Podcast Image: E16: GenAI's Impact on Walmart, L'Oréal, GE, VW, and Other Examples: What Works and What Doesn't

E16: GenAI's Impact on Walmart, L'Oréal, GE, VW, and Other Examples: What Works and What Doesn't

Chris Rod Max dive deep into how AI is transforming major corporations, from Walmart's e-commerce surge to L'Oreal's 10-petabyte beauty database. They break down Duolingo's AI-driven 45% revenue increase and explore the potential of AI in healthcare. Plus, a look at Volkswagen's ChatGPT integration and analysis of the Forbes AI 50 list. Is the US AI funding advantage leaving the rest of the world behind?

Host

Chris Wang

AI Innovation and Strategy Expert, CXC Innovation

Guests

Rod Rivera

🇬🇧 Chapter

Max Tee

VC Expert, AI Investor, BNY Mellon

E16: GenAI's Impact on Walmart, L'Oréal, GE, VW, and Other Examples: What Works and What Doesn't

In this episode, the hosts discuss the integration of AI into bigger corporations and explore use cases and results. They highlight Walmart's deployment of LLMs and generative AI for product categorization and customer experience improvement. They also discuss L'Oreal's use of AI to analyze skin and provide personalized beauty recommendations. The hosts touch on the challenges of data limitations and the role of startups in providing AI solutions. They also mention GE Healthcare's AI initiatives and the potential of AI in the healthcare sector. In this conversation, Chris, Rod, and Max discuss various use cases of AI in different industries. They explore topics such as AI-powered voice assistants in cars, the revenue growth of companies like Duolingo through AI content, and the emergence of AI startups. The conversation highlights the importance of data as the foundation for AI implementation and the need for companies to invest in AI to stay competitive. They also discuss the challenges of retention and the potential of AI to transform industries. The conversation concludes with a discussion on the funding landscape for AI startups and the importance of building holistic solutions.

Takeaways

  • AI integration in bigger corporations can lead to cost reduction and revenue growth.
  • AI can transform customer experiences and improve operational efficiency.
  • Data limitations and privacy regulations are challenges in AI implementation.
  • Startups can provide customized AI solutions for specific use cases.
  • AI has significant potential in the healthcare sector for diagnostics, personalized treatments, and data analysis. Data is the foundation for AI implementation and can have a significant impact on both the bottom line and top line of businesses.
  • Companies need to invest in AI to stay competitive and reap the benefits of AI technology.
  • AI can transform industries by improving customer experiences, increasing efficiency, and enabling personalized solutions.
  • Retention is a challenge for companies, and AI can help keep users engaged and provide personalized learning paths.
  • The funding landscape for AI startups is primarily concentrated in the US, which gives them a competitive advantage.
  • Startups should focus on building holistic solutions that address the specific needs of enterprises and collaborate with task forces to develop use cases.

Episode Transcript

Introduction

Chris: Welcome everyone to the Chris Rod Max show, where we discuss the latest trends and news in the AI world and interpret their second and third-order consequences. I'm glad to be back. Hi Max, hi Rod, how are you both today?

Rod: I'm doing great, it's great to be here.

Max: Good to be here as well.

Chris: Excellent. For those watching on YouTube, you'll notice we coordinated our wardrobe a bit - we're all in black today. Our main topic is quite exciting: we'll be diving deeper into AI use cases integrated into larger corporations.

Reflecting on AI's Impact

It's been a wild 18 months since OpenAI released ChatGPT in late November 2022. I think it's a good time to reflect on what companies have accomplished in terms of AI use cases and their results. This will be crucial going forward.

We've had episodes discussing whether we're deflating the big hype around AI, whether it's worth exploring, and we've seen AI company stocks tumble and recover. The key question is: Is there real potential to commercialize AI, and can larger companies use it to leverage their bottom line or increase their top line? That's what I want to explore with you both today.

Innovation Hubs and Promising AI Startups

Secondly, I'd like to look at how some enterprises are building innovation hubs around AI. They have working groups and are in talks with various startups and larger tech companies. Forbes recently released a list of 50 AI startups worth noting, and I'd like to discuss the use cases we're seeing there. Perhaps you've tried some of these startups and can share your experiences. Does that sound good? Shall we dive in?

Max: Let's go.

