AI is rapidly changing how we develop, market, and sell food products. But what does this mean for the future of food and beverage innovation? On the Epicurean Unicorn podcast, Andreas Duess, 6 Seed’s CEO and Chief AI Officer, sat down with Dr. Zheijing Wang to explore how AI is shaping the industry—from reducing food waste to creating data-driven consumer panels that eliminate guesswork in product development.
Key Topics We Covered:
- Why 82% of CPG products fail—and how AI can help brands launch products people actually want.
- The rise of AI-powered consumer panels—how brands can now test flavors, packaging, and concepts with digital consumers trained on real-world data.
- AI’s impact on food waste and sustainability—how AI can identify new uses for food byproducts, turning waste into profitable ingredients.
- How AI enables smarter product innovation—like designing rice products tailored for Ozempic users based on evolving dietary trends.
- Why AI won’t replace human creativity—but will instead serve as a co-pilot for product developers, chefs, and marketers.
AI in food and beverage is not about replacing people—it’s about removing friction, accelerating decision-making, and helping brands create products that truly resonate with consumers. If you’re in the food industry, this conversation is one you won’t want to miss.
Transcript
Brayden:
Hello and welcome to Epicureing Unicorn. It’s a very special episode for everybody today, mostly because you don’t get too much of me and we have a special guest host. We were able to have Andreas Durst visit with us, although it wasn’t really us, it was really a former guest of the show, now turned special co-host, Dr. Zheijing Wang. So very excited to have her come back. We did this because of her passion for AI and the company that Andreas has founded, which you’ll hear more of in that interview.
So let’s not have me go on anymore. Everybody, please follow me. We’re going to go down to Club Unicorn. We’re going to get this party started. So we’re going to come down the stairs through the door. That’s right. Watch your head as always. I see today we have a lot of AI-generated pictures of interesting foods. So enjoy. Enjoy with your eyes. Maybe not necessarily eating it today, unless you’re into paper. You know what? If you think it’s healthy and nutritious, have at it. Have fun.
Remember, if you’d like to chat with us or you have something nice to say, or not so nice, we’ll chat either way. You can always reach out to us at Epicureanunicorn at Parados.com. We’ll be happy to hear from you. You can also write a review of the show. Give it five stars. Anywhere that you want to. You could do it on Apple Podcasts. You could do it on Overcast. You could do it on Amazon. We don’t care. Wherever you want to write something nice, please feel free. And you get to come join us here in Club Unicorn.
All that being said, I will now take my leave so you can enjoy this conversation that Zhejiang had with Andres. All right. Hit it.
Zhixin Wang:
Amazing. Thank you, Brayden. And nice meeting you, Andres. It’s very exciting to have you today on the podcast. And you mentioned six-seat consultancy. So first question for you, it’s really what inspired you to co-fund the six-seat consultancy? And how has the journey been so far?
Andreas Duess:
So, yeah, I mean, those two things are really related. I have a long history in food. When I was at university, I went to university in London in the UK. I worked as a line cook in restaurants. And so really, I still cook for the people I love. I’m the main cook in our household. And then after university, I ended up in technology. I worked with early Google, early Facebook in the UK. And I worked for a marketing agency that specializes in working with technology clients. And I got exposed to early AI through a company we worked for called Autonomy, founded by Mike Lynch. And I then moved to Canada. I met the girl I was going to marry. We ended up living in Toronto. And I accidentally founded a marketing agency that specialized in food and drink, applying a lot of the focus lessons that I had learned while working in technology. And it grew from a, when I say accidentally founded, it grew from one freelance job into Canada’s largest independent agency for food and drink clients. Offices in Toronto, offices in Montreal.
And then about five years ago, I got pinged by a friend of mine from my tech days. And she said to me, I want to show you something, but you can’t tell anybody about it. And I said, sure. And it was an early large language model. And it was two things. It was absolutely awful. And it was very much the future. And, you know, I’m old enough to remember when digital replaced film in shooting commercials. And it was exactly the same. It was terrible. It was terrible. It was terrible. And everybody laughed about it. And one day it wasn’t anymore. And everything changed overnight.
And so I went home and immediately had a panic attack because I could see how this was going to change the world. And I was worried that this would completely destroy the value that we had built in my company. I then got very lucky that somebody was interested in purchasing. I was a 50% shareholder at that time. I had a partner. Somebody was interested in buying me out and taking over the company. And so I took the value out of the company. It gave me runway. And it gave me time. And it gave me the ability to sit back and think about what I wanted to do next. I wanted to stay in food. But I also wanted to combine this with technology, with data, and with AI. Because those are things I really care about. And so I started brainstorming for about a year with people in the industry. My friends who are marketers, friends who are running businesses, friends who are at large CPG brands.
And the problem that kept coming up time after time after time was we have a huge failure rate in launching products, bringing products to market. And the failure rate, the failure rate, the failure rate, the failure rate hovers between 75% and 82%, depending on whether you’re an independent entrepreneur or whether you’re a multinational. And the cost to the U.S. economy per year is around $24 billion. And that translates into just under $300 out of the pockets of every U.S. grocery buyer. So we said, well, how can we solve this problem?
