- 3 months ago
Everything, Everywhere AI-Powered Marketing
Category
🤖
TechTranscript
00:01Good morning, everyone. I am Aurélia Bittotti. I'm a partner at McKinsey & Company.
00:06I'm actually leading the marketing practice for Europe. I'm very happy to be here with you today.
00:12To start our discussion, I'd like you to invite you to imagine a future where marketers transcend
00:18creative boundaries and are actually able to propose the perfect offer at the exact right time.
00:25It would mean that AI and Gen AI would have actually enhanced the efficiency but also
00:30deepen consumer insights and also make hyper-personalization communication on a
00:35massive scale. This is actually not a myth. This is truly a reality. Gen AI and AI have massive
00:43impact on the B2B and B2C industries and at McKinsey we assume that the productivity boost
00:49would be massive, $4.4 trillion a year and specifically on marketing and sales, 5% to 15% productivity
00:57boost.
00:58So that means more than $460 billion each year. On a practical level, it also means that what used
01:07to take months to develop would actually be able to be developed in a couple of weeks or days and
01:12increase. And on top of this efficiency gain, it will also enable to boost creativity.
01:18I'd like you to embark with me on a journey into this future of marketing and we will be joined
01:24by three
01:25special guests today that I'd like to introduce. So Matthias Chayou, Matthias, who has been fans,
01:33Chief Media Officer at L'Oréal. Matthias, you have 20 years experience in various advertising and digital
01:39agencies across France, Italy, China and UK. Welcome. Sylvain Lebogne, Chief Data Officer at GCDECO.
01:47So you have funded multiple companies including Adrium, Acostar, Adledge, you have worked at Havas,
01:5355, MediaMath and so you have extensive background in ad tech, data and digital transformation. And
01:59finally, Ellen Meil, Chief Marketing Officer at VIEW, leading global digital out-of-home marketplace that
02:06combines data and technology, connecting buyers and sellers, always full transparency. And you have
02:11also previous roles, marketing senior roles at Parkopedia, O2, Tern and Microsoft Advertising. So
02:20thank you all for being here. So first of all, I'd like to discuss with you the use cases of
02:25Gen.AI in
02:26this specific marketing and media fields. So McKinsey identified four main use cases in the marketing
02:33field. Personalisation of marketing campaigns, so deepening customer engagement through more
02:38personalised and frequent interactions. Second is the analysis of unstructured customer data. So for
02:44example, interpreting customer feedback to shape better product recommendation. Process automation,
02:52so to improve the interaction between the marketing function and other functions. And finally,
02:57IT generation, so analysing competitors' moves, testing new product opportunities. And this enables
03:03to improve the efficiency of features, the time to market, the testing accuracy. Maybe we start with
03:09you Matthias and you can share a bit on the B2C side, two or three, yeah, one or two significant
03:14use
03:14case that you've been developing at L'Oreal. Yeah, so I'll give you two, but you started by talking about
03:19personalisation. And we see all around the world that the beauty becoming more and more individual
03:25every day. Which is why we've moved as a group from what we used to call beauty for all to
03:31now beauty
03:31for each powered by beauty tech. Beauty tech, we believe, is critical to empower consumers with
03:38better experiences. One of these experiences in use case, for example, is services. You know, 70% of
03:45people feel overwhelmed in the beauty market by all the choices they have. The number of brands, the number
03:51of products, but also the sources of information that are available to them through influencers,
03:58social networks, and so on. L'Oreal Paris Beauty Genius solves this issue. It's an assistant,
04:06a personalised assistant that can give you personalised recommendations, answer your questions,
04:12let you try on virtually products. All that in your pocket 24 hours a day. So that's one of the
04:19use case we
04:20see in beauty. And the second one, and I'm going to talk a bit about media because that is, and
04:24I invite
04:24you all to come see the stand and discover beauty genius if you haven't. The other one I'm going to
04:30talk about is media because it is what I do. And we've seen the media world more and more move
04:36to
04:37AR powered solutions. AI is not new in media, but the platforms are proposing solutions that are becoming
04:43better and better every day. And we've seen substantial improvement in most, if not all, of the metrics
04:51on media. So AI powered media as well. And one of the reasons we believe that this is working better
04:56is that algorithms are obviously better at math than marketers and they also have less bias. So when we
05:04optimise our campaigns, for example, to sales, an algorithm will have less bias than a marketer in Paris or
05:11London, to decide how much of our marketing budget is going to be allocated to Gen Z versus Boomers,
05:16for example. Yeah, amazing, amazing example. Maybe Ellen, I turn to you on the B2B side. Given your
05:23extensive experience in digital media, can you explain a bit the latest investment in AI you've
05:28seen with your client recently? Yeah, sure. And just before we get into that, I did try the beauty genius
05:33this morning. So next time I'm here, I'll have different hair and be 10 years younger, by the way.