Walmart's AI Integration

Chris: What prompted me to look into this topic more closely was Walmart's earnings call a few days ago. For the first time in an earnings call, a company of that size mentioned the deployment of LLMs and generative AI. They've used it in two or three key ways:

  1. Training a model with their 850 million pieces of data to better categorize products. They mentioned this is about 100 times more efficient than their current headcount for that type of job.
  2. Their e-commerce business has increased 26% year-on-year, which they largely attribute to using AI.
  3. They've developed an AI-assisted app that enhances the shopping experience. For example, if you're planning a princess-themed party for a 4-year-old, instead of wandering through aisles, the app provides a list of everything you'd need. This increases impulse purchasing.
  4. The app can also populate a cart with everyday items typically needed by families like yours.

There are many more backend use cases where AI helps employees do their jobs better. In total, Walmart estimates that major digital transformation projects now only require 1% of the staff they used to, thanks to AI automation and workflow improvements.

What's your reaction to this? Rod, would you like to start?

AI: Adapt or Die

Rod: We're seeing two things across industries and companies. First, they're rolling out AI initiatives. Second, in their reports and earnings calls, they're citing AI as their number one threat that could endanger their business. It's a bit of an "adapt or die" situation. If you don't aggressively pursue new AI initiatives to cut costs and improve efficiency, your competitors or new AI-native entrants might roll over your business.

Chris: Interesting perspective. Max, what are your thoughts?

AI Mentions in Earnings Calls

Max: A couple of things. Let's start with the earnings calls. I quickly Googled some facts: Out of the 199 companies on the S&P 500, AI is mentioned an average of 11 times per call. The median is about 5 times. For 12 S&P companies like Meta, NVIDIA, and Microsoft, it's mentioned at least 50 times.

It feels like AI is going to be a very common word across all earnings calls going forward. There's also the question of use cases. We've talked about how use cases were somewhat lacking in some larger corporates previously, but it's good to see that Walmart and others have been announcing more use cases, for example during CES.

Walmart's AI-Enhanced Shopping Experience

What I find interesting about Walmart's offering is that it's not just customer-facing. They're also focusing on internal operational efficiency and assistant-like solutions. The search function they're developing is particularly intriguing. Imagine walking into Walmart and saying, "I'm having a princess-themed party for my three-year-old daughter. What do I need to buy?"

Without this AI assistance, you'd typically go to Google, search for party ideas, get a list, go to different websites to compare prices, and then finally make your purchases. With Walmart's AI, the impulse buying potential increases significantly. You might be halfway down an aisle and suddenly remember your daughter's birthday is next week. The AI could suggest everything you need right then and there. It's about catching the customer at just the right time and place for that purchase decision.

Chris: Definitely. Rod, you pointed out that corporates seem a bit schizophrenic about AI - concerned about risks and data breaches on one hand, but trying to adopt AI to increase their top line or improve their bottom line on the other. And Max, you're highlighting how the entire consumer experience is changing, even for mundane daily tasks.

AI's Impact on Costs and Quality

Coming back to some of our past conversations, we've discussed examples like Klarna, IKEA, and McDonald's (which had a somewhat failed use case with IBM's Watson). One key takeaway is that AI really does have the ability to decrease costs. For instance, Klarna saw their customer support team reduced by a certain percentage, but the quality of service delivery also increased. So you have this interesting effect where you can decrease costs while also increasing top-line revenue.

L'Oreal's AI Integration

Today, we also want to talk about how companies are transforming the customer experience. Walmart is one example, and another is L'Oreal, which uses AI technologies to analyze skin and provide a better personalized experience for the end-user. Max, in our previous conversation, you mentioned some thoughts on L'Oreal's approach. Can you expand on that?

L'Oreal's Data-Driven Beauty Solutions

Max: Absolutely. I'm particularly interested in the overall experience L'Oreal is creating. They're utilizing 10 petabytes of data on skin conditions, hair formation, and other beauty-related information to help users find the right products and even personalize formulations to achieve individual goals.

This is fascinating because when the internet first became widespread, there were several tech solutions trying to use algorithms to match users with skincare products. Now, with the amount of data L'Oreal holds, they can transform the overall experience in a way that no one else can match.

There's also the possibility for them to unlock this capability for other brands to use. It's almost like the fintech equivalent of building a core banking system - like Monzo or Starling, who built their core banking in-house while running a front-end to compete with larger banks, but can still sell the backend to their direct competitors.

