- il y a 5 semaines
Personalizing the Customer Experience with AI
Catégorie
🤖
TechnologieTranscription
00:00The emergence of generative AI from the niche to the mainstream is going to change fundamentally society's relationship with technology.
00:10The combination with traditional more discriminative AI, enterprise IT, data science, the possibilities of customer experience are just going to
00:22be immense.
00:24The opportunity is for hyper-personalization, improved customer satisfaction, improved brand engagement, but balanced with new risks of privacy and
00:35security.
00:37I'm thrilled to be joined on stage here with four amazing industry professionals.
00:45Let me introduce Prakash Ramamurthy from Freshworks, Samir Hajjali from Louis Vuitton, Sophie Heller from BNP Paribas, and Eric King
00:56from Amazon Alexa.
01:00Prakash, why don't you introduce yourself and tell us why this topic is important to you.
01:05Thank you, Andy. Prakash Ramamurthy, I'm the Chief Product Officer at Freshworks.
01:10We are a SaaS software company that offers customer support software for businesses to be able to give first-class,
01:21hyper-personalized support to their customers.
01:24And we'll talk a little bit about it when we get to that.
01:29Sumia.
01:31Hello, everyone.
01:33I'm Sumia Jali, responsible for client development and digital omni-channel for Louis Vuitton.
01:38And this topic is very important to us, since actually part of our culture at Louis Vuitton, creativity, emotions, and
01:50therefore people are at the heart.
01:52And we strive, we craft the most exceptional pieces, serving millions of clients all over the world, serving with a
02:00high level of quality.
02:03And whether it is for a pair of shoes of 1,000 euros or a piece of fine jewelry of
02:0910 million euros.
02:10So human is very important.
02:12And on the other hand, AI, if I were to simplify, is about software algorithm and automation.
02:23And it can be perceived as a human-less approach.
02:27And this is the reason, and where it is at stake for us is scale and efficiency.
02:33And I will have the opportunity to say more afterwards.
02:38Sophie, what's your role and why you're interested in this topic?
02:41So I'm Sophie Heller, I work at BNP Paribas, and my job is really to transform the customer experience using
02:48technology to adapt to our customer new behaviors and expectations.
02:52When banking, people expect banking to be simple, but they also expect that when they have an issue or when
03:00they have specific projects, they want to be treated and considered as unique.
03:03And this is where AI can help us, and banks in general, to be simpler and to be more personalized.
03:11Maybe three very simple examples.
03:13First, by providing the advisor with the next best conversation to have with a given client.
03:20Or to push personalized insight on the mobile app to help people better manage the money.
03:26Or finally, to have a 24-7 service to answer basic queries thanks to virtual assistants.
03:35So at the moment, we have chatbots, but Gen.ai offers the possibility to bring that customer experience to the
03:42next level.
03:43Thank you.
03:44And Eric, same question for you.
03:46What do you do, and why is this topic of interest?
03:49My name is Eric King.
03:51I run Amazon Alexa in Europe and in Asia Pacific.
03:56And at Amazon, we believe, like many of the panelists here, that this is the most transformational technology of our
04:04time.
04:05And it's going to touch every aspect of the customer experience.
04:09And at Amazon, we're really investing in multiple layers of the technology stack.
04:15At the base layer, we're investing in our own chips.
04:19This is for companies that want to build their own AI applications and large language models.
04:26At the middle layer, it's for companies that are interested in leveraging someone else's model.
04:32They may not want to build their own models, but leverage somebody else's.
04:35And so through a service we've built called Bedrock, we're giving customers choice to do that.
04:40And then at the application layer, we're building our own apps.
04:44We've got 60-some-odd in development now, including Amazon Alexa.
04:47And these are the customer-facing applications, which we'll talk about today.
04:51Thank you.
04:52So the topics of interest, whether you're a technology provider, whether you're providing consumer goods,
04:57whether you're providing high-end goods or critical services.
05:01It's a broad, broad range.
05:03We're seeing a real trend for more and more hyper-personalization,
05:08really targeted customer interface and interaction.
05:13Sophie, what's the approach the bank are taking to this?
05:17I think that in terms of personal data, we really need to navigate between the use of data
05:23and the additional value that it provides to the customer in terms of additional better service or better advice.
05:31So we take a very prudent approach.
05:34We tend to test first the interest of the client.
