EP245. Could AI make your skincare more eco-friendly?

It won't have escaped you that AI is taking off everywhere and I'm starting to see the beauty industry increasingly use AI in more places as well. Now picture this. Picture a tool that can analyze every single ingredient in your favourite formulations, find eco friendly alternatives at the snap of a finger, and even spot untapped benefits in unexpected ingredients. Does this scenario sound slightly futuristic? Well, believe it or not, it's not. And we're about to find in this episode how it's already happening. So imagine a future where creating a sustainable or eco friendly beauty product doesn't mean reinventing the wheel, but instead uses data driven insights to reimagine what's possible. This technology could mean the end of trial and error formulations and it could replace them with smarter science backed swaps that deliver effective results and a reduced environmental impact. So this is a fascinating topic and it's amazing to see what's possible because AI has the potential to do great things for cosmetic formulation. That's also why I want us to look more closely at the role of AI in cosmetics in this episode. Can AI be trusted to guide us safely? Will it one day be able to take over the art and science of formulation itself? Can AI make your skincare more eco friendly? Settle in for this whole episode to explore all of this and and more. Welcome to Green Beauty Conversations in which I challenge the beauty industry to become more sustainable, champion, independent and green beauty and encourage you to join me in changing the beauty industry for the better. 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 tens of thousands of students in over 190 countries around the world who study with us to become organic beauty funds, 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 green beauty, make sure you subscribe to this podcast in your favorite podcast app. So in today's episode I'm joined by Abdu Kane, the CEO and VP of AI at Inflows AI, a UC Berkeley backed startup developing an all in one co pilot for cosmetic chemists who can then expedite sustainability and compliance checks. Abdou is a former Lead Data Scientist for R and D labs at l' Oreal Research, creating data driven solutions to empower formulation chemists and product evaluation teams. He has a Master's in Materials physics and a PhD in applied mathematics from the French Atomic Energy Agency. And Abdou has over a decade of experience using AI to drive engineering efficiency. Hi, Abdu, how are you doing?

I'm doing great, Lauren. Thanks for having me today.

It's awesome to have you here. I'm really looking forward to this interview. I'm so interested in what you do, and I know our listeners will be as well. So can we please start with some introductions? Please introduce yourself and your business to us.

So I'm Abdoul Khan. I'm a PhD in AI, and I have been working for years at L', Oreal leading the AI strategy for labs where I spent time, you know, at the chemist bench, you know, witnessing how they were creating products, what were they border next? And going back to our innovation lab and trying to create AI tools to import them.

Amazing. And now what business have you created?

So after leaving l', Oreal, we decided to create Inflows. And the aim of Inflows is how do we create AI tools for formulation chemists to help them accelerate the way that they create? You know, they create faster product in a more sustainable way. So we have started the company in Paris and we got investment from the UC Berkeley, and that's how we moved to California and continue to spread out.

In the U.S. well, I can see that there's definite demand for it. I mean, the beauty industry keeps talking about AI, but I hadn't seen all that much come up onto the scene. And then suddenly I heard about Inflows AI and I thought, this sounds absolutely fascinating. We've got to get these guys on the podcast. So obviously thrilled that you're here. So why did you create inflows?

You know, when I was at l', Oreal, I come to realize that formulation chemistry is not only about beauty product, it is all about chemical product. So what we did at that time was amazing. We were able to kind of speed up the way that we were formulating any kind of cosmetic product. But I came to realize that in the food industry, in the pharmaceutical industry, coating, you know, and you name it, all those industries, they have the very same problem. And then I just wanted to kind of have a much more broader impact by creating a company that can tackle all these kind of challenges in those industries that are very essential to human.

Yeah, yeah, I totally get it. Well, I mean, obviously the experience you have, I think, lends itself very well to putting a company like this together. So can you walk us through how it all works? Can you maybe walk us through a real world or a dummy example of how a brand might Use Inflow's AI then to improve an existing product. Sustainability. How does it all work?