The overall utilization of data, AI models, and the latest LLMs to create a better experience and more personalized output for end consumers is super interesting. I think L'Oreal isn't exaggerating when they say they've dedicated their entire life to beauty and have the world's richest database on the subject. The implications for smaller boutique beauty shops could be significant - how can they compete in selling the right products when AI can potentially replace the role of beauty advisors?

Chris: Definitely. I think the use cases L'Oreal mentions, like analyzing skin and giving makeup tips, aren't new ideas. It's almost like they're bringing back old concepts, but now they have the technology to actually make it work. And your point about them having this comprehensive beauty database that they can now harness is crucial.

Rod, I want to turn to you on this. In our previous conversations, we talked about data limitations - that all the models have already absorbed all the public data, and it's really about proprietary enterprise data now. What's your view on this?

The Value of Proprietary Data

Rod: Well, in L'Oreal's case, they claim to have a large proprietary database they can tap into. This is something we see in large organizations - recurring teams come up every time there's a new innovation wave, and leaders say, "Why don't we try this again?" Indeed, many of the things we're seeing now, like using generative AI to fill in product details instead of having a human type them, have been goals for e-commerce companies for at least a decade.

The technology is finally there to produce good results. The same goes for L'Oreal's case. We should also remember that cosmetic companies are, in the end, also pharmaceutical companies. They have not only customer data but also large databases of chemical compounds and ingredient combinations for new treatments and products.

By structuring this data and creating internal tools, a product designer or marketer can say, "I want to come up with a new product targeting this demographic, for this budget, with this impact." The brainstorming becomes much more efficient because L'Oreal already has this database with all available chemical compounds, products, and prototypes. Together with the marketer, it's possible to brainstorm and create portfolios of skincare products that fit the desired demographics much more efficiently.

Chris: Definitely. So it's not only the user experience that's affected, but it also helps in developing new products. This is something that Reckitt, another consumer goods company, has also mentioned. They're building task forces or innovation hub centers around AI.

AI in Product Development

They claim that AI allows them to develop new product concepts up to 60% faster and adapt and localize products 30% faster. Max, in your day-to-day work or in your community, do you see other companies creating these AI task forces? What's your take on the claim that they're really better or faster when it comes to new product development?

Max: Let me address the latter point first, then the former. The faster product development reminds me of Temu or Shein. What they do is run a lot of marketing upfront, then they have direct connections to producers in China to rapidly produce fast fashion products. Now we're taking the same idea and applying it not just to fashion, but to skincare and any other type of product you could potentially sell.

I think we'll really see this model work well. The question then becomes how the economics will work. From a consumer perspective, we'll probably be able to save more money. From a global resource perspective, it might lead to more efficient use of resources by finding better, more targeted products that people are more likely to spend money on, rather than buying solutions that only solve 80% of the problem.

As for AI task forces and use cases, what I'm seeing, based on my observations, is that many organizations already have some of these set up. Especially around the 2015-2016 era of data, ML, and analytics, they set up a lot of these teams. Now with generative AI, they're shifting or evolving to incorporate more use cases.

What's interesting here is that it's really a test of how good the underlying data layers are. If they're not up to par, companies will have to go back and solve that problem first. I always remember if a chief data officer comes in and says we need to spend X amount to build a good data layer, a business person will always try to constrain the budget, simply because they don't want to spend so much without immediate returns. Many senior leaders in big corporates won't be there for long enough to see long-term payoffs. But now we can really see the consequences of kicking that can down the road.

To some extent, the flaws in big use cases might be attributed to data connectivity issues and the skills of their internal talent. So I think the task forces are there, but they're often evolving from what has been done before.

Chris: Yeah, and maybe Rod, coming from more of the startup world - I'm sure you also speak to several companies and try to sell some AI features and services. What's your impression? Are these newly formed AI organizations open to startups and trying out startup features and solutions? Or do you think they'd rather partner with the OpenAIs, the Microsofts, the bigger players in the market?

Startups vs. Big Players in AI Partnerships

Rod: There's a bit of everything. Of course, the majority tends to adopt what OpenAI offers, or what the big players like Anthropic, Cohere, and so on provide. This is usually the first go-to, simply because of the ecosystem. You have to remember that, in the end, companies are made up of people. As a result, the team will gravitate towards technologies, providers, and vendors they've heard of, which tend to be the big names rather than small startups.