And we started looking into data inputs. What kind of data can we work with? And one of the things that we noticed immediately when we looked at data, when we talked to people at Nielsen, we talked to people who really have access to data. We identified this thing which we call the say-do gap. Which basically, see, people will say one thing and then do something entirely different. And that is specifically important on social media. So if you’re using social media to make decisions, please don’t, because people will lie to your face in order to look good.
And so we said, well, how can we circumvent this? And we started talking to people who had access to really good data. And we started building a data lake. So we have access to data from retailers. And it’s not Nielsen data, but it’s data that’s coming from credit card suppliers. It’s data that’s coming from the Instacarts of this world, etc., etc. So we know what people are buying. And then we have data. We added data from food service. So there is about four and a half million food service menus we have access to now. And so we know what people are buying in restaurants, when they’re buying it, how often they’re buying it, what time of day they’re buying it, are they repeat buying it, etc., etc.
Then there is data from people’s homes. And that data, for example, includes data from smart fridges. So the data knows when people buy a chicken and then eat it the day after and they buy greens and they rot away in the back of the fridge because they were a guilt purchase rather than something they really want to eat. And then all of this then gets compared to the social conversation, to the public conversation that’s out there. So if we know that somebody purchases the ingredients for something, let’s say, less than healthy and then eats that less than healthy food and then says on social media, I’m only eating organic vegetables. We know that’s not true. So we can tell the say versus the do.
But, for example, if somebody comes in and say, you know, I’m really like this. I want to eat high protein, high fiber. We know. And again, we can’t pin this down to the individual person, of course. But we know that people are making the purchase of these, let’s say, better for you ingredients and then they get consumed and then they get talked about in the public conversation. We know there’s a very high likelihood that these things actually happened. So the number of data points that we’re working with, around one trillion data points, constantly updated. And we have this data for the US, Canada, Mexico, Brazil, most European countries and Australia and India just came online.
So on top of this data then sits an AI model that allows us to actually make sense of all of these data points. Because otherwise, you know, we just as human beings, there would be absolutely no way we can do this. And then our associates, the people we work with, we use these data points to help our clients do two things. We help them bring products to market that people want to buy. And we help them communicate about these products in a way that people want to interact with. Nielsen released a study last year, I believe, in which it said that 70% of all marketing campaigns in CPG cost more than they earn. And our goal is to turn this around, turn the failure rate upside down and also turn the marketing rate upside down. And the way we do this is by saying, look, we know what’s important to a 35-year-old woman who works in healthcare, lives on the West Coast, versus a 60-year-old man who works in manufacturing and lives in Detroit. You know, we can really drill down into the importance, what’s important to those different people. And we can then help our clients develop products and develop communication that talks to them. And just to bring this to life a little bit, currently working with a company out of the U.S., a rice company. And I’m sure Ozempic is on your radar as a scientist. Ozempic and GLP-1 drugs, how they are changing the CPG landscape right now. So we are helping this company to develop a rice product specifically designed to support the need of Ozempic patients.
Zhixin Wang:
Wow, that’s a lot of information. Sorry, yeah, I know I talk a lot. I was just going to say, everything that you mentioned actually excites me a lot. I’m actually Canadian, so anything that started somewhat around Canada that touches Montreal, I’m very excited about it. I have a passion for food science and also technology combined. I love AI technologies. So all of that, it’s really, really exciting. And you touched, actually, on a lot of questions that I had for you today. So one of them, the next one was AI-powered market insight. So how do you really see within the next five years of how AI can be transforming the food beverage industry? And you mentioned using data, collecting data to help CPGs, right? Along that line, for the impact for food innovation, how do you think currently AI is already doing in this industry food and beverages? And also, what future application we’re talking about, like, more than over five years from now? We know how technology is so fast nowadays. So other than the near-term insight that we get from AI and the benefits from it, do you see any benefits from it, any downsides from it on the longer term?