Absolutely. Whenever you have a product in the market or a product that you about to launch, the main challenge that you'll be having is how do I select the best ingredients for my customers and how do I make sure that supply and resilient. So whenever I have to swap or whenever I have a shortage on one specific ingredient, how do I make sure that I can replace that very specific ingredient without messing up with the whole formula characteristic? And those are the challenges that today most of the time we use the expertise of the formulation chemist or we rely on some kind of claim from the suppliers and we ask for sample and we try it out. And it's a very time consuming process because you do a lot of batch try to figure out. Some tests are very long, like the stability test or the microbiology tests are very long to perform. So what we have created is a way to say if we can rely on some historical data or some trials that you did on the lab and create an AI system on the top of those data to learn the information between the input and the output and the outcome you're getting out of formula, we can very quickly create some, what we call digital twins. And the digital twins is how do you digitalize the physical experiment into a computer so you can run much more faster testing and then just simulate instead of doing actual experiment. And that's how we do in flows. So we enable AI tools in a very easy way for formulation chemists to just kind of create those digital twins and play in the digital world before going back to the physical texting.

So how does that fit with sustainability then?

Yeah, so if you say, for example, let's say sustainability is a broad kind of characteristic where you have several criteria. Let's say for example, like the energy consumption, if you are using a product that you need to kind of keep to have a kind of, you know, a liquid that you have to mix now you're consuming much more energy. So if you are able to swap this ingredient with one which is kind of liquid at ambient temperature, you can improve by the way the sustainability because you kind of consuming less energy. This is one of the way to do it. The second thing is about the biodegradability. If you have an ingredient, if after use is going to remain into the soil, into the water and aquatic ecosystem for a long time, you want to swap it out with something that's going to degrade much more easily. And then the challenge would be how do I make sure that my swap will be kind of stable within my formula and will degrade well after use. Those are kind of a way that you can enable chemistry to kind of figure out what other alternatives you can use for.

Okay, so the AI then has the ability, I presume, to trawl through databases of ingredients, then to look at specific characteristics to see what might be better suited for the sustainability of the formulation. Is that correct?

Yeah, absolutely. So for natural ingredients, which are compounds of several sub ingredients, it's kind of less easy. But when you have a pure substance, where you have access to the, the molecular formula so the chemical structure, you can pretty much guess what kind of reaction will happen in the nature once you release that kind of ingredients, because you can anticipate what kind of chemistry would be involved. And that's how we kind of find out what are the ingredients that may be stable in the formula and have very high likely kind of probability to degrade once it's in the nature.

Okay, I get it. So how long did it take you to build all of this? Because it sounds like there must be an enormous amount of data set behind the tool that you've created. So.

Yeah, so in general, when you talk about AI and data analytics, people think that you need tons of data to make things happen. But the challenge is that if you are in the cosmetic and the chemistry world, you don't have that enormous database. Right. You get started with Excel, spreadsheets, some kind of data here and there. And that's what's the kind of challenge that we use to deal with our clients. So our system is not kind of heavy, it's not demanding on high volume of data because we can get started with a very small sample of data from three data points, from five data point. You're not trying to kind of explain the whole world of chemistry. You're trying to, in a very narrow scope, which is your formula, to figure out among the alternative, the list that you have, which one is the most suitable. So it's more like a matching rather than trying to kind of have a broad explanation of what's happening.

I get it. So I presume that this, the tool you've created there, must help brands get new products to market a lot quicker then.

Yeah. So right now you have some guys are very skeptical to AI, so you have to have some kind of change of management to not let people think that AI will replace them, but more empower them. And that's the kind of challenge that you kind of were dealing with. In general, it's more like less about the AI, but more about the change management and the mindset that you need to adopt to say, okay, AI is like a fast car, so if you jump in, you still be the driver, but in the most faster way. And that's the kind of work we kind of doing right now to convince people to use it. But whenever they tried using it, they just kind of love it.

Yeah, yeah, I get. I love the analogy, by the way, of the fast car. I've never heard anyone put it like that, but that's completely right. It's also how we use it. So how are you seeing. Obviously, you've launched inflows AI, you've got initial customers in there. How are these first customers starting to use the tool? What sort of tasks are they using it to complete?