05:37So, you know, it was a great recommendation. But yeah, what we're seeing is that AI is really helpful
05:44for, you know, as a tool. It's for scaling, it's for reinvention and for growth. So I think the most
05:54immediate use case that we're seeing is in creative iteration. What you can do is take, you know, the source
06:00creative and actually exponentially enhance that and have many more iterations. So you go from 10s or
06:08potentially, you know, up to 100 to actually into the thousands. And Google's AdPixel team have
06:15actually done that. And they've taken 30 different creatives and they scale that up to 9000
06:21in a matter of minutes. And so what it does give you as a marketer is a huge amount of
06:27choice.
06:27And to Matthias's point is to be able to get that increased level of personalization.
06:34And in the out of home world and the programmatic specifically, which is the space that Vue sits in,
06:40what we're seeing is that because you have this extensive range of creatives now at your disposal,
06:46you can really utilize the real time nature of programmatic digital out of home, the contextualization
06:52of the locations, all of those different iterations, and make that ad relevancy really massively increase
06:58and become much more relevant to each consumer group. Yeah, indeed. Sylvain, maybe you want to speak
07:04about the outdoor advertising specifically, what are the benefits for your clients?
07:08So as you all know, JC Deco is managing some very, very premium assets all across the planet. So we
07:17are in 85 countries. And what we are seeing is that attention has become a very, very crucial metric
07:23and KPI for our clients. So we have built an AI and machine learning solution that can predict the
07:30attention that creative is going to drive. We have used a tracking studies to feed and train this machine
07:36learning solution. So now every before campaign or after when we want to understand more finely the
07:42results, we can understand what it draws the attention, what was the gaze sequence. And on top of that,
07:49we can put that in situ. You can insert the ad inside of a video and understand based on the
07:56dynamic
07:56around the asset, what is happening and how the attention is captured and what is the experience
08:03that is created for the consumer. Interesting. We'll discuss the impact of that right after. But maybe
08:09Helen, can you also explain what is the impact on the traditional media mix evolution of this revolution?
08:16Sure. So maybe just a show of hands in the room. We're going to do some interaction. How many people
08:21have a
08:22conversational AI tool in their home? So that would be Alexa, Siri, something like that. Okay. So at least
08:28probably getting up to 50%, I would say. And now imagine if one of those AI tools, you're asking it
08:36for a
08:36suggestion on a holiday. So, you know, Siri, where should I go on holiday this year? And you might give
08:41it, okay, I want, I want sunshine, I want beach, you know, you'll give it some tips. And what's happening
08:47now is that we're seeing and actually just this week, Google has started to introduce in their search
08:52results that you can have AI suggested search results. So this has a massive impact. If you're
08:58a brand marketer, and you know, you've got AI in there, AI suggestions in there, as well as sponsored
09:04results, you know, if you're that holiday company or that destination marketer, and you're trying to
09:10advertise and beat out the competition, you really need to think about how is your destination going to
09:17win out when the AI tool is actually having such an impact. And you know, there's Google Gemini have
09:24already started to work in this space. And I think that what you perhaps as a marketer would need to
09:29think about is actually what are, what's the mix? What's the touch points across the whole consumer
09:35journey? Because how much of you as a marketer can you as a marketer really influence? So, you know,
09:40as I said, I live in the real world space. I operate in the real world space. We do use
09:46tools and
09:46automation to increase that relevance to make that real time out of home opportunity really pop. But
09:53I think there's definitely potentially a shift away from the dominance of of display, perhaps,
09:59and into this much more cohesive mix of media channels. Indeed. Now, that's super interesting.
