- 16 hours ago
"As AI agents become part of everyday operations, hospitality is entering a new phase of transformation. The question is no longer whether AI can automate tasks, but how humans and intelligent systems should collaborate inside organizations built around service and human interaction.
In this fireside chat with Siddhartha Chatterjee, Global Chief Digital, Data, AI & Technology Officer at Club Med, we will explore the rise of human + agent organizations, the realities of responsible automation, and the tensions between operational efficiency and emotional connection. In an industry where trust and experience are central, what should remain uniquely human?"
In this fireside chat with Siddhartha Chatterjee, Global Chief Digital, Data, AI & Technology Officer at Club Med, we will explore the rise of human + agent organizations, the realities of responsible automation, and the tensions between operational efficiency and emotional connection. In an industry where trust and experience are central, what should remain uniquely human?"
Category
🤖
TechTranscript
00:11Good afternoon, everyone.
00:14Today we're going to talk about humans, agents, and the future of the hospitality industry.
00:20My name is Upton. I'm a former CNBC journalist, and currently I'm a content creator.
00:26I'm on TikTok, YouTube, making videos about tech and economics, and I'm also anchoring Viva Tech News every morning, which
00:33is upstairs on the third floor.
00:35Come say hi.
00:36So today we're going to be talking about AI agents, and as they become part of everyday operations, hospitality is
00:45entering a new phase.
00:46So the question is no longer whether AI can automate tasks, because let's be honest, we know that it can
00:52automate a lot of our tasks.
00:54The question is now how humans and systems should collaborate inside organizations.
01:00So I'm now excited to welcome Siddhartha, who is Global Chief Digital Data, AI, and Technology Officer at Club Med.
01:10Siddhartha, welcome to the stage.
01:12Thank you, Uptin. Thank you so much.
01:15How has your day so far been at Viva Tech?
01:20The day has been amazing. This is an event that I love, and for 10 years I've been following it.
01:26Last time I took the stage was in 2023, just after the arrival of ChatGPT.
01:33And I was talking about the potential that we have with AI, and after three years, it's just incredible what's
01:42been happening, and very happy to be here.
01:44Amazing. So let's start with Club Med. Let's talk about, we all know Club Med, but I want to know
01:49who is Club Med today, what is it like, and how is AI impacting and helping with tourism?
01:57Sure. So Club Med in France, you barely need to explain the brand.
02:04We are a pioneer. We are existing since 75 years.
02:08This year, last year, we celebrated our 75th anniversary.
02:12We invented the all-inclusive vacation concept.
02:15Today, we operate almost 70 resorts in 40 countries, 23,000 employees, 110 nationalities of clients, visitors, 1.5 million
02:25clients.
02:26And at the core of our value proposition is the human connection, so human touch.
02:31So that's where I think it's very interesting, you know, the way AI is allowing us to reinforce the human
02:39touch within hospitality.
02:41And to say in a nutshell, because it can be very complex, the goal of AI in tourism is to
02:48remove friction.
02:50Friction both at the customer side and also on the employees, you know.
02:57Because tourism is an extremely demanding sector, business.
03:01You know, there are lots of things to do.
03:03So AI can really help in removing that friction, and that's what we have been doing so far.
03:11And when you say removing the friction, can you elaborate on that?
03:17Yeah, sure.
03:19So the friction you have to look at from two sides.
03:22One is the customer journey.
03:26So today, you know, and customer journey is important because if you look at Club Med, our average basket is
03:32about 5,000 euros.
03:34So it is not a commodity transaction, you know.
03:37Our clients go through 11 touch points in order to make a transaction over almost between 70 to 100 days.
03:45And hospitality products, especially Club Med resorts, are quite complex to navigate.
03:50So there is, you know, there is discovery, there is booking, there is a lot of intermediaries, and then there
03:55is this topic of pricing.
03:57So how can we use AI in order to make those, all those steps as seamless as possible so that
04:05a customer can just express his desire?
04:09And the AI is able to do all the back-end work and get him closer to his dream vacation,
04:15you know, something we are doing with our use case, GM co-pilot within WhatsApp, and I'm going to come
04:19to that later on.
04:21So that's one example.
04:22And second is, from the employee point of view, the goal is to use AI and give back as much
04:32time as possible back to the employees so that they can reallocate that time in order to build on the
04:38service, the relational, the emotional aspect, you know.
04:42And as much little back-office, manual, you know, mundane task as possible, you know, so that's the goal, and
04:51that's what we are doing, and which will allow you eventually to increase your service quality, you know, make more
04:56acquisition, increase your conversion rates, and also grow the company without actually growing your staff in a linear fashion.