05:38So, for example, I was talking about this personalized insight to manage the budget.
05:43But we will say, would it be interesting in knowing how you could save on expenses?
05:48And if the customer says yes, then we will propose the service.
05:52As far as marketing is concerned, we have very strict rules.
05:55So whenever we will use tele-made, hyper-personalized profiling for marketing, we will ask consent before.
06:04And maybe finally, we spend a lot of effort to capture and to analyze the voice of customer on all
06:12channels and all customer journeys.
06:14So it's also a constant way to optimize and check if this is correct or not.
06:21And Samir, from a Louis Vuitton's perspective, how important is hyper-personalization to what you're trying to achieve?
06:27You know, our mission is to delight roughly 10 million, 10 millions of clients per year.
06:34And let's be honest, with 15,000 client advisors, there is no way to know each of them and personalize
06:44the experience at the fullest.
06:46So this is where AI comes into play because, of course, naturally, you tend to focus on the ones you
06:58know the best.
06:59And while each of them expects from us to know better than themselves what they may enjoy.
07:06So we use AI to augment our client advisors, whether in-store, in our flagships or any store, or our
07:19advisors in client services as well.
07:21So that we can use AI to hyper-personalize, to collect data, to analyze the client's behaviors,
07:29so that our client advisor's time, precious time, can be used to focus on building long-term relationships
07:39and help them to move from a reactive approach to a proactive one with a lot of authenticity.
07:47And as mentioned by Sophie, our approach is also to provide our clients with a trustful and meaningful experience.
07:58Trustful because as they consent to share with us their personal data, we owe a high level of responsibility.
08:07We know, for example, who they are, what is your wife's name or your kid, your wedding anniversary, for example.
08:15So this is very precious, and our client trust is more important than the fact of collecting data.
08:24So the use of data is really important for us, to us to really make sure that this trustful relationship
08:32lasts forever.
08:35Interesting.
08:36And Prakash, from a technology and a service provider,
08:41how important is hyper-personalization for you, and how do you know when to stop spending money in this area?
08:48I think in the support business, hyper-personalization is actually a necessity.
08:53It's not a nice-to-have.
08:56I'll give you an example.
08:57How many people here, show of hands, call customer support just to say everything is good?
09:04Nobody does that.
09:05But any time you call customer support, you already have a problem.
09:08So if the person, on the other hand, knows as much about you and your problem,
09:14they can serve you better.
09:16They can give you better comfort.
09:18Hoover saw the demo that came before this.
09:22The person's gentleman showed the Eiffel Tower written up,
09:25and AI was able to understand that.
09:28We won't apply that to customer support.
09:30Let's say you go buy a suitcase, and you take it on a flight, and it's broken.
09:35Instead of writing an email or calling them and explaining it,
09:39how about you take a picture of that and send it to the place you bought it from.
09:44If they know your phone number, if they can recognize everything,
09:48it can answer back in a nanosecond to say,
09:51oh, I see your luggage is torn, you're under warranty,
09:55I'm going to send you this replacement, and this is the address.
09:59They know your phone number, they know your order history,
10:03they know what you bought, they know it's under warranty,
10:06and they can respond back.
10:07That is a much, much better customer support experience,
10:11hyper-personalized, which gives a lot more comfort to the person who's buying it.
10:17So from a support perspective,
10:19being able to personalize the interaction for customers is super critical for businesses,
10:25and that's what we focus on in providing those capabilities for our customers.
10:29And Eric, you were telling me before the volume of devices that you have out in the world.
10:36On the hyper-personalization topic, can you ever do too much?
10:41Can it ever become uncomfortable?
10:42Is there a limit to what's appropriate?
10:45Well, we know it's incredibly important.
10:47As I was telling you backstage, we've sold about 500 million devices,
10:51and we have tens of millions of transactions every hour on Alexa.
10:56And what we know from listening to customers is the more customers feel like Alexa knows you,
11:03the more likely you are to use it.
11:05And so this started very simply for us with things like sense of humor.
11:10When we went from Spain to Italy to Saudi Arabia,
11:14we realized that the humor that we had trained Alexa on had to adapt to local norms and customs.
11:20Then we got into things like language, not just speaking different languages,
11:24but speaking different dialects.
11:26And we learned very quickly that in Italy, with many, many different dialects,
11:30the voice and the language and the way we said words in Tuscany
11:35had to be very different than it was in Rome.