10:06And I think we really see how transformative Gen AI is and all the different sets of use case and
10:11and diversity it can create. Today, we estimate that between 30 and 40% of marketers are actually
10:16already leveraging Gen AI. So it's the penetration is evolving. However, obviously, there's great
10:23potential and sets of use cases also imply substantial challenges, right? So from operational,
10:30technical perspective, ethical perspective. And I'd like to us to discuss a bit, how do you overcome
10:36that today? So what we see when we surveyed commercial leaders is that they see usually four main challenges
10:42to overcome. The first one is obviously to determine which technology solutions to prioritize, right?
10:48You have all sorts of solutions available. So which one to take? Second is established guard rates.
10:54There are massive issues of accuracy, compliance. So how do you avoid stereotypes, hallucination,
11:00and intellectual property entrenchments? Third is about data security, which obviously is super important.
11:07And fourth, that sometimes we underestimate is also related to the transition for workforce.
11:13How do you need to adapt your skills specifically and develop more technical, social, emotional skills
11:19and to reskill a percentage of the populations? The McKinsey Global Institute estimate that 7 to 8%
11:26actually of the workforce need to be reskilled by 2030. And that's quite massive. Maybe we start with you,
11:32Sylvain. Can you tell us a bit more about GC2Co? How did you overcome the technical challenges specifically
11:38of JNI? So what we have put in place at GC2Co is a three-layer approach. The first one is
11:44to make sure
11:45that when there are low-hanging fruits, where our AI solution already integrated in the day-to-day tools we
11:51are using,
11:51we train our team to use them. We want to make sure that they become comfortable in using some AI
11:57and Gen AI to
11:58be more productive, more effective, and also provide a greater service to all of our clients.
12:04The second layer is when we need a specific application, but when the data is not as strategic.
12:11If it's not strategic, but needs to be confidential, we will integrate a market solution, we will carefully
12:17choose it and make sure we can deploy it as fast as possible. Because what we are all seeing in
12:22Gen AI,
12:23especially, that doing a proof of concept is quite fast. But scaling it and getting to adoption is way tougher.
12:31And then there's a third layer, when you touch confidential information, confidential data,
12:36but also trust strategic topics. And for that, we build internally. We make sure we integrate the best
12:42open source LLN libraries, for example, but we make them run in a closed environment where we can
12:49control how we create the corpus, how we interrogate it, and what kind of prompts we can run.
12:57Mattis, maybe if we move from the technical to more the upskilling part,
13:01can you tell us a bit how did you make this transformation at L'Oreal from a change management
13:04perspective? Yeah. Say, I'll part Gen AI for just a second, but AI is not new to L'Oreal.
13:11We've been working on it for a few years now, I assume like both of you as well. We have
13:188,000 people
13:19who are dedicated to digital tech and data. And of those 8,000, 1,000 are dedicated to pure data
13:26projects. All of them have been involved in AI-related solutions and systems over the past year. So
13:34there's already a lot of training that's been done. Now coming to Gen AI, I think in light of your
13:39introduction, what we've done is built a Gen AI task force that has three core missions.
13:45The first one being defining guardrails. How do we use it? Because Gen AI does come with a bit
13:52additional, how would I say, complications versus AI in general. The second one is defining use cases we
14:02all feel comfortable with and we believe bring value to the group. And then the third one is to upskill
14:06everybody. So we have an ambitious upskilling program. And the last one is to provide all our
14:11employees with access to our L'Oreal GPT. And today 27,000 of our employees use it regularly. So
14:20quite a lot has been done there. Quite impressive penetration rate indeed. Super. So besides this
14:27technical and upskilling part, there is the ethical issue, right? So Helen, can you tell us a bit about
14:33this? How do you overcome these challenges and obstacles from the ethical perspective?
