EP183. AI in beauty - friend or foe?

Everywhere we look, everyone's talking about AI. Since ChatGPT burst onto the scenes earlier this year, all of a sudden we now have the potential to perform lots of tasks faster, smarter and more accurately. And we've certainly seen its benefits here at formula botanica too. ChatGPT assists me and my team in many tasks which have made all of our lives much easier. For example, I recently got ChatGPT to write me a press release. Took me five minutes. I had to frame my question properly. And then bam. There was my perfect press release. It was so cool. I changed two words in total. So the potential of AI is enormous for all of us. But I'm still seeing the beauty industry grapple with this idea of what AI could mean for us as an industry. So, for instance, could the AI of the future design our cosmetics or even formulate for us? What an idea. Could AI make all of us formulators obsolete? Perhaps? Could AI make the beauty industry more sustainable? We're in the days of infancy when it comes to AI, so who knows where the future will take us? So, to get out the crystal ball and answer some of these questions, I knew I needed to invite someone who is already completely embedded in beauty and is already using AI. So my quest immediately took me back to Cavallo, the ingredient search engine that aims to be the Google for cosmetic ingredients. I interviewed one of the two co founders at Cavallo, Timo Von Bargen, back in episode 79. So it's now time to hear from Cavallo's other co founder, Jan Chilvers. Settle in for a fascinating episode where we ask what role AI is going to play in beauty in the future. Welcome to Green Beauty Conversations, the podcast that challenges you to think about how you buy, use, make or sell your green beauty formulations. In my weekly episodes, I explore the sustainability of the beauty industry and encourage you to join me in changing the industry for the better. Hi, I'm your host, Lorraine Dallmeyer. I'm a chartered environmentalist, biologist and the CEO of award winning online organic cosmetic formulation school, Formula Botanica. We have thousands and thousands of students in over 190 countries around the world who study with us to become organic beauty formulators and entrepreneurs. Visit our website@formulabotanica.com to try our free online formulation course. And if you want to be the first person to know what's happening in greenview, make sure you subscribe to this podcast in your favorite podcast app. So in today's episode, I'm joined by Jan Chilvers, co founder of Cavallo the world's largest beauty ingredients platform. Cavallo's mission is to make the consumer goods industry more sustainable and transparent by helping over 5,000 brands across 145 countries better connect, collaborate and transact with the right suppliers. Prior to working at Cavallo, Jan worked at Monitor, Deloitte and Clarion on a variety of innovation and digital strategies focused on launching and scaling sustainable consumer centric solutions to market. Hi Jan, how are you doing today?

Very good. What about yourself, Lauren?

Yes, I'm good. And you know what, I am really pleased that we're tackling this topic on the podcast because this is a fascinating one and I think we're going to get a lot of people tuning in to listen to us about this. So we're going to talk about AI. So I guess the first question I have is firstly, can you introduce the company that you run and can you tell us a little bit about how you're using AI so far in your systems?

Absolutely. So my name is Jan, I'm the co founder of Covalo. Some of you may know it's under the old name of Canberry and we were initiated a bit like a within client with the ambition to become a Google for ingredients. And a bit more than two years ago we spun off Launch Covalo and more and more we evolve to becoming the data platform for the consumer good industry. And so basically what our mission is to help the brands launch transparent and sustainable product faster by helping them connect, collaborate and transact with the right supplies. Data has always been actually at the core of what we did already back in 2018. My very first project when I was working at Client was a digital formulator project. So it's funny to see how suddenly the topic of AI became probably a bit more prominent with ChatGPT and other LLM models. But I think it's a topic on which many companies have been working on for a while now.

So tell me a bit more about how you're currently using AI at Cavallo then.