11:38And so then we started to get to very specific individual personalization.
11:43And so when I walk into my kitchen in the morning, my device recognizes my face, my voice,
11:51and it gives me my commute information for the day, my news of the day, my schedule for the day.
11:58And that's very different than what my son, who's 16, is going to get two hours later when he gets
12:03up
12:03and interacts with the device.
12:05Thank goodness.
12:06And so all of these things, like we heard earlier, like Prakash said,
12:11we believe are sort of a necessity.
12:14I used to say when is too much, too much.
12:17And that's where we also think it's important to give customers choice.
12:21And so we've made it very easy for customers to turn off these controls if they don't want the experience.
12:26We believe very much that the more personalized the experience, the better.
12:29But we also understand that that's an individual choice from customers,
12:32and we'll give them the ability to toggle that to meet their own needs and preferences.
12:38Okay.
12:39You've all actually touched on, really, the next topic I wanted to look at,
12:43which comes down to the nature of that relationship between the human and the computerized machine,
12:49and more broadly, our relationship with technology that I hinted at at the beginning.
12:53And that's the subject of authenticity.
12:56What counts for authenticity?
12:58You talked about sense of humor.
13:01But I guess there are other elements that feed in, as well as regional precision of language.
13:08Let me stay with you, Eric, in terms of what authentic means for you.
13:12And perhaps you could go into a bit more detail about some of the other topics.
13:15We were talking about personalization.
13:19I think a slightly different take on that is the concept of personality.
13:23And we've worked very hard to give Alexa a particular personality.
13:28We're working on a new version today.
13:30This is going to be much more conversational.
13:33It's going to be much more two-way.
13:35But one of the things we're clear about maintaining are the personality traits
13:39that have come to be known as sort of the pillars of Alexa.
13:44Humble, humility, helpful, empathetic, inclusive.
13:48These are sort of the reasons customers tell us they like to use the device on a day-to-day
13:56basis.
13:58Part of that, in terms of building customer trust that we think is very important,
14:03is not only knowing the right answer, but also knowing when you don't know the right answer.
14:08So an important part of the systems we're building are to know when to engage and respond,
14:14but also know when you don't know the answer, the correct thing to do is not necessarily answer at all
14:21costs.
14:22It's what's an elegant off-ramp or what's a way to tell a customer that you need more information to
14:27answer it.
14:28And in our business, trust is absolutely key.
14:31And if we answer the wrong way or if a customer says lock the front door but we lock the
14:36back door,
14:36we've got a problem.
14:38And so building both a personality but also trust into the system,
14:44both knowing when to answer and when not, has been a sort of key area of focus for us.
14:49Yeah, I think you've got a really important point there.
14:51I work for an engineering business, and taking this technology into engineering
14:56where precision and correctness really matters is a core challenge for industry
15:01to move away and broaden the development of the tech from not just the correctness of the model
15:07but also the wider aspects of confidence.
15:10I'm interested, Sophie, from a banking point of view,
15:13what's your take on what authentic means and perhaps the role of the regulator, I guess, as well?
15:19Yeah.
15:19Of course, trust is probably the most valuable asset a bank has.
15:24So we do everything we can to maintain and to develop that trust.
15:28It means that all the models, all the AI models, have a human supervision.
15:34So it's extremely important for us.
15:37And also, we have really this cautious approach.
15:40So as I mentioned before, we already have chatbots in production.
15:44Now we are working on Gen AI virtual assistant.
15:49But we will not go into production if we don't have enough security
15:54that it's going to be perfectly correct.
15:58So, yes, we also prefer largely to say, I don't know, rather than hallucinating.
16:05And in addition to the trust, we are also heavily regulated.
16:09So that brings a lot of, I mean, not only AI Act, but also bank secrecy,
16:16but also e-privacy, GDPR, and all those things.
16:20So we have a lot of things about how we use widely the data, but also where they are stored.
16:27So it comes also with a lot of limitations.
16:29But I think that they are absolutely essential to make sure that the security of data is preserved.
16:37So the topic of authenticity is really quite far-reaching.
16:41And from a services provision point of view, what's your view on sometimes saying, I don't know?
16:47I think it's actually good.
16:49It makes it be real.
16:51So what we allow our customers to do, first off, we have protection within the AI capabilities we provide
16:58so that prompt engineering is supervised and all of that.