Definitely, when a company starts an AI initiative, when they want to do something AI-related, the very first thing they'll do is work with OpenAI. It's only after they realize they want to do something a bit different, or they hit some limitations, or they don't like certain aspects, that they start exploring what else is in the market - what else can be provided by startups or smaller companies that aren't these big names.

Chris: So basically, you're saying people normally go for the gold standard, and then they realize the gold standard isn't really adapted to their personal, customized needs. And then either they customize it themselves, or they find an alternative solution. And this is really where startups can also get in. Is that what you're saying?

Rod: Exactly. And this is also, as with every IT decision, a case of "nobody ever got fired for buying IBM." It's the same here. The executives have heard of ChatGPT and OpenAI. If you as an IT leader say, "We'll be using OpenAI technology," then this resonates with them. Whereas if you say, "We'll be exploring some unproven startup in this space," then it's a harder sell. There are always chances, but as a startup, you have to fight against the brand power and recognition that these established AI players have.

Chris: Definitely. Coming back to these task force ideas and some of these partnerships, I think GE Healthcare has also dabbled in AI. They've started with an AI task force to explore and understand how AI can help them improve. They've launched a couple of AI apps for their clinicians to help them make better decisions based on the data they feed the model.

GE Healthcare and AWS Partnership

They just announced about a month ago that they're partnering with AWS to explore more AI services and build them together. Honestly, I hadn't heard about AWS's LLM model before, and I had to look it up. It's called Bedrock. Rod, have you heard about it before?

Rod: Yes, but it's not an LLM, it's a service. AWS, compared to all the other players so far, hasn't really released dedicated LLMs like we've seen from Meta or Mistral. Instead, what they've been doing is investing in companies, startups, or large scale-ups. The other thing they've been doing is creating platforms like Bedrock that enable companies to easily deploy open-source models like those from Anthropic, Meta, and so on, in a scalable way within their ecosystem.

So, let's say you're an application developer, like a front-end or back-end developer, and you say, "Hey, I need to start adding AI to my products." Bedrock enables you to do this in a very straightforward and easy manner.

Chris: I see. So Amazon is coupling it with their cloud service to put an AI layer on it and sell it to corporates. At least that's how I see it. But I think also here, and Max, maybe you can weigh in on this, healthcare seems to be a very interesting field for AI use cases. GE uses it to streamline healthcare operations, which I think is comparable to other industries, but they also increase diagnostic and screening accuracies and obviously help and support doctors.

AI in Healthcare: Untapped Potential

Actually, I think there was one statistic which I found incredible: the healthcare sector is responsible for about 30% of the world's data generation, and 97% of this data is actually untapped because of its unstructured nature. What's your take on that?

Max: I think that's super interesting. I wouldn't be surprised because of the amount of data that comes out of all these medical devices. The thing about medical and healthcare devices is that many of them are becoming electronic, so a lot of that data is captured but then stored in a less efficient manner compared to what we see in big corporations or e-commerce.

The question becomes, who is going to extract that data and help create services on top of it? I remember looking at a company that was using ML specifically to increase the chances of IVF success. They were trying to improve the overall insertion of sperm into an ovum with more precision. That alone could improve many lives because if you don't succeed the first time, you have to keep injecting yourself until you get results.

And that's just one use case. Can you imagine the amount of complex data we have about our bodies and the amount of data one would need to get to that right, precise, or personalized offering? It's potentially game-changing.

This also links back to L'Oreal's use cases we discussed earlier. A lot of people in the Western world are quite sensitive about having acne, so they might not want to talk about it with others. They might feel more comfortable talking to an AI to help them find the right information or triage their concerns.

The same thing might happen with healthcare. One thing I can think of is the NHS in the UK, which is a free service. I have healthcare records with them probably in three different places. When I broke my leg, when I had the flu - I wonder when they're going to consolidate that data. They never actually consulted the data and realized that I had the flu quite often or broke my leg quite often. Therefore, the first thing to do when I go in for emergency services should probably be to push me to an osteopath to look at my leg first.