We are using AI in different areas. So the very first one is data aggregation. So we have like little piece of software that we call crawlers that extract information on a daily basis from different data sources, including suppliers, websites. We always inform them of that approach of course, and those crawlers are able to identify relevant information and then input that information in our data model. Recently we were also able to improve that process by combining our existing crawler technology with OCR technology, which is optic character recognition technologies that allow us to extract information from static documents like PDF and now with all the large language models like ChatGPT, what is interesting is that you can also contextualize that information. So until now, we usually needed to have what we call an expert in the loop that review the information and make sure that the contextualization is correct. And now those large language models can do that in a more accurate manner. So, for example, when you have green, it's not always, or it used not to be so easy to know if green was the name of the product was green, the color was green, the sustainability claim. So now you can really contextualize that data and that allow us ultimately to aggregate data in a more effective but also accurate man. That's the first bit. And the second bit you really need to structure the information to make it comparable. So you need to see that the data aggregation and structuring is the foundation on which you can build intelligence. Once we structured that information, we are able, for example, to leverage our search engine, which is also like looking for the most accurate results. It's going to look at the different attributes, but also claims it can also identify alternatives, which ultimately means that people can do a very natural search without having to overthink. And we are doing all the work in the background to find synonyms, to find alternative and ultimately bring up the most accurate result. And the last bit is what we've been working on for the last two years is like a recommendation engine for sustainable alternatives. And this is basically trying to leverage all the information that we have on a supplier level, on the ingredient level, but also taking the location of the buyer and supplier to identify the product that ultimately will minimize carbon footprint, taking scope one, two and three into into consideration. Of course, you can use this in many different ways. So you could apply the same logic to optimize for price, optimize for lead time, optimize for performance, optimize for the stock that you have already. Now, networking capital is a big topic for many companies. So before you are buying new ingredients, sometimes you want to see what you can do with existing stock. Bottom line is you can do a lot of things that are really taking a certain burden away from R and D formulators, but also procurement people. The only bottleneck like in most of the AI project is really having access to the core data and making sure that the data that you are leveraging is correct. There is a saying like trash in, trash out.

Yes.

And that's why this initial data aggregation is definitely not the sexiest part of the process. But it's the most important One just.

To paint a picture for our listeners. I mean, you want to be the Google for ingredients. It sounds to me like you are already the Google for ingredients. You've obviously got all these crawlers out there, you're aggregating data. The optic character recognition. Sounds fascinating. How many ingredients have you got in your database at the moment? Roughly.

So today we have a bit more than 80,000 ingredients on the platform from about 1,000 suppliers. We decided to really have a vertical focus. So to focus now, really now, today on personal care. In the near future we want to focus on the consumer good industry. So basically all formulation driven businesses.

Amazing. Well, let's talk a little bit more about AI, because I think this is a topic that a lot of people are talking about at the moment. It's obviously really hot, it's fairly newish now for us, but I haven't seen many people talk about AI in relation to cosmetics. So what role do you think AI could play in relation to cosmetic product development, for instance, in the future?

I think when you have new technologies, a lot of people tend to be a bit scared. Okay, will it like threaten my job or my role? Actually we've never seen such a low unemployment and such a high need for people with the right expertise. So that's the first thing probably important to mention. AI and all technologies, their purpose is not to replace human, is to augment them, so to also provide them the capabilities and access to the right information at the right time to take the right decision. So for me, AI would be like really as a way to optimize the process and to for example, lower the time required to launch a new product or to bring it to market. So if you say that on average you need six months to four, five years for larger groups to launch a product to market, ultimately a trend is not lasting for ages. So if you have a trend that lasts for six, seven years and you take five years to launch your product, then you already lost a lot of the life expectancy of the product. You are also losing competitive advantage because you probably arrive on the market where already a lot of alternatives are existing. So ultimately the speed to market is in my view a key competitive advantage and something that many brands should be focusing on.

So how do you think a brand could use AI then in order to undertake cosmetic product development? Because people are probably listening to this thinking, what do I do? Do I just open ChatGPT and sort of go design me a serum? How does this work?

So I think like the very first part is what kind of product do you want to build and for whom? So it's identifying market opportunity, identifying target customers, and being able to validate your assumption as quickly as possible. Once you have your product brief, you need to find the right ingredients, and there we can come and support you to really help you find formulations, to find the right ingredients for it. And then you get to a phase where you need to make sure that the formulation is stable, that it's regulatory compliant, that it's match your retailer requirements. And this is something again that you need to have considered as early as possible into the process and then you launch the product. But there you again have topics like pricing. You might be able to justify a higher price when you're the first product in the market. As competition increase, you might want to have a certain flexibility. So you have a lot of dynamic pricing tools that exist and that can also help you optimize ultimately that price volume relationship in the market. So there are really a lot of tools that can help you really master each one of the steps that lead to launching a product today. I mean, someone told me earlier this week from, I think it was an Accenture study, that almost 50% of product launches are a failure. And that is just a massive investment that brands are doing so, working in very short iteration, getting continuous feedback, and rather launch few products really well, and then expand from there, then trying to boil the ocean, and then ultimately spend a lot of money on products for which there is probably not a real.

Demand, fair enough speed to market. Absolutely. I think that's a great benefit for many of the product designers out there. But what about formulation? Do you think AI could play a role in actually creating the formulas themselves as well?