17:03And there are times we also allow our customers to see how the board is answering,
17:08where it is picking up the information from, because it's usually picking up from their repository
17:12so that if the answer is not accurate, they can go back and look at it to see, to correct
17:18that.
17:18So they have, it's like she was saying, it's very supervised from their perspective.
17:23The other way we help with authenticity is with the agents.
17:28I think Sophie talked about their agent assist telling them what the next best action is and all of that.
17:35If you provide the agents with more information when they are dealing with a customer,
17:41most of the time they are irate because that's when they're calling support,
17:45they get a chance to be more authentic and represent the brand of the company better
17:52because AI is helping them.
17:54For example, support centers have a very high attrition rate.
17:59That means one out of three, one out of four requests is going to a brand new agent
18:04who is probably terrified because they don't have all the answers.
18:08So authenticity can be interpreted in multiple ways.
18:11The bots being accurate, bots being able to say, I don't know,
18:16but also AI helping agents so that they can focus more on the customer
18:21and represent the brands well and be more authentic about it.
18:25I like the fact you've picked up on the emotional state of the user on your side.
18:31And Samir, that perhaps connects into one of the things you're talking about.
18:34And the value of, you know, with a high-end product, you talked about the emotion and the attachment.
18:41What's your view on authenticity with that?
18:43Authenticity is key and I fully agree.
18:45Trust and transparency are key.
18:48On top of that, we also use the collective intelligence, if I may say that,
18:56meaning that when we provide our client advisors with a recommendation,
19:00product recommendation, for example,
19:02they also have the ability to adapt what has been recommended
19:07so that the AI learns from our human advisors.
19:13And on the other way around, also they keep on learning
19:19because we launch hundreds of new products per week.
19:23So it's really a way to step up the game when it comes to really make the most of our
19:32inventory
19:33and make it suitable for all our clients because they are all different.
19:38And from a market to another, for example, China is a very specific market for us.
19:45So trends can be different from France or the US.
19:48So you take also into consideration external insights and trends and so on.
19:57So it's a never-ending process.
19:59And also, let's be humble, it's also an evolving process.
20:06And we keep on exploring, testing, learning, sometimes failing, and that's okay.
20:12And that's the way we have decided to move forward.
20:17So you've all talked about actually the quite wide range of stakeholders
20:22that these systems actually have to support.
20:25And it sounds like there's quite a lot of tailoring in terms of the different stakeholders
20:29or the local geographic variations, depending on what you're trying to achieve.
20:34There are places quite a lot of demands on the tech providers.
20:36You've got to build equipment that does that.
20:39Perhaps we can talk a little bit about the journey from producing first demonstrators.
20:47And this is my first time at CES.
20:50Sorry, CES.
20:51Viva Tech.
20:52Viva Tech.
20:53Viva Tech.
20:54And looking at the scale of the venue and the nature of the demonstrators is just breathtaking.
21:01But there's a journey, isn't there, from demonstrators through things that are good,
21:07the things that can become useful, things that actually really become positive and value-add.
21:17Sophie, can you talk a little bit about how the tech is developing in the bank
21:22and how you navigate that journey from good to useful?
21:27I think, actually, it's not very different from innovation in general.
21:32I mean, when you innovate, you don't innovate for the purpose of innovating.
21:36But you innovate.
21:37I mean, innovation is good as long as it provides value.
21:40So either for the client, either for the bank or for the society.
21:44So I think that it's good to just, it's common sense,
21:48but just to remind what is the problem or the issue you are trying to solve
21:52and can we solve that better with AI?
21:54So we have this very value-driven approach.
22:01We really start to, with the customer, what is the issue?
22:04So we invest a lot on non-technology things like UX,
22:09talking to people, understanding their problems.
22:12And then we test.
22:14We test, we take feedbacks, we improve continuously.
22:17And this is what we've been doing also for digital innovation.
22:21And I think it's the same for artificial intelligence.
22:25So it's both for the customer, of course,
22:29though the fact of having implemented also agile organization,
22:32when we can really try and deliver quickly something that we can test,
22:37put in the hands of the customer, get some feedback, make sure that it works.
22:41It's the same for employee, so be it digital innovation or AI, the same.
22:47So what is the problem?
22:48What is, what, how do we want to help them?
22:51And then does it work?
22:53How we adjust?
22:54So I think that by doing this, it's, it's, yeah, very quickly,
22:59if it doesn't work, you stop it.