14:38Sure. So I think there are, you know, thankfully the tech has moved on. I'm going to start there.
14:43But there were some very well-publicized examples where people had got it wrong in the early days.
14:49And I think that when you look at the source data, it's really important to make sure that that is
14:53unbiased and inclusive. So, you know, there was some well-publicized examples, I won't name names,
15:00but, you know, where facial recognition was identifying darker skinned people and not as humans.
15:06I'm just going to put it out there. You know, that was obviously a very bad experience. But I,
15:10as I said, I think both the awareness of the need for the source data to be really, really concrete
15:17and
15:17unbiased has improved massively as well as the technology. And so, you know, in today's world,
15:23if you're doing a campaign and you're using mobile face recognition or voice recognition,
15:29what I would say is think about how inclusive the source data is and making sure that it's really
15:34offering that opportunity for the world that we live in today and to be, you know, not making
15:41assumptions. And then that way you get the best output, both for you as a brand, but also for
15:47the consumer as well. I was having a conversation with Mastercard recently and they said that their
15:53approach to AI is that, you know, they have very strong core values in their company brand values of
15:59authenticity. So they've made a conscious decision that they will not use AI generated imagery in any
16:06of their advertising. But what they do do is use AI to listen and decode social media conversations,
16:14right? So they're aggregating that level of insight and being able to just quickly jump onto micro trends
16:21that they're seeing and insert themselves in a very organic way, whether that's in social media or
16:26utilizing that in the out of home space or, you know, any of their media channels and really become
16:31very relevant. And so they're using AI for that purpose, um, to be, to increase their relevance
16:37and act in a very fast manner, as opposed to, you know, they're going to go down the AI generated
16:42imagery
16:43route because for them that doesn't align. So just make sure it's you're sticking to your core values,
16:48I think. Yeah. I think Matthias as well at L'Oréal, you've established very solid guidelines and
16:54commitments right on the ethical part. Can you also elaborate a bit on how did you manage that?
16:58Yeah. So we've been working with, with external partners to establish seven principles of trustworthy
17:03AI that go from, uh, uh, the deploy, the, the development, the deployment and the usage, uh,
17:10of, of, of AI solutions. Uh, one of these principles, for example, which is critical in
17:15Gen AI that was a bit different with when it was just AI, uh, is, uh, the, the, the absolute
17:21necessity to
17:22have supervision, human supervision of AI systems, uh, and more importantly of the usage of the
17:30recommendations that, or the solutions come with. Uh, the second one is we've, we've, um,
17:35we have a responsible advertising and marketing, uh, communication policy, which covers many things
17:41such as, uh, responsible content, how we behave in media, marketing to children. Uh, and we've added AI
17:47to that as well. Uh, so to, to the point you were making, Alan, as well, that what we've taken,
17:51uh,
17:52the stand that we would not, uh, we do not, would not use life-like, uh, face, uh, body, hair
18:00or skin
18:01that's been generated by AI to show a product benefit or a product enhancement. So this is, uh, some of
18:07the,
18:07the rules that we've, uh, we've, we've, we've, uh, decided to, uh, to apply to our, uh, to our, to
18:13our
18:13marketing. Great. Thanks a lot. No, I think the, this is really super interesting. And in the end,
18:19we see like the real challenges for organization to address all those mitigation strategies,
18:25robust governance, and importantly, the human oversight, right, of the strategic and operational
18:30levers. Um, I, I would like now to turn out more to the positive impacts, right, that we briefly
18:36mentioned at the beginning and what, yeah, can generate actually, uh, generative AI and AI.
18:42So in our work with our client, what we see in terms of impact is super positive. It means that
18:47we're able to create capacity and nearly double cost effectiveness in every really marketing work.