So we have two main topics that we address. The first one is about raw material replacement for several reasons. It can be the challenge, it can be the compliance, it can be the sustainability aspect, but at the end of the day, it's just replacing one ingredient. So we have a tool which is called Swap Flow, where you can input any chassis that you're working in. So we're going to control because whenever you have to swap ingredient, you have to be aware about the context of the use, because swapping one specific principally in a shampoo will not be the same in a mascara. So you have to be aware about that. So the tool is very easy to use. You just put the list, your chassis, the list of ingredients you'll be using for your product, and then you tell the one do you want to swap out, and you run the system. The system will provide you a couple of alternatives that might be suitable for that. See, this is one of the use case, and the second use case is you have one ingredient that you want to use, but you don't know precisely what is the amount, the exact amount you should use in your formula. So it's like kind of troubleshooting the recipe because you want to fix it down. So we have another tool called Design Flow, where you put the list of the trials, you put your objectives, and then you let the AI work to figure out what would be the best amount of this ingredient to kind of make all your properties green.

Amazing. So you've got sparks flows and design flows, Is that correct?

I say swap, like swapping something. Swap flow.

Yeah.

And the other one is Design Flow because you're designing experiments. So we call it Design Flow.

So do you plan to launch any other flows then?

Yeah, we're very exciting we have just released about one week ago what we called Formula Builder. And the Formula Builder is like a studio. Whenever you want to make a video, you have a studio where you edit and you all do things for the video. Like imagine you have a digital studio where you want to make a formula. You say, hey, I have my benchmark, I have the list of Inkey, and then I want to now create my formula. So you just input the list of inky, you tell us what are your preferred suppliers or what are the preferred trade names that you want to use, and the system work out to create a first prototype of what the formula should look like with the percentage, the price and all the things that you need to know to kind of create the first version of your product. Because most of the time when you create a product, it's a very challenging journey to kind of figure out what are the prices of your ingredients. So you have to ask to suppliers, you get a call and then sometimes when you get the price, you don't have any kind of management system to store there. So what we're creating is by having a way to store the prices and all the information related to ingredients. Whenever you get started for a new product, you have to hit the list of ingredients and the system will just pre fill from all your historical knowledge, the formula, and you're good to go.

Wow. So it sounds like there are dozens of different use cases that you could probably put in place with using AI. So obviously you have a lot of work ahead of you in creating more facets to the business. But coming back to what you've built already, what were the biggest challenges that you faced in translating lab based cosmetic chemistry into an AI framework?

The main challenge is that whenever you deal with what we say, natural language processing, like the AI for just natural language, we know pretty much that if you want to swap a word in a sentence, you can pretty much figure out what are the synonyms, right? But when you deal with chemistry, you say, okay, this is a novel ingredient from a plant based ingredient out of, let's say Australia. And then you get the ingredient, but you don't know what it is inside, Right. So it's very difficult to kind of characterize the ingredients and know precisely what are the actives inside. So there's a lot of work that needs to happen. Most of the time this work is not kind of done 100%. So you have an ingredient, you have the name, but you don't fully know it. And you know, maybe one function out of maybe 10 functions that this Ingredient may perform. So that's where you can see that one ingredient is a multifunctional. Sometimes it's used for one function, another one, another one. And maybe five years later somebody discovered that there's a new function that is hidden in the ingredients. And that's the main challenge we're going after. And I think that in pharmaceuticals and the food industry, they have the same challenge. It's about how do you characterize the ingredients to make sure you have your 360 view of what all applications that you may do you can do with that ingredients.

Yeah, I totally get it. So do you think that AI will ever be able to formulate for us?

Quite frankly, I know that people might not like what I would say, but I think that it's just a matter of time. Because formulating is about applying science on some materials to make a finished product. And the science today for human, if you want to learn that science, you read books, right? The AI system is able to read books. You have to understand books, you have to understand chemistry. AI is able to understand chemistry. So I think at some point AI will be able to formulate. But now there is a kind of sensorial attributes that a human need to kind of evaluate where maybe AI may struggle. But I think that the combination between human and AI would be the future.

Still coming back to that fast car analogy where you're going quickly but you still need someone to drive it. Because you're quite right. I mean, AI can't smell yet. Maybe that will be possible in the future. Who knows? AI doesn't have skin yet. Who knows what terrifying future awaits for us. What do you think are the biggest risks in terms of using AI then in terms of cosmetic formulation?

I mean, one of the risks might be maybe if we skip the evaluation and then the safety assessment of product and we go straight from an AI idea to applying this product on skin or a human body. Now you can create some harmful chemical effects and then it's going to be difficult to kind of say who is the responsible of those kind of harm.