23:00So we, yeah, you, you make sure that what you implemented is tested, safe and useful.
23:07And Samir, what's, what's your view?
23:09I understand the volume of products that you launch on the market.
23:12How do you manage that journey from good to useful across that suite?
23:15Yeah.
23:17We, we, we, we have decided to, to, to make choices.
23:21Bold choices start from the customers.
23:23And actually it implies to, to, to have a discerning approach when it comes to the customer experience
23:34and, and also use our AI capabilities.
23:38What is merely trendy topic versus what is a lasting value that we are providing our clients with?
23:47And, uh, uh, to be clear, the, the three, the, our top three priorities focusing on the client experience
23:54is firstly, um, of course, personalization.
23:59So know our clients better to serve them better.
24:02And, uh, and, uh, and, uh, not everywhere, but personalization at the right moment to, to,
24:08at the right place, uh, with an omni-channel approach, which, uh, uh, lay me to the, to the, to
24:14the point number two.
24:15Uh, we strongly believe that, uh, uh, uh, an innovative, modern and, uh, up-to-date retail experience,
24:22luxury retail experience is to, to, uh, to provide our clients with a seamless approach
24:28and, uh, and they decide, uh, where they want to purchase.
24:31Our role is to really trigger, to, to inspire them.
24:34And, uh, at the end of the day, they will, um, end up with a purchase, uh, on, on, on
24:40our, in our stores,
24:41our calling, our client service, or purchasing online on the web, on the app,
24:46or a mini-program we chat in China, or on Kakao in Korea, for example.
24:51And last but not least, our third point is, I was, is, uh, uh, about, uh, really pushing, uh, uh,
25:01augmenting our capabilities in our client service once again.
25:04So, for example, we have leveraged the speech-to-text technology, uh, so that we can, we have 14,000
25:13calls
25:13or contacts to, with, uh, our client services, uh, per day in 24 hours, uh, all over the world.
25:20So, uh, we, we are currently, uh, uh, uh, experiencing this, these, uh, uh, technologies to, to, to, in real
25:28time,
25:29provide with our client advisors with, uh, uh, uh, uh, uh, increasingly, uh, amount, uh, of, uh, of, uh, of
25:38data.
25:38And, uh, so that they can, uh, uh, uh, have, in real-time recommendations, so that they can adapt, uh,
25:45their, uh, response
25:46and be, uh, relevant, more and more relevant as long as the machine is, uh, learning
25:52and also managing the cultural, uh, differences and languages.
25:57Thank you.
25:58And, Prakash, your view on how this technology, can it be, is it about operational effectiveness
26:05as you make your way from good to useful, or is it also differentiating in the market for you?
26:11Yeah, both of them. Um, for example, at Freshworks, we have been investing in AI for the last five plus
26:17years.
26:18Uh, but what has happened in the last 18 months, it, it's a pretty monstrous wave, and, uh, our goal
26:26is to ride it.
26:27Uh, just in the last 10 days, if you look at it, what OpenAI announced, what Gemini has announced,
26:32what Bedrock has been working on, we have the ability to plug in, the way we have structured our, uh,
26:40architecture
26:40is to be able to plug and play all of these models, then train it with our, uh, information,
26:45train it with the customer's data for them to get benefit out of that.
26:49So, it is, it's, it's a little bit of an arms race, uh, uh, right now to be able to
26:56provide these capabilities
26:57for the customer, but the thing that excites me the most is, for the longest time, as software has been
27:05built,
27:06humans have been trained to understand the software, right?
27:11If you, if you look at any, any time, uh, software gets rolled out, the agent has to be trained,
27:16uh, to use the software.
27:17You understand the quirks of the software.
27:20With AI, there is an opportunity to kind of us gain back the pecking order, uh, and make AI understand
27:27us.
27:28So, we just ask it to do the things.
27:30It goes to figure out which configuration it needs to tweak, whatever it needs to do, as opposed to us
27:37knowing
27:37this is how you sort, this is how you pick the information, which is how we have been, uh, drilled.
27:42So, that, I think, if, if software companies adapt that and provide that to the customers, the ones who do
27:49that will be able to differentiate.
27:51And I think that's where we feel we can leverage this enormous amount of innovation that's happening in such a
27:57compressed period of time,
27:58which is, it, it's one of those waves that happen once every 10, 15, 20 years.