Andreas Duess:
So first of all, I think the long-term impact of AI. I think of AI similarly to electricity. that it is an underlying power that allows us to do a lot of things. It allows me to switch on the lights in my house. It also allows you and me to have this conversation right now. It allows people to drive their cars, to take trains, to take public transport, but also to do a million of other things. And that’s what’s going to happen to AI. That AI will be everywhere. And it will power a million things. Lots of which we don’t even know it will power today. But I think people will become less and less excited about it. I don’t get excited about my power, my electricity provider, my power provider. You know, I know vaguely who they are. I know what they’re, you know, where they come from and all this kind of stuff. But I don’t really stay on top with the latest development in water turbine development. You know, it’s not what I’m interested in. The same thing is going to happen to AI. I forgot the name now. But there is a story from the 1700s. It’s called something about a paradox. I might have to look it up. Where steam engines, when they first arrived on the scene, were so power hungry that you could only operate them in close proximity to a coal mine. Because they needed so much coal to operate. And as a result, their application in society was not very widespread. And a lot of people called them a fad. A lot of people said, ah, steam engines, nonsense. They’ll go away. And all that kind of stuff. And then James Watt came about, engineer. And he increased the efficiency of steam engines significantly. So within five years, he decreased the amount of coal a steam engine needed to put out more energy than before, more power than before by 75%. Two things happened. Shares in coal mines collapsed because people thought that now you need less coal to do what you were doing before. The opposite happens. Steam engines popped up absolutely everywhere in the most ridiculous use cases you can think about, which is exactly where we are with AI right now. AI uses a lot of power. And then we have Deep Seep coming along, which uses a lot less power and a lot fewer resources. And all of a sudden, AI pops up. And it’s perfectly fine for that we currently have ridiculous AI applications where people really shake their head and say, that is just stupid. Chances are that’s true. There were really stupid steam engine applications when this first happened. But over time, steam engines took over power generation in big cities. They took over transportation through locomotives. They took over agriculture by putting the first usable tractors out and mechanizing farms. And they powered the entire industrial revolution. And at the same time, use of coal, for better or for worse, absolutely exploded. And people who were running coal mines were getting seriously and very, very wealthy. So that’s where AI is going to go. Right now, everybody talks about AI as something that’s interesting. What is going to happen in the next five years, we are going to talk about the things powered by AI, which will be really, really interesting. And the underlying power, I’m not suggesting that we’ll lose interest in it completely. And I’m not suggesting that we’re going to stop researching those things. But they are less and less and less interesting when compared to the actual outcome. So that’s one of the things where I’m predicting that this will very much happen. The other thing that’s important to keep in mind is we call this meaning making versus sense making. And what we mean by sense making is if we talk about food and let’s talk about perhaps a food startup as an example. The majority of food startups that I’ve ever seen are based on emotions. I wish to celebrate my grandmother’s heritage. My child can’t eat XYZ. And I want to put a product in the market that allows children to take part in life to the fullest extent. So they’re almost, most of them are emotionally driven. There’s very, very few food entrepreneurs who come into the market and say, I just want to make a lot of money. And, you know, I’m going to do it this way. Now, that’s meaning making. Humans are really good at being emotionally driven and creating human meaning. And then we use the AI to make sense of this. So here is my gluten-free cookie that I want to bring to market. Or here are my grandmother’s samosas that I want to bring to market. Now, AI, work with me. What ingredients make the most sense based on the flavor profile that we have access to in your data? Who will buy these? How much will they cost to make? And while we’re at it, what are the standard margins that distributors ask for? So I now have a really well-informed co-worker working with me to help me reach my human goal. When AI goes wrong, it’s typically for one of two reasons. Reason number one is people outsource the human part of the work to AI. Or they’re trying to outsource it and say, you know, do this for me. Whereas when things with AI go typically right is when you say, do this with me. Work with me on this. And help me create my emotional outcome. And help me work towards the outcome I want to achieve. And correct me and direct me when it feels and looks like I need to change or adjust my dreams in order to make a commercial reality out of this. Nobody needs, the market doesn’t need another chocolate chip cookie. The market doesn’t need another undifferentiated product that enters a crowded marketplace. And a lot of people fail because they don’t know this. And strangely enough, even in the big companies, even in this larger, larger brands we talk to, innovation and this kind of thinking is very, very often a pain point. I never forget. This was a significant player in the CPG market. And I spoke to one of their VPs about the innovation process. And he looked at me and he said, it is the worst day of my month. We have innovations meeting every first Monday of the month. Everybody scrambles and Googles the night before. Everybody comes back with the same ideas. And everybody is just stressed about it. Right? And so giving people the tools and giving people this co-worker that allows them to reach their goals, that is a really good use of AI.
Zhixin Wang:
Yeah, definitely. I think you answered perfectly a lot of the questions. So predictive analysis is one of the next that we’re going to ask. How AI can help the food brand to respond to consumer trend. You mentioned analyzing what the humans, I’m telling the human what it may need emotionally, and the AI can help me analyze the data to see what would work better. And personalized consumer experience as well is something you touch on. How can it personalize it to my liking, for example, or a specific set of niche consumers if I give a data or an instruction to the system? So I really like also your thoughts of it’s not AI taking our place as human, but really us humans having a buddy, so something extra to work alongside with. So I really, really like this thought. And following through that more into the industrial scale. So when we talk about supply chain optimization, there’s a lot of data going into that. How do you think AI or technology could be used to leverage that sector? How can we reduce, we talk about in food industry, a huge amount of waste, food waste? How can AI maybe help humans reduce the waste, making something that’s currently negative into a positive?