05:03So you talk about building a human plus an agent organization.
05:08So you have humans, you have AI agents, but I want to know, what is the strategic vision behind that,
05:14and what does it already look like in practice right now at Club Med?
05:20Yeah, absolutely, and I have Andrew in front of me.
05:23He's my boss, and he says something.
05:27Hello, Andrew.
05:28Wave your hand, Andrew, so we all know.
05:30And my great team is also here.
05:32So, but something we believe at Club Med, and we are really proud of it, is that we were very
05:39quick in anticipating that we can't use AI to optimize existing broken processes and use it as a tool, okay?
05:49The current capabilities of AI, which is the capacity to reason, to plan, to execute tasks, will be, which will
05:58mean that in five to ten years, companies will no longer operate the way they operate together.
06:04So, and that's what Andrew says, is that you need to shake up the organization, you know, and make the
06:10AI drive the business.
06:11But at the same time, you need to embark the humans or the employees in order to do so.
06:19So what we do is a very structured process called process re-engineering or process design.
06:24So for every function, we have dedicated profiles called process designers who are in charge of mapping their entire operating
06:32model,
06:34and then, along with experts, redesigning that operating model with agents, part of that operating model.
06:41And stop looking at AI as a tool, but rather as a digital collaborator, you know?
06:47So this is what we are doing function by function, whether it's HR, whether it's finance, commercial, sales, sales.
06:53And then, obviously, with the right kind of teams, right expertise, starting to put in place this new operating model,
06:59you know, which takes time.
07:01But over a period of a few years, you know, I'm pretty confident that we'll be able to reinvent totally
07:07the operating model of Club Med,
07:09where humans and AI agents will collaborate together and not just use AI as a tool.
07:18So responsible, quote-unquote, responsible automation sounds kind of abstract to a lot of us.
07:25So can you give us a concrete example of automation without dehumanizing?
07:30Kind of like what you're saying right now.
07:32It's not just going to be a tool, but a practice, so to speak.
07:37Like, yeah, so there is a big myth within the world about AI replacing the human being, okay, the human.
07:51This is not going to happen.
07:53And if I come back to my first point, which is AI is meant to remove the friction, okay?
07:56So I'll give a concrete example.
07:57If you take our use case, which is an industry pioneering use case called GM Copilot,
08:03which is an AI orchestration agentic platform, which is deployed in WhatsApp live in 23 countries,
08:12with a rate of autonomy of 48%, means one-third of all incoming customer queries are coming within WhatsApp,
08:21and this AI agent is able to answer or handle in full autonomy almost half of this, okay?
08:28Now the question will come that when an agent is having so much of autonomy,
08:33what happens to the human or the travel agent?
08:36You know, we call them the travel experience designers.
08:39Now what is happening is now the travel experience designer is moving to a more emotional connect with the customer.
08:48So he has more time to listen.
08:50He's not just looking for responses.
08:52He's building inspiration, right?
08:56So, and inspiration and emotion is what actually moves the needle when it comes to our clients.
09:02You know, so they are not interested just to know whether there's swimming pool at Punta Cana
09:06or what is the timing of Kids Club.
09:09So these are all information that you have within a database,
09:12and an agent can actually take that information and give it to the client, okay?
09:15But what the client is looking for is having a real connect with our travel experience designers
09:20to discuss their feelings, their aspirations, their memories.
09:26You know, what is the memory that they want to create with their family when they go to a Club
09:30Med resort?
09:30And these kind of technology gives that time back.
09:33So that's one example.
09:34You know, I'll give you another example, which is internal side.
09:38You know, we are in a seasonal business.
09:40So every season, we need to reassign our GOs, 23,000 GOs across the world in different resorts.
09:48Now, for 20 years, this has been a manual process.
09:52So you have 50 to 100 people who spend half of the time,
09:55who used to spend half of the time to manually reassign GOs to different resorts.
10:00Now we have put an AI in place where that reassignment is done automatically,
10:06and what do our HR business partners, they do with this 50% of time that they are gaining?
10:13They use it to come up with strategies, to use their critical thinking, to use their empathy,
10:18to use their, you know, creativity,
10:21to build strategies for the career development of our GOs and our employees.
10:25You know, and no longer in manual processes, administrative processes.
10:30So I'm giving you a couple of examples, which highlights how it is not dehumanizing it,
10:35but on the contrary, it's actually humanizing more.
10:38So your example on WhatsApp, just to clarify,
10:42that would be somebody, like, messaging the WhatsApp team,
10:47the Club Med team via WhatsApp, correct?