28:03And we just happen to be in the middle of it, and we want to make the most use of
28:07that, uh, to deliver value for our customers.
28:09So, it has gone from an interesting science experiment to really being able to build differentiated solutions for our customers.
28:18I agree.
28:18And the pace, obviously, is, uh, is significant.
28:21And, Eric, I guess with, with the product, the, the nature of the interaction, I guess, is very closely attached
28:27to, to the brand.
28:28What, what's your view on how you use this to build the value in the brand and to be differentiating?
28:33Yeah, I, I agree with Prakash.
28:36We've, we've, we've gone from science experiments to, to utility for sure.
28:40Uh, when we launched Alexa 10 years ago, it could do 13 things.
28:44It could play a song.
28:45It could set an, uh, an alarm.
28:47It could do a timer.
28:49Um, I had the fortunate, uh, the good fortune to visit the Red Cross in Spain last week.
28:56Uh, and this organization has deployed Alexa devices in thousands of elderly customers' homes across the country.
29:02And those devices are being used not just to sing songs, but to keep, uh, the client's homes safe and
29:10secure, to remind them when to take critical medicine, to keep them in touch with their caregivers.
29:17An important side effect, it's being shown to combat, uh, loneliness, uh, which we know is, uh, can precipitate cognitive
29:24decline.
29:25So, even with the technology we've got today in the box, we're seeing this go from sort of fun gadgets
29:31to sort of life-enhancing technology in many cases.
29:36Um, you asked about differentiation and, uh, with sort of, as I mentioned before, we're working on a new version
29:43of Alexa.
29:43Uh, one of the things we talk often about is going beyond gathering and presenting information to taking action.
29:53And that means things like connecting to hundreds of thousands of smart home devices.
29:59Uh, it means not just being able to book a vacation.
30:03I'm sorry, research a vacation, but book a vacation or plan a meal and then order your groceries.
30:08Uh, and so there's a lot we have sort of within our roof at Amazon where we're looking at and
30:14saying, how do we go from helpful assistant that can be fantastic to research a paper or write a poem
30:21to actually make my life easier and do something?
30:25And so that's what we see as one of our key assets in the investment areas for us.
30:30No, very interesting.
30:32Thank you.
30:33Let's move on to, um, perhaps the, the, the flip side of some of these conversations.
30:37We've talked about the, the positives, the engagement, authenticity, um, clearly within the industry, there's also, uh, the reverse side
30:45of it, which is the, the threat, um, of bad actors, of people using the technology for, for reasons that
30:51we don't want them to.
30:52I'm interested, um, uh, Sophie, what's, what's, what's, what's your view on this?
30:57Is this something that do you, uh, do you, uh, do you, how do you actively manage it?
31:01Is this something you manage inside the bank?
31:02Is it at the interface?
31:03What's, what's your view?
31:04Well, I think it's, uh, it's a major concern, but I think for everyone, uh, public institutions, uh, all companies,
31:11um, especially for banking.
31:13So maybe the first thing is that we have very strict frameworks about how we can use the data and
31:20with, which type of solutions and where they are stored and processed and et cetera.
31:24So we are extremely strict at it also because it's regulated, but also for our own, uh, policy.
31:30Um, and second, I think that, um, as you say, with the technology, you have new fraud schemes, new fraud
31:37techniques every day.
31:38So you need to invest a lot of the investments are dedicated to cyber security team that are detecting new
31:45scheme, um, avoiding them, protecting the bank.
31:49But there's also huge effort to train all employees because it's, it's really the job of every employee also to
31:57protect the bank.
31:57So we need to maintain a very strong level of awareness about those topics to all employees, uh, front to
32:04back.
32:04So to make sure that we try to avoid, uh, all those threats.
32:10Yeah.
32:11I think education, I think is, is, is going to increasingly become a key theme for all of us.
32:15I think particularly governments and wider education on the topic.
32:18Because I think, well, we're all aware of the amount of noise around the topic.
32:23Not all of it always accurate.
32:25Samir, what's, what's your view on this?
32:27Is this something that, um, given the interaction with the advisors, is, is it less of a worry for you
32:32or is it something that you pay attention to as well?
32:34Yeah.
32:35I just wanted to jump on this, uh, training topic.
32:37I was in Singapore last week meeting the teams.
32:40Half of the questions were about what are we, what do we plan to, to do when it comes to
32:45AI, gen AI and so on.