Andreas Duess:
Yeah, I think you touched on something so important there. And one of the ways this can happen, I would, before I go to the waste question, I wanted to sort of touch, circle back slightly on something you said at the beginning. And that’s about getting the right information, the right products, the right communication to the right people. One of the things that we do today, and I said earlier on how we can create, how we can actually target data to the right people in a very specific way. What we’re doing now as a company, we are setting up AI-powered consumer panels. And so what we have, we have created AI people. And we started doing this by having people come in, paying them money, and asking them 500 questions. And then we found out that we actually don’t need to do this, and that there is enough data available in the general data pool that allows us to create people with great backstories and a great sense of reality through data alone. Then once we have these people, we then train those models with specific data about the market. So for example, sourdough. I know sourdough is a Pareto’s product. You have the sourdough museum, I believe, which is really exciting and really interesting. So let’s say we have a person, a 35-year-old woman lives in Austin, Texas. She works in tech. She’s married and all this. So we have this person. And then we train this model with bread consumption, for want of a better word. And then we do this to, let’s say, five other people. And then we can ask these models questions about their preferences as they rely to bread that are an overlap of around 90% to 95% with a real person. So now as a company, you have a consumer panel available to you. At any time that you can ask questions about flavor preferences, you can ask questions about packaging preferences, you can ask questions about what’s important to them at this particular stage in your life. And as they’re constantly updated with the latest observed consumer data, those panelists will always be completely up to date with what’s going on under the market. And the important thing is, this is not data that’s six months old. It’s not data that’s a year old. This is data from last week. So if you’re a product developer, you can say, I want to talk to Jane and James because the product I’m working on is for a Jane and for a James. Talk to me about the things that matter to you right now. And all it is, it’s a clever way to create an interface that’s very human in its nature, because talking to other humans is natural to us. It’s an interface that allows the user to interact with data through the interface of a human avatar in order for them to do the one thing that makes AI really, really, really useful. And that is talk to it. Talk to it. We have been trained over the last, I don’t know, how long have PCs been around? 40 years, perhaps? We’ve been trained as human beings that there is a right way and the wrong way to use a personal computer. And if you do it the wrong way, you get no result. And in worst case scenario, you break it somehow. Right? There is a right way to use Microsoft Word. There is a right way to use Excel. There is a right way to use Zoom. And now, because of this, people are looking for a right way to use AI. Except there isn’t one. Talking to AI is very, very similar to talking to a human. So when we’re working with Microsoft or when we’re working with typical software, it’s a one shot. You tell it what to do or you, you know, what to do and it comes back with one correct way of doing things. But with AI, there are many ways of doing things. And there are many different answers coming back. And so the more we can move away from the one shot software model and the more we can move towards a conversational model, the better. I’m going to share with you something that I use in my personal life. So I was at a tech event last week. And one of the speakers spoke about something that resonated deeply with me. It was a really interesting observation. It was something I’ve been thinking about. And I thought to myself, this will make a really, really good newsletter for our company. I left the meeting. And I live very central. I live very downtown. And I decided to walk home. And it was like a half an hour walk home. I put in my headphones. I fired up ChatGPT in advanced voice mode. And I said to ChatGPT, look, I’ve just been to this meeting. The speaker spoke about the following things. You know this is important to me because ChatGPT, our account, has a really good idea of who I am and what I do because I tell it an awful lot about myself. And I said, I want to put this in a newsletter together with some of the thinking that I’ve been sharing with you. Can we work on this? And so we had a conversation. And it was a conversation with a digital co-worker that I was having for half an hour. By the end of the half an hour, as I was coming around the corner to the street where I live, I said to it, that’s excellent. Really like this. Turn this into a draft for the next newsletter. I’ll come back to you. And then I opened it up the next morning. And there was a fully fledged draft. Really well thought out. Really well researched. And again, meaning making versus sense making. I wanted to express something. And then I worked with the AI to arrive at that goal. And the AI was my co-worker. But it was conversational. Right? So back to the question that you really asked me earlier on before I started segueing like crazy. You started talking about waste. One of the most powerful use cases in my belief for AI is the question, what if? Typically, what if questions in traditional business are really difficult to deal with. So one of the most powerful things that you can do with AI is to explore what if questions. And what if questions in traditional business are really expensive. Because they cost money to research. And they cost time to research. And so very, very often, what if questions just aren’t asked by people. Because I’m already working full time. I’ve got all these things going. But for example, if I can go to an AI model and say, you know what? What if you’re going to do with them? I found that there is a vitamin in it that’s really easily extracted, and there is a market for these vitamins here. So why don’t we take this waste product and turn it into the profit? Why don’t we turn it into something more useful than currently sending it to a composting farm? So that’s where AI comes in as an incredibly useful partner. And I know there are people out there right now that are putting together a huge database of ingredients. You probably use some of them yourself. databases of ingredients, database of waste materials, analyze the ingredients or the things these are made of, the vitamins, the oils, the fiber, whatever is in there. And then AI is incredibly good at pattern matching and figuring out what to do with those.
Zhixin Wang:
Yeah. I actually really enjoyed the conversation so far. It made me think while you were speaking as well, how do I use AI on a daily basis? I think it’s a great way to do that.. the AI robot, if we can call it as such. What would you give as an advice for an individual consumer, but also on the other hand, for a startup or full-size company in the choices they made determining which system they might want to go for? So there are so many available, right? Sometimes too many choices can make it a little more difficult. So what would be your advice on that?