And this is where those final checks and balances obviously from a human is so incredibly important. And on the flip side, then what do you think the opportunities are?

The opportunity would be, I think that accelerated innovation having way much better product in the market, more sustainable ones, cheaper product. Because we will able to kind of figure out how to make products cheaper with less ingredients. We have less complex formulas. Because right now when you see some formula, they have up to 50 ingredients. I don't know, think that you need 50 ingredients for shampoo, all those kind of things. There are historical kind of chassis that we have been kind of using over years and didn't change them because we didn't want it to kind of mess a little bit the formula. But I think that AI will kind of speed up the way that we create safer and more sustainable product for human.

So how does Inflow's AI then address regulatory compliance? Because you were talking about safety, obviously, and obviously regulatory compliance requirements are different across different global markets. There are variants in ingredient restrictions, sustainability criteria. So how does the tool you created address that?

So today we don't have a tool specific to compliance, but we have some R and D or what we call chat with unstructured documents because most of the regulation is on PDF files on website, some document which do not have a specific format from a country to another.

Right.

So there's a huge task right now in the AI space where people are trying to make AI understand documents that doesn't have a structure. So we kind of rely on those technology and the improvement that's going to happen on that space to say once those technologies are very mature, how can we plug them into where the regulation is coming out to kind of understand what is about? And if we are able to understand with a machine, what is the regulation now? We can apply them systematically into formulas, but we're not yet there.

No, I get it. I mean, there are a huge number of variables in there as well. So then in terms of transparency, moving from compliance to transparency, what role do you see AI playing in consumer transparency, such as helping beauty brands disclose ingredient sourcing or sustainability data?

I mean, there will be still a challenge about confidentiality where you source your ingredients and what are the kind of, you know, secret sauce that you use to make your product. But I feel like AI will help consumer understand better what is inside the formula. From the Inke list, if we assume that the brands are putting all the ingredients in the entities. I'm smiling when I say that, but I assume that brands are doing so.

I hope so.

Yeah.

They're supposed to. Yeah.

So if you assume that right now you don't need to have this kind of chemical knowledge, chemistry knowledge and to kind of understand what is inside your product, you can just leverage AI to help you out understand if this product best for my. The best for my skin, it is for to tailor for what my needs, etc. So it's going to get you a little bit more understanding about how to pick the right product. Just leveraging AI to help you like a guide to navigate through a formulation space and then formula space.

So that's where we are right now. Where do you see the future of AI in terms of cosmetics? Because you're obviously looking at what the potential is as well. Imagine five, ten years down the line. What do you think the cosmetics industry might be doing with AI at that point?

I mean we see huge trends on what we call computer vision where you use a camera or mirror to kind of look at you and then provide a diagnostic and after that provide you a way or cosmetic routine. Maybe that's the future where you can just stand up in front of your mirror and just have a, you know, scan of your skin, your hair, etc. And then the system will kind of from in the back in the restroom, just kind of craft a personalized formula for your skin, for your hair and just make it happen so that you have the most tailored routine for yourself.

Wow, that sounds futuristic. I can't quite picture that one at the moment. But who knows what might be possible, I suppose. Let's come back to Inflows for a minute. Tell me more about Inflows. Is it available for any formulators or what sort of entry level capacities can Inflows handle?

Yeah, so today we have a premium version where everybody can access to, to kind of test a couple of modules. Like if you have a, let's say you have a product and you want to know what are the closest product in terms of chemistry in the market, you can use one, what we call the search flow. If you want to swap out ingredients, you want to, I want a shampoo without, let's say given ingredients, you can just filter out this ingredient and have other products in the market and then you don't need to pay to access to that. So we made just publicly available for everybody who want to test it out. And then the name of the site is co Chemist, like co CEO chemist. So it's a co chemist working along your site to just help you out, navigate the formulation and the product landscape. And then we have a couple of modules that you need to subscribe. Like our business model is a subscription where you have to subscribe to be able to use those modules and they are very easy to use. So we in general onboard people, by doing a 10 minute, 15 minute training, they get started that you show them the use cases and then they're good to go in general and they can use it by themselves.

