

That conversation gets recorded and we have some custom tools that we programmed where we then upload that conversation and the custom tool. Pulls out the client's needs, lists the core points that need to be addressed, maps them against our capabilities and writes a proposal and it's done in five minutes, you know, rather than in two days.

So those are the benefits of data collection cannot be overestimated in your own company, you know, serve data. Collect the stuff, record everything you do, put your data somewhere where AI has access to it. And it doesn't really matter which tool it is. You can also have projects and chat GPT, boulders where you can put stuff in.

[00:19:00] We run our business, or I run all my businesses on a software called Notion, where you can put all your content in and then Notion AI has access to it and can manipulate it and work with it. But that is not to be. Underestimated the importance of this, as I said, you have specialist software does specialist AI that does one thing or a couple of things.

And then you have, I think, for most startup founders, for most people who are just launching a business, they don't have like, you know, when we at succeeds, when we work with a client, we started 50, 000. And the majority of that money is data cost, because the data that we have is so detailed, it is so good, it is so in depth, that it costs money to purchase it.

That's not money a lot of startups have, but most startups can afford 20 bucks for a chat GPT pro subscription. And. There, what you can do [00:20:00] is work with ChatGPT to help you solve your problems in a way that makes sense to you. So, for example, I'll give you an example. There is a woman out there.

Rebecca Van Edwards or Jennifer Van Edwards, I always get this wrong. She is a, an expert in human behavior. she's been on YouTube, she had a bunch of TED Talks, and she's really great in getting you to do. What you, what she wants you to do in a non creepy way. So to give an example here, one of the things that I find difficult to do and that I absolutely hate is when I have to write follow up emails to people who don't necessarily want to talk to me.

But I want to talk to them because we met at some place and perhaps there's an opportunity for business and so I'm highly motivated to stay in touch with them. And There is always this creepy email opener just following up, or just checking in. And it immediately makes you sound like a rejected lover standing downstairs waving some flowers, hoping to get some attention.

And so I was writing one of those emails and I was [00:21:00] trying really, really hard not to be needy in my approach. And I thought, I wonder what, you know, when AdWords would do. And then so I fired up ChatGPT and I said, you know, this person on ChatGPT said, sure, she does the following things. These are things that are important to her.

And I said, okay. Look, here's my terrible email. What would she do with this? And chat GPT came back and said, well, first of all, that email sucks. And second, here is how it should sound like. And so to give away one of my email openers that has increased my response rate by 50 percent easily nowadays, instead of saying following up or just checking in, I say thing.

Jordan, the last time we talked, you said that XYZ was important to you. We can deliver this. Are you still interested in moving forward with this or whatever, right? Or you say something along the lines, Jordan, I was just talking to our mutual friend, Sachi, and was thinking about the things that you said to her.

Here is da da da da, and you go straight in. And just [00:22:00] small things like this, where you can set up ChatGPT as your business teacher, can make an absolutely massive, massive, massive difference.

Jordan Buckner: I love that. I'll have to use that as well, but I do like the idea of when you think like, I know someone has figured out better ways of doing this.

Maybe I can't read the entire book right now if there is a book or watch all their TED talks or webinars, but you can actually get what Is close to their advice because a lot of our public figures with the tool, like Chat GPT and apply that directly with your business. And like you said, just have it rewrite the email for you, but then learn from it so that you can do it better in the future.

I would really like that as a practical example. Andreas, I really appreciate this conversation. You know, one thing that I am always excited about is the future of the industry. And I think that a lot of businesses are starting to integrate tools like AI and like you said, it's just going to become like electricity.

They're just going to ask questions of their business is going to really be a game changer. One being able to run a more efficient way. [00:23:00] Linear efficient team achieve more with fewer people, but then also add people in the right roles who really love what they do, and I see it really as a multiplier instead of a subtraction for a lot of businesses, and I'm excited to be on that journey and to see you at the forefront of that journey.

I appreciate you being on for this conversation. We'll have to catch up as AI tools develop even more. But for now, where's the best place to learn about the work you're doing? Or any kind of closing thoughts , for brands who are interested in how this might help their business.

Andreas Duess: Thank you. Yeah, absolutely. If you're a small brand, like if you're a large brand, chances are you're already talking about this internally. If you are a small brand and you are worried about AI or you're not quite sure where to start, one of the things that I think is a really, really good, really good place to start is by having a conversation with AI.

You know, it's like we've been trained. Forever that using a computer, there is a wrong way and there is a right way. And [00:24:00] if you do it wrong, you get a bad result. The computer crashes or what have you. Something happens along those lines. But with AI, with a large language model, the same Skills that make you a strong manager also make you a strong user of AI and that is tell it what you wanted to do, give clear feedback and work together with it.

So really, that's if you want to be an AI expert, you don't have to learn about prompt engineering. You don't have to learn about prompting. You don't have to do anything. any of those things. If you want to figure out exactly what's going on, one of the best things you can possibly do, put your headphones in, launch chat GPT.

I believe the advanced voice mode is free for everybody. Launch that and start talking. To it about a problem you want to solve and have an a half hour conversation and you will be absolutely impressed and surprised by how much work you get done that way. 

Jordan Buckner: I absolutely love that. I've used that before and encourage our [00:25:00] listeners to do the same.

Andreas, thanks so much for being on today and chatting all things AI. 

Andreas Duess: Thank you for having me.