Andreas Duess:
I give you my tech stack. And I genuinely think that going back to what I said earlier on, what will happen is that there will be software available that’s AI powered, but the fact that it’s AI powered will almost be redundant. You know, there are a lot of meeting takers, meeting note takers right now. You know, where you’re in a meeting and it sends you all the notes afterwards and it’s beautifully formatted, et cetera, et cetera. That is sort of a thing that’s becoming table stakes. It’s normal. That’s what we expect these days. And the fact that it’s AI powered is neither here nor there. All we want to get is great meeting notes out of our meeting. So a lot of things will happen that will allow us to do our jobs better, connect better, do things that we want to do. And the fact that the AI powered is neither here nor there. So one of the things, for example, that we offer, we have an AI model that’s been trained on 25 million recipes, all legally licensed. And that AI model is then connected to the data firehose that we have access to. And now if we are working with a test chef, for example, like, you know, who runs in a test kitchen, that test chef can come to us and said, look, I need inspiration. We’re talking to millennials living in Boston. But we are also selling to boomers living in Austin, Texas. And I need to make sure that the recipes we’re putting out are relevant for these groups. And then what our AI model does. And again, it doesn’t matter to the chef that this is AI. The chef just wants these results. Our AI model goes, reads the data, understands what’s going on, looks at its training, 25 million recipes, spits out 200 recipes or 100 recipes relevant for the needs of the test kitchen chef. And then the test kitchen chef takes over and says, out of these 100 recipes, these 10 really hit the mark for me and let me test them, improve them and perfect them. But what the test kitchen chef now doesn’t have to do is guess anymore and Google for two weeks, right? That person can now focus on the thing that makes her or him a better chef. And doesn’t have to focus on endless, mindless, looking through pages, trying to find inspiration. And that’s where AI really, really shines by helping humans to do more of the things that humans do. Now, on a personal level, I honestly, if you’re looking for an AI model that helps you to do things better in your personal and your professional life, I honestly think that the chat GPT is a great Swiss army knife. You’re right. It does an awful lot of things reasonably well, better than a lot of people need. Then there is a search engine that I personally highly rate called Perplexity. Perplexity is has replaced Google for me 90% of the time. And it is an AI powered answer finder and research partner that does a really good job. And then if you want to go into coding, if you want to play around with coding, if you want to play around with creative writing, there is another model called Claude that I really like. So those are the three I personally pay for to work with me. And, you know, you said earlier on, how can how can you use this in your personal life? One of my personal favorite use cases is to use AI as an expert. So, for example, part of my job is following up with people via email who don’t always want to talk to me because they’re worried I might want to sell them something. Right. And I’m not very high in the priority of emails that need answering. And, you know, there is always a sort of awkward email opener just following up, you know, and it’s basically it screams desperation. It’s like, you know, a spurned lover turning up your front door and you really have more important things to do at that particular moment in time. And so there is this woman. She lives in Texas, I believe. And her name is Jennifer. Jennifer, Jennifer, I believe Jennifer van Edwards. And she’s an expert in human behavior, how humans interact with each other. And she’s very public. There’s lots of YouTube content about her. She’s written a ton of books. She’s been interviewed on a million podcasts. So her knowledge has been ingested by the large language models. And it’s part of this. So I was writing one of those emails and I was dreading it, quite frankly, because I needed to get an answer out of somebody. And I wanted to know if the project was dead or not. And if it was dead, you know, we could stop thinking about it. And I was writing it and I wasn’t really I was reading my own email and thinking I this just doesn’t sound as great as it needs to sound to get an answer. And so. We I. Uploaded my email to chat GPT. And then I said to chat GPT. How would Jennifer write this email? Teach me. Teach me. And chat GPT came back and said, well, Jennifer would never write a terrible email like this. You know, instead it would do the following things. It would start with empathy and demonstrate expertise. And it would start like this. And here are some openers that will get you better results. And one of these openers that I’ve been using to increase my response rate by an easy 50 percent. Other words, the last time we spoke, you said. And another opener is I was just reading, seeing this and thought about how this would be relevant to you. And just changing from, hey, just following up on this project. Are we still on just changing my language very slightly and adjusting it to immediately instead of being somebody who asks for something is somebody who brings value into the conversation from the get go. And that’s completely changed the way people respond to emails of that nature for me. For me, one of the most powerful and I’m teaching this to my children as well. One of the most powerful uses of AI. Again, don’t tell AI to do stuff for you because it’s knowledge in air quotes is so broad that it’ll do a terrible job. It’ll be, you know, a race to mediocrity. But if you say, look, I want to communicate to Shazing that she really should work with our company and we bring great value to her. And, you know, she’d love the experience of working with us. How do I do this that you don’t open the email and feel either vaguely annoyed or bored with what I have to tell you and just, you know, hit the spam button so you never have to hear from me again. And so that’s really, really powerful for one of my children. He, my oldest son, is quite heavily dyslexic. So super intelligent boy. But words dance around his eyes, you know, the written words and all those kinds of things. So one of the things that we’re doing with him now with his teacher’s approval, all his coursework goes into Google Notebook LM. And one of the things that Notebook LM is really good at is to create a podcast out of written content. So when he gets something from school, it goes into Notebook LM, translates it into words, and then he can listen to it. Right. Right. Again, don’t do it for me. Do it with me. Don’t be my substitution. Be my mentor, my teacher, my supporter, my minion. That is the one thing that is incredibly important to remember. And the other thing that’s important to remember, in my estimation, is what if? What would happen if? And don’t go for the one shot. Go for the conversational approach. And that gets you further and farther than you ever thought possible without the help of the guy.