So I think the disparate data is what I'm trying to highlight. Even with medical devices like those from GE, the collecting and storing of data is what we're missing a bit. This makes it very unstructured and, I would say, undiscoverable for any ML and AI applications. But I do believe there are more and more use cases emerging. I believe we could live to be over a hundred years old if that data is properly surfaced because we would be able to do a lot more prevention rather than focusing solely on cures.

Chris: Definitely. I think your point on the fragmentation of data, and I suppose regulations on data security and privacy, are probably the biggest obstacles for data crunching and really utilizing that 97% of data that's not being structured or discovered.

Volkswagen's AI Integration

We've talked about different kinds of companies. There are maybe two or three other companies I want to highlight because I found them very interesting. One was actually Volkswagen, who announced earlier this year an integration with OpenAI to provide what they call a ChatGPT-assisted voice assistant. It really reminds me of our last show. Rod, do you have any views on that kind of use case - having a more sophisticated conversation in the car while you're driving? Do you think this Volkswagen integration is a smarter Siri? Or do you think it's a bit pointless? What's your take on this?

Rod: Yes, it's hard not to think back to what we were discussing in previous shows. Companies in the past have tried to integrate various types of assistants like Siri, Cortana, Bixby, Alexa, etc., with mixed results. In the end, it turned out to be much more difficult than expected, and the technology wasn't really there. Maybe in the end, there was some functionality, but it was more of a "dumber" functionality. In the case of a car assistant, it was pretty much limited to simple commands like "play this song" or "turn on the volume" or "call my parents."

But now that we have much more robust assistants, such as those provided by OpenAI, it makes sense to revisit these old ideas and say, "How can we make our vehicles much smarter?" Not only in terms of communication but also in other areas like diagnostics, providing information about the car itself.

We also have to consider that vehicles are now essentially electronic devices. You don't have buttons and knobs anymore, but just a big tablet, a big screen in the car, and everything works via touchscreen. The downside of that is that it's much more difficult for individuals who aren't used to this technology to find different settings. It's also less user-friendly for many people - not everyone has the necessary tactile skills to navigate touchscreens easily.

So being able to just communicate with the vehicle and say, "Change this setting" or "Please find the shortest route from A to B" becomes much easier. It's possible to have dialogues where the car can respond and say, "Do you want to stop for an ice cream on the way?" It becomes much more dialogue-based, conversational, and ultimately, a better user experience.

Chris: And Rod, would that now incentivize you to buy a Volkswagen?

Rod: I would say that this is becoming standard. The big challenge in everything AI-related is that one doesn't really have an advantage for long. It becomes very limited. We're seeing a race to the bottom in terms of AI becoming cheaper and more powerful. They're becoming so much more robust in terms of the use cases they can cover.

What Volkswagen offers today, or what Tesla or anyone else offers, you'll soon see all major car manufacturers and all major models offering. Not only that, but we'll also see retrofit kits. If your vehicle doesn't have this functionality available, there will surely be many startups and even product manufacturers on Alibaba and similar websites offering the ability to retrofit your car with these smart assistants, even if your car was a "dumb" car from the 90s.

Chris: So Rod, basically your answer is no. You wouldn't buy a Volkswagen simply because they now have ChatGPT as a voice assistant.

Rod: Exactly. It will be something that's just normal, just like every vehicle has a radio, and you don't buy a vehicle just because it has a radio inside. The same thing will happen with smart assistants and AI inside cars.

Max: I just want to add to that - imagine if you have a lot of devices that you have to talk to. I feel sorry for introverts!

Duolingo's AI-Driven Success

Chris: Definitely. One last company I want to talk about, because I couldn't believe the numbers, is Duolingo. Duolingo actually reported that they were able to increase their revenue during the quarter by 45%, while everyone else reported declining sales. I thought that was really remarkable.

Basically, what they did is create AI content that's also helping them on the bottom line, as we discussed with the Klarna example. They reduced their contractors who generate content for Duolingo's language experience by 10%. But at the same time, they also created a new subscription tier that offers AI-generated feedback, conversation, and voiceovers in various languages. It seems that customers or users are willing to pay for it, which actually increased their revenue by 45%. Isn't that incredible? Max, what's your reaction to this?

Max: Definitely. The ability to converse better and learn languages more effectively using various AI tools to change the learning experience - is it worth paying for? I think I would pay for that. It would be quite interesting. Because ultimately, if a tool can help me reach my end goal quicker, which is learning a language faster based on AI conversations, that would be super helpful.