18:54Uh, we are, we are seeing like, um, three to 15% increase in revenue when you actually really
19:00activate, uh, AI properly. And we also see sales ROI boost in between 10 and 20%. So, uh, this is
19:08really a massive impact. Obviously it takes some time between one and three years to develop that,
19:13but really it's not about only R and D and fun proof of concept. It's really active impact, uh, on,
19:21on the, on the business side. Um, maybe Mathias, you want to highlight on how is L'Oréal planning to
19:26invest, uh, in Gen. AI. And how do you see this technology really enhancing the return on investment
19:34for you? So, so we've built, um, we've built, um, um, um, marketing mix allocation tool that's called
19:40BetIQ. Uh, Bet being our beauty engagement touch points. So pretty much every interaction we have with
19:46consumers. Um, and BetIQ leverages AI, uh, through marketing mix model, um, optimizer. Basically,
19:55it looks at the response curves of over 50 different touch points and what is the best
20:00allocation of our marketing mix between all these 50 touch points. Uh, that enables us to make in real
20:07time, um, improvements to our marketing mix, uh, investments. And we've seen through this use of
20:13this tool, some significant, I won't quote the numbers, but it's not far from, from what you've
20:18said, uh, uh, improvements in, in ROI. Uh, now the point I wanted to make though, is it was more
20:24than just
20:24the AI itself. We've been using AI to power this tool in the sense that you can have all the
20:30optimization you want. If you don't have the robust data behind it, um, the system's not, you know,
20:36rubbish in, rubbish out. Uh, and we've been working with a French startup, uh, called the Grasp, uh,
20:41uh, that we accelerate at station F within beauty tech and with HEC, um, that enables us in something
20:48that's, uh, I know advertising and marketing people are very passionate about, enforce techno,
20:53taxonomies and naming conventions, which is critical, uh, to succeeding in an AI world. And
20:57not only working with Grasp, we've been able to, uh, enforce those taxonomies and naming convention,
21:03but we've worked with them and their AI solutions to correct historical data. You can't imagine the
21:09number of people who miss Mel La Roche-Posay, for example, AI solutions enable us to go back in time
21:14and fix historical data. And that's been a really big game changer. Yeah.
21:18Impressive. Ellen, what about you? What's your vision of the future of marketing and media parity in AI
21:23in the coming years? Um, so, well, I think actually we're already starting to see the future. Um,
21:28you know, the immediate use cases, as we mentioned around creative, creative iteration, um, across
21:35multiple channels and really being able to make each channel work to its maximum, you know, to its best,
21:42to its most relevant self. Um, and I think that is seeing that scale and the ability to utilize
21:49things like dynamic creative optimization in, in programmatic out of home, it's already in other
21:54media channels, but it's now coming here and the ability, if you've got, you know, multiple creatives,
22:00um, that make the relevance improve, then that's what we're already starting to see. Um, and it's,
22:06it's, you know, it's a tool, it's something that you should use, but to Matthias's point,
22:10it doesn't take away from having that human overview and the strategy. And that still needs
22:15to come down, you know, to, to a real person, I think, or a team of people. Um, so that's,
22:21that's, I think how we're going to see it evolving in the future is, you know, it's a tool for
22:25use.
22:25It's a tool for scale and growth and speed, you know, speed to market is, is huge these days. And
22:31the ability to tap into those micro trends, um, to improve your relevancy as a brand is really what,
22:39where the future is. Um, and that, I think that's what the consumers are craving as well. Um,
22:43Matthias mentioned, you know, it's not, it's not one stereotype of beauty. It's everybody has their
22:49own personalized, uh, journey and their personalized look. Um, and we see that with multiple brands
22:55in multiple sectors that the consumers are really looking for that level of, um, contextual relevance,
23:00highly personalized, um, and that's really where AI can, can support, I think.
23:06Indeed. Indeed. Maybe, Sylvain, so you, you have been highlighted those exciting use case, uh,
23:11at the beginning of this conversation. Can you tell us what is the game changer in terms of outdoor
23:15advertising of GNI? So there are already been a few changes that, uh, have been powered by, by AI.
23:23Matthias was mentioning marketing mix modeling. Now it's fast. Now it's accurate. Now you can use a lot of
23:29different variables and get simulation and make very, very quick and precise decisions. Um,
23:36Ellen, you, you're part of the programmatic revolution in, in, in out of home. So same stuff.