Zhixin Wang:
I really like that thought again. It’s about the whole relationship, right? Daily lives on relationship with our parents, with our kids, with the pets at home. Now, AI is there, so it’s also a relationship with the AI. And it’s a relationship to foster on both sides. So we are teaching the AI how to help us. But at the same time, AI gives us feedback. And then it’s a conversation, as you mentioned. So maybe the first feedback is not exactly what we want. But then the conversation continues and then might end up in solution of a bigger challenge. And I like the information you provided. It’s not one size fits all. It’s my needs and X, Y, Z. Then I look for a model that will fit those needs. So I really like all of what you said. And it comes to the question, so challenges and opportunities. What do you think are the biggest challenges and also the biggest opportunities that we could see with AI within the food and beverage industries?
Andreas Duess:
Yeah, there are many challenges, right? I mean, and again, those challenges are historical in many ways. You know, when the first printing press opened in Paris, it was burned down by an angry mob of medieval scribes who were all saying the things that people say about AI today. It’s not authentic. It’s not genuine. It’s not authentic. It’s not genuine. And by the way, we’re all going to lose our jobs. And, you know, the French are fairly good at rioting. So they immediately rioted, burned down the house in which the printing press was. Didn’t change a thing. So I think it’s important to listen to people who are scared of this and who are worried about this, that it will change their lives, that it will take their livelihood away from them. And I think we need to listen to this with compassion and understanding. And I think anybody who tells you that jobs won’t be impacted is not telling the truth. Jobs will be impacted without a shadow of a doubt. But there is a clarifier there. The jobs that we have today will be impacted. I read two weeks ago that in the last census, 71% of all jobs mentioned in the census didn’t exist in 1955. Not too long ago, right? So will there be change? 100%. Will there be dangers? 100%. It will be really easy to impersonate people. It will be really easy to twist the truth in many different ways. So we need to educate our children. We need to educate ourselves. We need to educate those whose job will be impacted by AI to work with AI rather than despair and pull back. And there are, you know, on a societal level, on a broader level, out of the food industry, there are many conversations that need to be had. What is genuine? What is truth? Should we start paying people a universal basic income? You know, how is AI going to affect society? But I think one of the things that’s really important to remember, every single time there have been prediction of mass collapse of the workforce, and there have been many of those over the last hundred years. The last one, I think, was in 2014, when the New York Times and Oxford University had out a study. New York Times reported on it. That by 2024, 50% of all jobs in the US would be lost to automation. And of course, 2024, we had nearly full employment and people were generally very happy with that work. So I am a tech optimist rather than a tech doomsayer. But I do think what is important for this optimism to work is the red thread that’s been through the entire conversation we’ve been having, right? Use AI to make humanity stronger rather than something to replace it.
Zhixin Wang:
Yeah, definitely. And it’s a lot of time, I think, food industry is leaving the mundane work to the AI system and keeping, or I always like to say, even though we’re in science, it’s really a science and arts. It requires a lot of art part of the science aspect of it, of developing products. A lot of the managers in food industries do product development. The development side of it can be different thoughts that AI wouldn’t be able to do, but could help us fully achieve the end goal. So, yeah, I really like that. So, can I summarize in a way that the biggest challenges are humans, but also biggest opportunities are humans too?
Andreas Duess:
Yeah, I think that’s fair. You know, I do think if we use it right, like in my own company, one of the things we didn’t predict would happen, but we now, part of our business now is help clients, help our clients to implement AI in their company. And to teach people to make that shift from there is a right way and a wrong way to there are many, many different right ways to do this. So, that’s a business we never thought would be relevant for us, but it has become relevant because AI is so deeply ingrained in our own company. I think of the AI models that we have access to at Succeeds, I think of them as co-workers. Right? So, I think the capabilities of somebody who is a good manager, clear feedback, clear problem definition, shared responsibility. I think those traits are also exactly the kinds of traits that make somebody good with AI.
Zhixin Wang:
Yeah, definitely. Yeah, I totally agree. It’s, I really like to think about as well as the AI being someone who works with you, that’s willing to do everything, anything that’s possible. And us as human managers said, if we manage it well, give a good feedback, constructive feedback, it can give back much more. And yeah, it’s, at the end of the day, it’s endless possibilities, right? And in terms of AI and humanity, I did a little quick search on congratulations on the book. I think it was from end of last year. So, that’s very exciting. Do more with less. What are some key takeaways for businesses? Quick summary of what people should take from reading the book.