Wow. Well, that's really exciting. And obviously we'll include all of the links in Our show notes as well. So how do you envision Inflow's AI's platform evolving over the next three to five years? What do you think are the next big steps for you guys?

I mean, today, if you ask the founder and even myself, what is our big dream is to create what we call the all in one copilot for cosmetic chemists. Because right now, where you are in a journey of developing a cosmetic product, you have several tools. One for compliance, one for supply chain, one for pricing. And then you have to jump in and out from one tool to another one. Right. So what we want to do is create under the same roof the all in one platform where you will have all the models that you need seamlessly connected from the workflow, from the idea of creating a product to the finished product and the trustability that you want to have on the scale up, all under the same platform. So that's why we kind of creating the building blocks for this kind of journey. And now we're assembling the building blocks to create the story exciting.

Well, obviously big vision to create the copilot for every single cosmetic chemist. And do you think that the big beauty brands are starting to embrace AI more than the smaller ones at the moment?

I mean, that's the beauty of AI. Today you can be a team of 5 people, 10 people, and competing with the biggest brand of the world. Because AI is not something that because you are a thousand people, you're going to be more efficient. It's about the mindset of the people using it. And I feel like today AI is kind of putting more equity in the competition because now with AI you can compete with a team that is maybe 10 step more people than you because you're kind of leveraging AI if they are not doing so.

Interesting. So it's going to give smaller brands perhaps a competitive advantage. From what you're seeing. Are you seeing the bigger brands embrace this quicker though? Because I would imagine that they might be earlier to the table in terms of AI or not.

I mean, we'll say some, but big brand they have. The challenge with the big brand is that they have a lot of product in the market that are very successful and they are very kind of, I wouldn't say afraid, but they kind of very hesitating on touching those products because they don't want a consumer feel that something had changed and that's why they don't want, they don't leverage that much AI today. They more like rely still on human qualification, human testing to make sure that they can substantiate all their claims before go to market. And then I think that that's where maybe for smaller brand that might be more kind of prone to take risk to say, hey, you know what? We don't have all these thousands of brand and thousand people. We have small product. Maybe we are in a niche, we can leverage AI and then just navigate through the blue ocean.

Interesting. It's interesting because here at Formula Botanica we always say you don't need to be a chemist to formulate. But I suppose there's an iteration here for you guys which is you can have AI to formulate as well and to see if people actually buy into that message. Because it sounds to me like this might be the future for certain cosmetic formulators out there. Abdul, it's been an absolute pleasure talking to you today. I think what you've built is absolutely fascinating. If our listeners are interested in finding out more, where can they find out more about you?

You have my LinkedIn, so I will just share with you my. You know, I'm very open on LinkedIn and you have my email. I can just send it over as well. It's akaneflosai.com so I can just share with you so you can put it in the notes. So I would be happy to kind of connect with any folk in your community to kind of see how we can work together.

And the website obviously is inflowsai.com.

Absolutely.

And what was the other one? Cochemist.com?

Cochemist AI.

Cochemist AI. We will put all those links in the show notes. Abdul, it's been an absolute pleasure talking to you today. The best of luck with Inflows AI. I can't wait to see where you take it. Thank you so much for joining me.

Yeah, thank you so much for having me today, even my pleasure.

I would love to hear your thoughts on this episode. How do you feel about AI coming into the world of cosmetics? How could AI help formulation? 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 covered today, please do delve back into the podcast archives. Some great Examples include episode 183, in which I asked Jan Chilvers of Cavallo to talk me through the way that he sees AI changing the beauty industry. Or episode 184, in which I asked ChatGPT to formulate for me. That was already recorded a year ago and let me tell you, spoiler alert, it was pretty disastrous. To be fair, I haven't tried it since, so I should probably have another go to see if AI has improved at all. There's a wealth of information in our old episodes and I do encourage you to go back and listen to them. Thank you for joining Abdu 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 the podcast on Apple Podcasts or Spotify or whatever your favorite podcast app is, and stay tuned for the next episode. Follow Formula Botanica Formula Botanica on Facebook, Instagram, YouTube, TikTok, Pinterest or LinkedIn and come and give me a follow at Lorraine Dmeyer on Instagram and LinkedIn. Visit our website at formula botanica.com and sign up for our free online Formulation course today. Sam.