23:40AI is allowing anyone to transact in real time to buy premium inventory in out of home. And for me,
23:47the next frontier is about a creative production. Um, today, one of the barrier to entry for a media
23:54like ours is how do I, how, how am I going to create the right creative with the right message
24:01and being fast enough to react to those micro trends. With GNI, you can have the creative ideas
24:07and then have a different iteration of your creative idea and your message being prepared
24:12so that you can then pre-test it with a solution I was, for example, explaining at the beginning,
24:18and then pick and choose the right one and decide to publish it on, on, on the right asset
24:24and, and creating the right experience for your consumers. Amazing. Indeed. Maybe to make the link
24:29because Matthias, you already spoke about the marketing ROI, but can you also elaborate a bit
24:33on how do you bridge the gap between analytics and creativity? Uh, so I'm going to start by, by, by,
24:40by saying, I think Gen.AI obviously will, you know, on, on every, on, every aspect,
24:46improve creativity from ideation to production to execution, uh, and operations. Uh, we, we've built,
24:53uh, uh, a content lab called, uh, CreaTech, which is a safe place for people to, to experiment.
25:00Uh, um, and we've been using NVIDIA WPP engine, uh, to build already, uh, over a thousand beauty images.
25:08I'm not talking about product pack shots because that's quite, uh, that's quite straightforward.
25:12Uh, and trying to redefine and create new codes of, of, of, of beauty. Um, but let me go back
25:18to,
25:18to media again, cause that's what I'm, I'm obsessed with. Um, you know, the, all the AI and Gen.AI
25:25solutions,
25:26uh, enable us to gain incredible time efficiencies. Uh, and then in turn, our vision is to reinvest
25:34a lot of, of that time saved, uh, in, uh, uh, building, uh, incredible idea, ideas for campaigns
25:41and make sure we build our brands in, in, in, in the long term. And I, I believe the past
25:4610 years
25:47in marketing and rightfully so, we're about just performance marketing. Uh, I believe the next 10
25:53years will also be about reinstilling more magic. Uh, and so oddly enough, we talk about tech,
26:00we're at Viva tech all the time. I believe that tech is going to enable us to bring more magic
26:04to
26:04the table because it'll leave us more time, uh, to do so and enable us to build our brands,
26:11not only in the short, but also in the long term. Okay. Thank you. Thank you everyone for your,
26:16your insight. I think that's really truly amazing to see how this kind of revolutionize customer
26:21experience and our works, um, in marketing. Maybe to conclude this panel, I think we'd like to,
26:28yeah, you'd take away with three main building blocks of what successful transformation, uh, be,
26:34can be led, uh, in the Gen AI, uh, space in marketing. First one is to make sure that you
26:40actually align the Gen AI strategy with your overall business strategy. This is super important. It means
26:45to have a clear value driven roadmap of use cases and avoid, and we have been seeing that too many
26:51times,
26:51super isolated proof of concept in all different domains. So clearly, uh, prioritized based on
26:57value and based on technical visibility. Second, um, is to make sure that you select one or two
27:03archetypes for technology implementation. There are many, many different solutions and technical
27:08solutions that are available. So it's super important to really have solid partnership with
27:13credible vendors and progressively developed internal technical foundations. And I think the main
27:19pitfall is to take those numerous technologies and really have a poor integration of both the internal
27:25and the customer data. So super key. And thirdly, as we discussed already, is properly to ensure the
27:31change management and the support of the workforce transition. As I was mentioning, to reskill people,
27:37identify the roles that are going to change, and make sure that you have reskilling from the technical,
27:42um, social and the emotional skills at the same time. So I think we have exciting times ahead. Um,
27:49I hope that this panel was interested to you, that it was more lively and more concrete than any
27:55generated tools that we would have delivered. And I wish you all a very good Friday and very good
28:00weekend. Thank you very much, Mathias. Thank you. Ellen and Sylvain. Thank you.
Comments