Andreas Duess:
I think, for reading the book, the main theme that the book is all about, it’s all about the friction reduction, right? Because the things that stop us from doing the things we really want to do are very often related to work that is outside of our core capability. And in business, there are two different ways to overcome friction. And one is, and very many people use this as a default. You know, let’s say we’re in a room and the door is stuck and we want to leave the room because there is an opportunity outside that room. Very, very often, people just try to break that darn door down. They push against it, they shoulder against it, they push it, it’s heavy, it needs energy. Hey, I need the hand here, three more people come and help me push that door down. AI, on the other hand, helps us oil the hinges. And solve the problem where it matters. So rather now than having to expend all this energy against pushing a door that’s stuck, getting three more people and having to pay three more people to help me push and everybody is sweating and grunting and it takes a long time. What happens if I apply some WD-40 to the hinge and all of a sudden that door is openable by the push of a single finger? That’s if you use AI intelligently is a really good way to do this. And I’m going to share with you a little story from when I was writing the book. I was writing a chapter. And there was a story in there, in that chapter, that I think made a really important point. But I couldn’t make it work. It stood out. It wasn’t part of the flow. And I struggled with this for about two weeks. And then I remembered I had built myself an AI editor. So, and the way I did this when I first started writing the book, I wrote a chapter. I had it edited with a human editor. I had a structure what the book wanted to say, who it was for. And I say, you know, this is how I wanted to sound. This is the format. This is how chapters should flow. We start with an anecdote. We go to data. We go back to the story. So the editor, the AI editor, had a really good idea. And I did something I had never done before. I uploaded my trouble chapter that was giving me so much. grief. I uploaded it to the AI editor and I said, I can’t make this work. Should I just delete this? Because I don’t think it’ll fit. And then I used words which I had never used before. I said, what do you think? And I felt a bit silly using that language. But I thought, eh, what if? You know, the worst can happen. I hate it. And the AI came back and I said, no, no, no, no, no. Don’t delete this. This is really important in the context of what you’re trying to do. And it illustrates your point perfectly. Your problem is that you’re putting it into the wrong place. It doesn’t belong where you’re putting it. It belongs way over there where it illustrates the point and flow and works with the narrative. And that, for me, was such an absolute eye-opener that the AI could look at the hole that, in my head, is difficult to keep, but it could look at the entirety of the book, could look at the context in which language flowed within the book, and then said, that belongs there. You’re trying to put a round peg into a square hole. And that was an incredibly, that was an incredibly, that was a real eye-opening moment for me. And it was a moment of incredible usefulness. And I think for companies, there are many pegs and many holes all the time. And very, very frequently, we try to put a triangular peg into a, you know, round hole or what have you. But when we’re using AI in a business context, and we’re feeding it information about the things that are important to us, our values, the things we want to achieve, then the AI can really help us to put the right solutions in the right places and solve the right problems that really take us further. And that, for me, is one of the most powerful use cases.
Zhixin Wang:
Yeah, that’s, that’s amazing. I really like that I will do more with less. It reminds me of a manager once in a discussion as well, saying, work smarter, not harder. So I think AI really help us to work smarter, not harder in that case as well. One of the few last questions that I have for you as well, future of food and drinks with AI technologies coming up there, people being scared of using it. And we’re going to talk about maybe cell culture, meat coming into the play, which is also full of technology. How do you think in the future would make people accept this technology easier? Or do you also think that just like in history, it’s going to be rejected first, but a new normal will maybe be established. And it’s really, as people understand, it’s us working alongside AI and not AI. AI will, as you mentioned, will likely take, replace some aspect of daily lives. But in general, what do you think the future would look and how would it make it easier for people to accept it?
Andreas Duess:
We have never seen a more fragmented consumer market than we’re seeing this today. You know, when I was 20 and I threw a dinner party because I always liked to cook and I had friends around for dinner. A dinner party consisted of me saying, come to my house, we’re having braised lamb shanks. So I didn’t even tell people, bring a bottle of red. And then we ate and frequently drank way too much as you do when you’re 20. And before anybody sends you angry mail, I grew up in the UK where the legal drinking age is 16. You know, so we used to have long dinner parties long into the night. But basically I cooked and people came and ate. We still have dinner parties today. But now when we have dinner parties, I send out a quick email and I say, hey, is there anything you don’t eat? And very frequently people say, well, I’m gluten free or I don’t eat this or I’m vegetarian or I’m vegan. You know, I’m vegan. And I don’t mind because, you know, food is all about inclusiveness. And so we make something happen that gets people around the table that I would like to spend time with. But this is so for me in my own home, this is something that’s relatively easy to deal with because I have the skills and I have the willingness to do this. But for brands, this is really difficult because now you have to work with so many different needs and so many consumer desires that it’s incredibly difficult for brand to hit them all. And invariably, you have to choose where you focus. And then AI here will help us invisibly in the background to solve these problems. I was in Saskatchewan recently at an agricultural event and I saw a combine harvester that has AI powered sorting mechanism that sorts the grain based on its protein content. And so in the back, it has two storage containers, one for high protein, one for lower protein grain. And those can then be split at the harvest level already into a higher protein one that’s being sold for, let’s say, bread flour or pizza flour, et cetera, et cetera. And a lower protein grain that can be used for everything else. So that is, again, a way where the fact that this is AI is like the fact that it’s electricity. Nobody is excited about the AI, but everybody is excited about the fact that now a higher value grain can be separated at harvest, put in its own supply chain, and then make more money immediately from that point onwards. You know, higher profits for the actual farmer. So we’re going to see more and more and more of these things appearing. And from, again, from a, for food companies, you know, our clients, clients that work with us with six seeds, they have a really good idea of what’s important to their clients. And we can go and we can go and we can go and say, look, you know, you are selling, I’ll give you an example. We worked with a commodity group last year and it was all about blueberries. And blueberries for the longest time have been sold on the health benefits at the commodity level. So I eat more blueberries, they’re healthy. And we sort of looked at this and we said, yeah, that’s not. But, and then it was, honestly, it started as a what if question. Because we looked at the accepted wisdom, we looked at the results this marketing campaign we’re getting in market, and I’m not going to lie, they weren’t amazing. And I said to, I said in a meeting, I said, well, what if people don’t actually buy blueberries because they’re healthy? What if there’s other drivers there? And what if they’re not reported? What if there is a say-do gap? And so we looked into the data that we have access to, we fired up the, well, the AI isn’t fired up, it’s always running. But we started asking questions and interacting with our model. And what our model told us is that people don’t buy blueberries because of the health benefits. That’s accepted, that’s there. One of the main purchase drivers for blueberries is that they’re pretty. People buy blueberries because when they bake something, blueberries look like little jewels. And they’re a good looking fruit. And that was one purchase drivers that exceeded the health benefits of blueberries. So that allows our clients now to say, okay, we are going to keep the health blueberry message as an underlying message. We’re not going to completely move away from this. But every single photograph we’re going to put out, we’re going to focus how pretty blueberries are. We’re going to focus on. We’re going to focus on the market. We’re going to focus on the market. We’re going to focus on the market. We’re going to focus on the market. We’re going to focus on the market. We’re going to focus on the market. And that is just a practical application of the things you can find where all of a sudden your sales and your relevance explodes in market.