I've used Duolingo before. How it works is that when you sign up, you tell them what level you're at in a specific language, then they teach you how to pronounce certain words. Now, if you make it more dynamic where you can speak back to it, that's quite interesting to me. It's a bit like how I could learn a new language by being amongst people who speak that language.

So I feel that Duolingo could be quite game-changing when they start doing that. But coming back to my previous point, you just have to keep talking to your phone to learn that language. That's one part of it, which is interesting.

The other part is that we're seeing a lot of people utilizing AI to create content, especially on the marketing side of things. Instead of just having a person write copy, you can have the AI do the first pass and then have someone else edit it down the line. This is very similar to the Klarna use case of reducing the amount of agency fees they're spending on marketing.

So I think they go hand in hand, but I'm more interested in the experience side, as you can probably tell. It changes how you and I would behave around an application because it would help us get to our end goal quicker, either to learn the language or to be able to communicate with someone's parents, and so on.

Rod: I would say for a company such as Duolingo, the challenge is retention. Many people have the motivation of "This year, I will learn a new language." But in reality, most people may not have the habit, may not be able to fit it into their schedule, or may just be lazy. So companies like Duolingo need to figure out how to keep these users engaged, how to keep them learning and progressing.

The classic path has always been gamification - having functionalities with points, medals, and so on that show slow progress and get people a bit addicted to the product. But now, with AI, we can have a virtual teacher or tutor that understands where I'm struggling. Is it grammar? Vocabulary? Pronunciation? Based on this, it crafts personalized learning paths that are adjusted in real-time to what I'm saying.

This way, the user is more interested because they see progress faster. The user realizes, "In this area where I was struggling, I'm getting better." But for the companies, this means that the user doesn't cancel their subscription. As a result, revenue increases not just because new users are coming and converting, but because existing ones aren't churning out.

Chris: Definitely. Rod, you're making a really good point about retention, and I'm really curious whether Duolingo can actually keep up that revenue growth. Yes, we're all very curious and willing to try new AI features, but I wonder how many of the newly signed-up users are actually going to stick around within the next six months or so. But I think this remains to be seen.

Conclusion

This has been a very interesting conversation on different use cases. I think after 18 months, this is really a good time to review, and I've learned a lot from our conversation today. Maybe I'll just summarize it quickly for our audience.

If they wanted to do something in their company, I think one key takeaway is that data is truly the foundation. If you do it right, it really does have an impact on your bottom line and your top line. We spoke about the proprietary data of L'Oreal and Walmart, and both of them were able to use generative AI effectively because they were able to clean up their data and use it in a structured way.

This actually helped them build better customer experiences, whether it's helping Walmart customers find items more easily or even helping them discover products they hadn't thought about before. On the L'Oreal side, we discussed the enhanced customer experience in finding the best personalized products for users.

Another thing we talked about was how we may not need entirely new ideas, but we can look into the backlog of old ideas that failed in the past. L'Oreal tried skin matching before, but it didn't work out so well from a customer experience point of view. We've seen other companies like Reckitt really harnessing AI, being able to reduce the time to come up with new concepts by 60%.

Also, bigger companies like Airbus and Volkswagen are now building these AI workforces or task forces to really look at use cases. So revisiting some of the old ideas may also help your company.

Last but not least, I'm really happy to see that AI not only decreases costs but also really helps on the top line. We now have some concrete figures in our show, with Walmart reporting a 26% growth in e-commerce and Duolingo claiming a 45% increase in revenue. I think that's really exciting, and I can't wait to see what else people, corporates, and companies come up with.

With that, I actually want to turn to the other side. We also spoke a little about strategic partnerships with AI-offering companies. Yes, there are bigger players, but we also talked about startups. As I mentioned at the beginning, Forbes came up with a list of 50 AI startups worth mentioning.

Some of the top companies we've spoken about a lot, like OpenAI and Anthropic, and the significant funding they've received. But I think it's interesting to look at the bigger themes that arise in these 50 startups. There are three top use cases that are getting funded in the startup world:

  1. AI model development (OpenAI, Anthropic, Adept, Cohere)
  2. App development and deployment (Replicate, Scale AI)
  3. Image and video generation (Midjourney, Runway)

Interestingly, none of them really tackled some of the corporate use cases we've discussed. What's your take on this? What was your impression when you went through this list?