So if you look today in the investment of AI, the data is very important. But the second thing, like now 80% of investment going into AI companies is going into infrastructure. So what you need to see is like, if you want to optimize for one variable, it's fairly easy. So let's say you have a formulation with 30 ingredients. You have, let's say, 25, 26 ingredients that are fixed, and then you want to optimize for the last four. The computational power for this is much lower than if you want to optimize for all 30 ingredients. And now let's assume that the number of ingredients that you are using is also open. And that's the biggest limitation today, AI, if the application is not really narrow and well defined, it can take months to get to the result. And it's not something that is so accurate. And that's where having formulation expertise to narrow down the scope of a search or an application, that's what can ultimately lead to the best result with AI. And that's why I'm talking about augmented formulators, because without that technical expertise, you are not able to narrow down the scope enough to get quick enough results from this kind of models.

That makes sense. Well then let me ask the follow up question to that. Do you think it's possible in the future to create a digital formulator without that technical expertise?

I mean, if you look far enough in the future, yes, sure. I mean, Google launched a quantum computer last year that was 300 million times faster than the previous computer.

Wow.

There are technologies out there that are coming, but again, the initial application are extremely broad. And then it's a matter of breaking down those technologies to verticals, topic segment specific applications. We mentioned Google in the past. Why did Google not address the chemical space? Why did they leave room for companies like Corvalo to exist? And I think it's the same with AI. You can do everything, but ultimately most companies need to really focus and that focus is leaving the room for smaller players and to also tap into those niche needs, those niche markets. And I think for indie brands that's also the case. You have probably some areas that are addressed by the very big brands, but you have other more niche market niche customers that you can build strong relationship to, that you can address much faster than anyone else. And that focus is again like a key because it allows you to be fast, it allows you to work in iteration and always be in touch with the market.

Well, let's talk about those indie brands for a minute because this podcast is listened to by many indie formulators and indie brands. I'm sure many of those indie formulators listening to this episode right now might be feeling slightly indignant because, and I've heard this a lot from people, they feel that the secret ingredient that they provide to their formulation is like a care and attention, almost like love and passion. And they feel, and I'm speaking for them here, but I would imagine they feel that that might be missing from an AI formulator as well as obviously the ability to choose the very best ingredients. So do you think that this might get overlooked in the process of using AI for form and product development?

Let me ask you a question, like, you put a lot of care, you pick the best ingredients. How do you make sure that your customers are aware of this and value it? Everyone puts care, but why does a consumer go with a specific brand rather than Another one. And the essence of it is that you are building a relationship, you're building a story that people love. You are building credibility. And this is what people buy. You know that elasticity, price elasticity of the utility line of cosmetic products. And you see that people trust more expensive products more than cheap cosmetic product, because they have the feeling that if it's good, it has to be expensive. And therefore they are not trusting cheap product. And that's deeply ingrained in the way consumers are looking at product. And that need of trusting the provider, especially with something you put on your skin, that you put in your body, is something that is particularly important. So I would say that if you are an indie brand founder, what I would be focusing on is really nurture the relationship with your customers, give them the feel that their opinion matter in a way that would not be the case with a larger brand. Truly understand what they are looking for, and then bring a product where you have the credibility and legitimacy to address that need. And that's something AI can't really do, even if AI. Let's take regulatory topic as an example, because I believe that this is probably one of the areas where the biggest disruption will happen. It's extremely well documented, which means AI model can analyze the data quite well. It becomes overwhelming for humans to absorb all those new regulations. And therefore we almost need AI models to always stay on top of regulations. So you have a lot of investment that are made in that direction. And models can get to 80%, 90% accuracy, but for the last 10%, you still need that human touch. I strongly believe that if today we were to launch the perfect model, consumers or brands would not trust the model at 100% and would still value to have that quality stamp on it. Would you go on ChatGPT and ask him to populate a contract for an acquisition or a very strategic project and then just take it and blindly sign it?

No way. No way.

Because the stake is too important and you don't feel probably equipped to assess if the result is good enough. It's not overlooking some aspect. So you would want to have someone with the expertise to review that contract at least. And that's a bit the same formulation. It's like you have a marketing person say, I want this wonderful product with an SPF 30, a 50% gross margin for decent demography, and then there is a formulation that pop out. How can that person truly assess the efficacy of the formulation, the quality of the formulation? You put it on your skin, you don't have an instant result. So it might feel good, might smell good, but you can't really assess beyond that if you don't have the technical expertise.

I like that. So what you're effectively saying is AI is currently missing humanity, which we need to bring to the process of formulation and product development. So let's talk about sustainability then. Do you think it's possible to use AI to make our formulations more sustainable?