Zhixin Wang:
Yeah, I think the blueberry thing really surprised me. Think of thinking outside the box, right, which AI helps us do that. So I think that’s really amazing. And we could talk about AI days. It’s something with my partner as well. He works in tech over food as well, which is my love language as well that we discuss a lot on a daily basis on the weekend and going to grocery shopping, future trends. When we talk about that, you mentioned a lot of trends that are currently ongoing and also potentially in the future. If you have to choose one that is the most exciting for you, that’s AI driven, what would it be? And the last question would be, what advice in one or two sentences would you give to the younger generation getting to the workforce that are utilizing those new technologies?
Andreas Duess:
I’m going to answer your last question first. I have three teenage boys roaming around my house, eating everything that hasn’t been nailed down. Honestly, I could have a conveyor belt to Costco these days and we still wouldn’t have enough food in the house. But they are at an age where in a couple of years from now they will go to university. And my advice to them is study the humanities. Study philosophy, study languages, throw in some religious studies by all means, right, for all kinds of different things. So, throw in everything that helps you figure out what makes humans human and then use that knowledge to figure out how AI can help us to be more human and focus on the things that make us human. So, for the younger generation, that really is my main advice. The other one, it’s exactly the same, what we talked about repeatedly in this conversation, have conversations with AI. Use it as a conversational tool, not as a one-shot tool where you put one question in and then you expect one answer to come back. That’s a really important one. That’s a really important one. And then as far as food and drink is concerned, I think the main thing that is going to happen is that AI will allow us to create food that nourishes us, that feeds us, that is high in nutrient value, that is better for us all around. By helping us to grow food and make food that answers fundamental human questions. Fewer pesticides, fewer chemical treatments. Again, you know, when it comes to, I know Pareto’s is big when it comes to sourdough. How can we use sourdough perhaps in order to achieve certain outcomes that increase gut health? How that can we improve to help us to improve mental health for our customers? How do all these things work together? And again, I think if you’re a food scientist or if you’re involved with formulation of these things, if you have access to AI and you can say, hey, tell me. You know, for me, and you’re a scientist much, much more than I am. But I read something years ago which really stuck with me. And it said the most important words in science are not “Eureka, I found it.” The most important words in science are “Huh, that’s weird.” And I think that AI allows us to dive into those “Huh, that’s strange.” A lot more effectively, a lot faster, and a lot more with a lot better outcomes than what we currently can do.
Zhixin Wang:
Yeah, definitely. As a scientist, if I look back 10 years from now when doing lab work, I would have loved to have AI at a fingertip and ask, “Why does it not work? What did I do wrong?” So I think, yeah, the future is very exciting. Conversation could last forever, but Seek of Time, I think was amazing. We went through a lot of questions. Safe Seeds as well, a very exciting company that does a lot in terms of AI and helping companies from small to large solve the world problems. And, yeah, I actually look forward for having a potential future discussion with you on AI, food technology in general. And I’ll thank you so, so much today for answering all my questions with such precision. Fantastic.
Andreas Duess:
Thank you, everybody.
Brayden:
I’m back. It’s me again. I’m only here, though, to say thank you very much to our guest. He successfully caused me to send, Jijing, an email after that saying, “Hey, alright, I think there’s someplace maybe kind of, sort of, we could potentially consider forming a committee to think about the benefits of potentially using AI for something tangentially associated with work. Maybe kind of, sort of, think about it, get back to me.” So, congratulations. That’s a huge step for me. So, again, thank you to my special guest host and our guest. It was a wonderful conversation. And thank you to all of you for listening. We appreciate each and every one of you. And until next time, take care, stay well, and be seeing you.
Hey, everyone. Thanks for tuning into our wacky world of edible exploration with Epicurean Unicorn, a Paratus Corporation production. If you have questions or any little thoughts on today’s topic, send us an old electronic piece of mail at [email protected]. In the meantime, keep on talking about the magic of food with your friends, fam, I don’t know, even your local ice cream man to keep culinary curiosity alive. Bye. Bye. This has been a Studio 47 production.