Rod: For me, what surprised me is that it shows how it's always possible to use technology, to be part of the AI wave and transform yourself into an AI company. For example, Notion, the productivity software with easy-to-use pages that many people use - I wouldn't have necessarily associated it with AI. For me, it's just a tool similar to Google Docs to create documents and maintain knowledge management. But it's part of this list, labeled as an AI company. Notion has heavily invested in AI.

This idea that if you go all-in on AI, it might not get the results you want, or it might be a waste of money - that's not necessarily true. In the case of a company like Notion, they have very successfully transformed from being just a SaaS productivity software to now being considered a relevant AI player by Forbes.

Chris: So what exactly is your take on this, Rod?

Rod: My take is that for any company, especially those who are not AI natives, there is always a possibility of adopting the technology, innovating, and then transforming oneself into an AI-first company.

Chris: Got it. So basically, you're saying any startup can become an AI startup if they execute it correctly.

Rod: Exactly, because if you look at many of these companies, they were started way before the generative AI wave, in 2016, 2017, and so on. They were already doing other things - they weren't AI infrastructure providers. They were, for example, doing discovery or analytics. But they decided to integrate AI heavily into their offerings and processes. And now they have successfully transformed themselves into AI-first, AI-native companies.

Chris: Got it. And Max, what's your take?

Max: The first thing I noticed looking at the list is that 30 or 50 of them came from the US. It's amazing how much dry powder these companies have and how much influence they have. That's one observation.

The second thing that caught my eye is the founding years of these companies. I agree with what Rod was saying - a few companies started in the pre-generative AI world, like Databricks and Cerebras Systems. And then there's also a whole host of companies that were built on top of or in the wake of ChatGPT and OpenAI.

So my first point is that yes, while some players have done something else before and could now utilize AI to transform into something new, OpenAI also created almost an entirely new category of AI companies that may not have existed without the ChatGPT realization.

Secondly, if all sorts of software can turn into AI applications, then why would some of this software still be around in the longer term? I heard someone on a podcast mention that if you believe in AGI (Artificial General Intelligence), why are we still building software? Because that AGI could do everything for you. So I'm trying to think, even in a non-AGI world, what are the different processes that one would need to do?

Today, AI helps you with image generation, writing words, and content generation. So then the question is, what else can it do for you? What else can you do on top of that? Probably a lot, right? Because once you can do content generation, you can come up with instructions, which then can turn into algorithms, which will then turn into actions. I don't think that's too far off - we just need some time to get there.

To answer your question about the most interesting company on the list, I'm going to say Midjourney. Why? Because they've raised zero money. They've just been building. Previously, I read that they didn't even have a website - everything was done through Discord. But I'd like to hear both your takes on interesting companies.

Rod: I want to add to Max's point about the location being predominantly US. My take is that this correlates to the investment sums that these companies require. In this list, many of them have raised 100 million, 200 million, even 600 million dollars. This amount of money is only available in the US.

We have the case of, for example, Hugging Face, which started in France but is now labeled here as a New York, US company. However, the company has its origins, founders, and a large part of the team still in Paris. This is just due to the fact that to attract top talent and compete, you need to buy very expensive hardware, and these are very cost-intensive companies.

The only place where you can find the resources to finance these types of operations is in the US. Innovation is happening in many places - Europe is having a lot of innovation. But if you really want to grow and be able to compete, then sooner or later, you have to rebrand yourself as a US company.

Max: Are you moving to the US, Rod?

Rod: I guess this is also something that anyone in the AI space can learn for themselves, right? When and how should I need to be in the US? Do I need to do it? Because in the US case, innovation is happening there, and this is also where the opportunities are in terms of talent, market, and what's possible to do.

Chris: I totally agree with you, Rod. I think if I look at the list and some of the most noteworthy companies they display, the use cases aren't so particular, right? There's this one company, Abridge, I think they're helping doctors do some of the menial tasks they need to do for documentation of a patient's state of health, etc. So they build a service around that. Or another one, Harvey, I think it's called - they help lawyers draft legal documents and review contracts.

Honestly speaking, even at the Technical University of Munich, you have similar use cases. So I would totally agree with you that the question is, where do you really have the resources to get off the ground and also the network to really distribute it? And I think I agree with both of you that the US obviously has a different scale of resources when it comes to that.