32:46And especially the younger generations who are already used to use, uh, uh, this technology, you, you, on our everyday
32:53life, when you use Google map, you use AI actually.
32:56So, so, and the presentation we, we witnessed just before with the open AI, uh, uh, for O, Omni.
33:05So it was really impressive and it gives a, it provides us with a lot of, uh, ideas to explore.
33:11Coming back, uh, to the phrase, uh, of course, um, we need to manage.
33:16The, the, the, uh, the, uh, the, the, the concerns of, uh, uh, uh, our clients, but, uh, also both
33:23our clients and our, uh, our, uh, uh, teams.
33:27And, uh, on that, I think that we need to partner, uh, with, uh, I don't know, the, the, the,
33:33the governments, the institutions and, uh, uh, uh, and so on, uh, to first, uh, set the right standards, uh,
33:40to, when it comes to data privacy.
33:42And, uh, and, uh, and so in Europe, we have GDPR in, uh, in the U.S. CCAA and so
33:47on, uh, but also, um, uh, ensure with an ethical use of AI, if I may say that, uh, to
33:55avoid any biases.
33:57And, uh, uh, we know that, uh, that this is really important because behind the machine, behind the software, uh,
34:05there are a lot of, uh, data scientists, talented data engineers and so on.
34:10Uh, uh, sorry to say that, but most of them are men, uh, and, uh, uh, so, so, uh, uh,
34:17that's the reason why we need to also, uh, uh, uh, invite, encourage our daughters to, uh, to really follow
34:24this path.
34:25I have a son who is, uh, data scientist, for example, and he's part of our daily discussion, so, uh,
34:31that's a point of, uh, of, um, of concern.
34:34And also, uh, promote transparency and, uh, and educate the people and, uh, and, uh, and our, our teams.
34:42But I think also that it will, uh, it will be much easier for the next generation.
34:48Thank you.
34:49And, Prakash, the threat of bad actors for you, is it, is it in your scope or, again, at a
34:53boundary issue?
34:55Oh, absolutely.
34:56It's in our scope.
34:56Um, our customers expect us to make sure that we keep their data safe in providing them value.
35:04So, uh, we take a lot of effort in making sure each customer's data is completely siloed from the others.
35:11That's one aspect.
35:12Uh, we also have a lot of, uh, security framework around protecting the prompts and prompt engineering so that a
35:19bad actor cannot, uh, cannot break into that.
35:23Um, so, all of those things are part and parcel of this.
35:28And, uh, that's the expectation a customer has from us.
35:31And we have all the safeguards we can internally in our system when we deal with customers' data.
35:38All the things that she was talking about in terms of GDPR and others, we respect all of that.
35:43Uh, but most importantly, we, we make sure that we are the absolute guardians of our customers' data while we
35:51provide them the value of using AI.
35:54So, that's integral to how we develop our capabilities.
35:58It's integral to how we deploy it, uh, on the cloud.
36:02And, Eric, this must be key for you.
36:03It is, and, uh, many of the same themes we heard before are, are, are, are exactly what we talk
36:10about internally at Amazon.
36:12Maybe an interesting take that, that I haven't heard yet is how we're using the generative AI tools and the
36:18ML systems on our own platforms, uh, to keep customers safe.
36:23So, an example of that is how we're using these tools on our retail website, where sort of trust is
36:29queen or trust is king, um, we're using, uh, very similar tools to the ones we're talking about to, uh,
36:36deal with fake reviews.
36:38This is a huge problem, uh, in the industry today.
36:41And, uh, last year we used, uh, our own ML algorithms to, to take down 200 million, uh, fake reviews.
36:48Uh, similarly, fraudulent counterfeit products on the website.
36:51Uh, we used these systems, uh, to monitor, uh, billions and billions of, uh, page changes to different retail websites
37:00on a daily basis.
37:01And, uh, we're able to use AWS tools and a suite of those to catch about 99% of, uh,
37:09challenges before they even get to the vendor.
37:11So, we're sort of walking, we're eating our own dog food, as we say, using, uh, using the technologies on
37:16our own, uh, retail websites.
37:18And the other thing that I'm encouraged by is on the education front, we've also believed that educating not only
37:28internally and our customers, but the next generation of employees has been super important.
37:34And so with initiative we've got called AI Ready, we've made a commitment to train 2 million sort of non
37:42-Amazonians on generative AI technologies in 2025.