Yes. Well, I think AI is very good, is to bring you the right information or the right knowledge at the right point in time. So if you want to make a product more sustainable, what are you looking at? Are you looking at transportation? Are you looking at the origin of the, is it based out of coal, oil, gas? Because the same ingredient with the same properties, depending on the origin, can have like 6x different carbon footprint. So that's where ultimately helping us focus on the right elements, helping us highlight the key information on ingredients and telling us what are the implications on the formulation. I think there's, there is definitely a use case, but I believe again, people want to understand finding the right information and saving search time and making sure people focus on the right thing. I think that's where AI will be focusing on, at least in the, I would say the next five to 10 years.

It is about finding that information. You're right. And as you said earlier, trash in, trash out could also mean greenwashing in, greenwashing out. So your answer is only going to be as good as the information that's out there and available, which is, will make it very interesting. So do you think that AI would be able to make the skill of formulation itself more accessible to people? Because in theory I could go to ChatGPT and say, formulate this for me and it would spit out a formulation. And I have tried this. And in fact, next week's episode that drops after this one will be with me challenging ChatGPT to formulate for me. But do you think formulation could become more accessible to everyone because of AI?

I think so, because, like, how does a typical course look like? You're the expert, but people usually enroll and they do three, six months, one year, sometimes longer to learn specific skills. And you have almost this information overload. So you have a concept, you understand it, but then you don't apply it for a while. And then when the day comes where you need that knowledge, then you need to have a refresh or you need to find that information again to make sure that you don't have just a vague memory, but you really understand what needs to be done. I think AI can help probably change the way we are learning. And it's okay, you have all the knowledge you need, but instead to have a linear approach to learning, you have a use case approach where you have an application and now you need point of information A, B and C and you bring those in that specific use case. I also think that this is the best way to really anchor knowledge because they always say like when you read something you have like 10%. When you tell something you have another percentage. But then the best way to learn something is to teach it or to apply it. And if you can bring the knowledge at the moment of application, that's probably where you get the best learning experience.

Okay, well, we'll see in next week's episode how well ChatGPT formulates for me. So I suppose my final question for you then is what do you think is the future for AI in the beauty industry?

I think probably the type of skill set that people need will evolve. I strongly believe that people will need to be more generalists or T shapes experts because the very deep knowledge will be accessible more easily. You need to be able to ask the right question or the right prompts. You need to be able to make sense of that information to see if that seems accurate or not. I think it will be more about knowing a bit of many things rather than knowing a lot of one thing and then not being able to cover the full breadth of knowledge required.

It may well help all of us move faster and in a more sustainable direction as well, by the sounds of things. Jan, thank you so much for joining me today. It's been a real pleasure talking to you. Where can people find out more about you and Covalo?

So we have the covalo.com website. The sign up is for free when you have a professional email address. We also intend to be on different fairs. We launched last year partnership with Real Exhibition so we have launched together the In Cosmetic Discover platform. Some news will come around this fairly soon and then if you want to just Simply connect on LinkedIn or per email, you can easily find me on LinkedIn or at yann.chiverso.com Amazing.

Thank you so much for joining me today and I can't wait to see what you do with AI moving forward.

Thank you very much. Have a lovely day, Lauren.

Thank you. I would love to hear your thoughts on this episode. How do you feel about AI? How do you feel about AI designing your formulations or your cosmetic products? Tell me what you think. Leave us a comment on our social channels as both the Formula Botanica team and I love hearing from you. And if you want to hear more about some of the topics we've covered, please do delve back into the podcast archives. Go and listen to episode 79, in which I interviewed Jan's business partner at Cavallo, Timo von Bargain, about the challenges of sourcing natural ingredients for indie formulators. Or head to episode 131 in which I look at another futuristic form of formulation biotechnology and look at its potential for us to perhaps even grow our own cosmetics in the future and then hand them on like sourdough starters to our friends and family. And also stay tuned for next week's episode in which I'm going to challenge ChatGPT to formulate for me and I'll be sharing my results with you in my Opinion piece which will come out next week. There's a wealth of information in our old episodes and I do encourage you to go back and listen to them. So thank you for joining Jan and I for this latest episode of Green Beauty Conversations. I really hope you enjoyed it. If you did, please do leave us a five star review so that other people can enjoy these conversations too. Make sure you subscribe to our podcast on Apple Podcasts or Spotify or whatever your favorite podcast app is, and stay tuned for the next episode. Follow Formula Botanic at Formula Botanica on Facebook, Instagram, YouTube, Tik Tok, Twitter or LinkedIn and come and give me a follow at Lorraine Dharma on Instagram. Visit our website@formulabotanica.com and sign up for our free online formulation course today.