I think the second observation I made, and this is really linked to the previous discussion we had, is that a lot of the startups are building features. They're building the video editing function, or they build this speech assistant function. But to really bring it all together, this is where I think there's a bit of a gap between what startups are currently offering and what companies or corporates are looking for. They want all-in-one solutions that startups simply can't deliver, just due to sheer size and security concerns.

I think I would want to encourage all the listeners who are working in startups to think a bit more holistically and also to really talk to these task forces to make sure to try out some POCs (Proof of Concepts) and really help to develop further the different use cases that startups are offering. Because from my previous role in innovation, I do think that there's a gap in understanding what enterprise really needs and really building for that. So there's a bit of a language barrier, I would say. That is at least my take. How about you guys? Do you have any final words on the two topics we discussed?

Rod: Definitely. The companies that are not investing are being left behind. Things are moving extremely fast. Those who are making these investments in AI are not only reaping benefits but also getting very positive PR from that. If we look at Walmart, in the end, this is retail. This is something that's not so exciting. How many retail companies are out there in the world? There are so many, but who is talking about them? Well, not many people are talking about Walmart because of use cases. The same thing happens with L'Oreal.

They're in FMCG (Fast-Moving Consumer Goods), which is almost like a commodity. The brand power is very limited. But now we're talking about them today because they're doing big AI investments. So if a company is hesitating about going all-in on AI or not, then they are wasting their time. They should have made the decision a long time ago, and they should be investing extensively in AI.

Chris: Definitely. So basically, you're saying anyone should actually be looking into AI and really think about how they can harness it just to stay competitive. Max, what are your final thoughts on this?

Max: My last thought would be let's get building. I can't wait to see some of those use cases. I've been trying to use AI myself, right? Try to do, I don't know, write up survey questions all the way down to researching some financial ratios on larger corporates for competitor analysis. I think those are all interesting use cases, but they kind of fall under the search bucket, if you will.

So that's one part of it from an enterprise perspective. I think I echo what Chris was saying, which is, you have to have some sort of design partner in order to really help you solve problems. One thing that I heard recently is that, yeah, their solution is cool, but how does it solve my problem like today? Always keep that in mind. You're building cool stuff to help solve problems. Otherwise, you're doing experiments. They're still cool, but may not have the immediate commercial impact one would be looking for.

And then, I guess, lastly, just on funding. I definitely echo what both of you were saying around funding in the US is just on another scale. I guess this is more for regulators. Maybe they need to step up and do something to attract the right talent for their respective regions. Otherwise, this is not just a competition within the business world. It is also competition from different jurisdictions.

Because the way I think about it is you want to have multiple types of AI models because it's a reflection of different types of thinking and culture in different worlds. It's more about having multiple models to help you get better results for local personalization, to help you not eventually have everyone fall into the same way of thinking. So I'm interested in seeing more solutions coming up from different parts of the world as well.

Chris: Max, thank you for your very last long thought at the end. I think I agree with both of you. Funding is an issue, jurisdiction, etc., not new topics we've spoken about for the first time. But again, I think just to wrap up this show today, I thought it was super interesting to dive into the different use cases corporates in all different kinds of industries have started to implement. I am also very happy to see that there are also consequences and effects on the top line, which I think the general market has been very skeptical about.

We also spoke a bit about startups and how they actually serve and cater to these use cases of bigger companies. And I think with that, I want to wrap up today's show. Thank you guys for listening. If you like the show, please subscribe. Leave some comments, give us some feedback. We always love to hear from you. And with that, have a great rest of the week.

More from the Podcast

Post Image: E21: Is OpenAI's $157 Billion Valuation Justified? Chris Rod Max Weigh In

E21: Is OpenAI's $157 Billion Valuation Justified? Chris Rod Max Weigh In

In this episode, we dive deep into the recent challenges faced by OpenAI, including employee departures and leadership changes, and explore whether its $157 billion valuation is justified. Chris, Rod, and Max discuss the difficulties of rapidly scaling AI companies and provides insights on potential investments in the AI sector. We also examine Hippocratic AI, a healthcare AI startup valued at $500 million, and debate the future of AI in healthcare. Join us for a fascinating discussion on the current state and future potential of AI companies.

Rod Rivera

🇬🇧 Chapter