37:46So that's taken partnership across the industry, it's taken partnership with state and local and national governments.
37:53So just some of the unique things we're doing at Amazon, but many of the customer promise, data safeguard things
38:01that I heard from my colleagues here are absolutely core to our investments.
38:05So unpicking the topic of new approaches to customer engagement, there's a vast set of underlying topics that we all
38:13have to deal with.
38:15We're clearly, as we begin to close, we're making our way, we know, still through large amounts of the hype
38:20curve.
38:21If I can ask you with your closing comments, and let's start with you Eric and come back down the
38:26line, what do you see that you think will survive the collapse of the hype curve?
38:31In terms of the types of things you're developing, what do you think will persist?
38:37I'll just say three things.
38:38It'll be a little bit repetition, but hopefully it'll sort of wrap up some of the things we've talked about,
38:43and I'll speak specifically about Alexa.
38:45The first is conversational.
38:47You'll see us invest a lot in real-time two-way communication that feels a lot more human-like, and
38:54that's a very important part of what we deliver.
38:57Second is connected, and by that I mean real-world action.
39:02I mentioned that before, but we want customers to be able to do something with that data, take an action,
39:08and this goes beyond playing a song, but like I said, it goes to home control, ordering food, planning a
39:13vacation.
39:15And then, finally, the theme of this panel, which is personalization.
39:20That's going to continue to be a huge investment area for us.
39:22Thank you.
39:23Sophie?
39:24I would say that we've started to invest in AI long before the Gen.I. hype.
39:31This is why we have now, thanks to a very cautious and value-driven approach, over 700 use cases in
39:37production.
39:37Now we are investing a lot in LLM and Gen.AI.
39:41This is also why we've been part of the roundtable of Mistral, and we have a partnership with them.
39:46And by the way, you are all welcome to look at our corner.
39:49It's just in front of the audience, and Mistral is there.
39:53And in the future, so we'll continue to invest also a lot in our talent development.
39:58So we have other 3,000 data experts.
40:01We keep also on recruiting new staff, upskilling the staff, but also, as you say, training the whole workforce about
40:08AI.
40:08So I think it's, for us, it's really a long-term commitment and investment.
40:13But on the other hand, it's just the beginning of the journey, because all the examples that we have, it's
40:19better do what we already do.
40:21But AI is also about doing new things that we don't even know, and this is to be started.
40:27Yeah, I agree.
40:29Samir, what are you going to take forward?
40:30So first and foremost, AI, personalization data, and trust are synonymous.
40:38Personalization and transparency are synonymous.
40:41And it's a never-ending topic when it comes to learning.
40:46Point number two, we need to really keep on learning, keep on teaching, keep on coaching our teams, keep on
40:55coaching the AI as well.
40:57So, and last point, we strongly believe, and this is a strong conviction, that AI and human are complementary.
41:10So AI will never replace our client advisors or our touch base with our client, which is the most precious
41:20asset at Guivito.
41:21Well said.
41:22And Prakash, what's going to survive the hype curve in your area?
41:25First off, I think with this, the hype curve itself is going to be compressed, because we're very soon going
41:33to get past the disillusionment,
41:37because we already know what it's capable of doing, and there's features and capabilities being delivered.
41:45So just like we don't talk about Internet or smartphones anymore, because we just think that's just how life was
41:51supposed to be,
41:52I think in about three years, we won't talk about multi-turn conversations with AI and smart answers and all
42:00of this stuff.
42:01We just said, hey, why did it take us so long to figure this out?
42:04And it'll become the way of life, and the companies that adapt that and serve the customers are the ones
42:10who are going to survive.
42:11So I think it's going to become real very, very soon, if not already.
42:17Well said.
42:18It's not often, I think, in one's career, one finds oneself working in the middle of a genuine industrial revolution.
42:24Thank you, my colleagues, for taking part in this conversation, and enjoy the rest of the conference.
42:30Thank you.
42:31Thank you.
42:31Thank you.
42:35Thank you so much to our panel there, and thank you to Andy for that fantastic moderation.
42:41Okay, we're going to stay focused on the theme of AI and human interaction with our next two sessions,
42:48both of which are going to be looking at how we can upskill and reskill both our workplaces and our
42:55workforces.
42:56That's starting here in just a few seconds.
42:59Thank you.
Commentaires