10:50Yes, so messaging is becoming a future channel for our customers.
10:56In China, this is a case for years now with WeChat,
11:01and WhatsApp is the competitor in the rest of the world,
11:05and I think I have partners from Meta also over here.
11:08Hello if you're here.
11:10And why is messaging becoming a promising channel for the future?
11:14It's because of the asynchronous nature, right?
11:16Me as a client, as a premium client,
11:18I don't want to call up the hotline and stay online 10, 15 minutes
11:22before I join a travel agent, right?
11:24I want to send a message saying that, you know,
11:26I'm having this dream of a vacation.
11:28You know, can you give me some information, some recommendations,
11:31some pricing, et cetera.
11:33So earlier, and more and more traffic is moving towards that
11:35in terms of customer relation,
11:37and earlier there would be our travel agents
11:40who would actually give the answers manually,
11:42and for that they would have to go on to different systems.
11:44So today there's an agent, this agent, GM Copilot,
11:46is answering queries on product,
11:49answering on pricing, and also doing simple recommendations.
11:52You know, so that's already saving a lot of time for our clients.
11:56And after that, if they want to continue discussing
11:59to go more deeper into the whole plan,
12:02they always have the ability to join a travel experience designer.
12:05And that is exactly in luxury.
12:07This will be the role of AI.
12:09You will always have AI to win time and go faster, remove friction.
12:12But if you want to talk to a human,
12:15there will always be a human that you can connect with.
12:18So that's a really great example of, I guess,
12:22if you think about your sales funnel,
12:24someone who's interested enough in Club Med,
12:27they're already quite, you know, integrated in the funnel,
12:31and now they can talk to the AI agent on WhatsApp.
12:35How about before they even get to Club Med,
12:38i.e. they're on ChatGPT or Claude or Google Gemini,
12:42and they're, you know, what are you doing
12:44and what is your team doing to ensure that Club Med
12:47is spewing out the results as a consideration for their vacation?
12:52Now, this is another hot topic in generative AI.
12:56And why is it a hot topic?
12:57Because this touches the distribution model, okay?
13:01And in terms of, so we, Club Med, we are a direct D2C brand,
13:05as in our customers are in direct contact with us.
13:09More than 70% of our sales are generated through our direct platforms,
13:13whether it's online, whether it's offline, you know,
13:15call centers, website, e-commerce, WeChat in China.
13:18They are all, their 70% is direct.
13:21And the way we are acquiring our customers is through SEO,
13:25through SEA, through display advertising, et cetera,
13:28the classical method.
13:29But now what we see is, and which is normal,
13:32you know, the customers are moving towards the LLM platforms,
13:36ChatGPT, Publixity, Gemini, DeepSeek, et cetera, et cetera.
13:40Even though the traffic that's coming towards our channels,
13:43direct channels is quite low today,
13:46but we can see an exponential growth in the traffic.
13:49And there are some, there's a statistic of booking.com,
13:53which says that within 2030, more than 30% of, you know,
13:57transactions will happen through these channels
13:59and by agents doing itself, okay?
14:01So we are building a very structured strategy to counter this
14:05because we want to keep our direct distribution,
14:07but at the same time, we want to be competitive
14:09in terms of pricing,
14:10but also in terms of visibility on different platforms.
14:14So the first step for that would be to be visible in the AI models.
14:18And we have done our initial research,
14:21and already we are pretty good, well-placed
14:22because we are among the top three results
14:24when there are questions in our field on tourism,
14:28all-inclusive, luxury, et cetera.
14:30We are within the top three brands,
14:31but there are ways by which we can continue
14:35to further enrich our content on our own platforms,
14:37whether it is a website, social media, et cetera.
14:41And what we call is, so I have France here
14:44who's working on this from my team
14:46to make this content machine-readable, you know,
14:50and be accessible through protocols such as MCP.
14:55So that will allow us to be more visible on these models.
14:58And once that visibility is there,
15:00then we have to think how we can, you know,
15:02continue to keep that direct contact with the client
15:05and not delegate the whole journey
15:08or the booking journey to the LLM platforms, right?
15:12So that's a different topic of how to protect our margins
15:15and keep that direct transaction with the client.
15:18I see.
15:19So in hospitality, efficiency and emotional connection
15:24are often opposites, right?
15:27They're in the opposite direction.
15:28So how do you manage that tension
15:30between efficiency and that emotional connection?
15:37So it's quite a difficult question to answer,
15:42but in a easy, you know, if I want to summarize,
15:45give a short answer to that,
15:48is all that we have been doing so far
15:52is in order to do that, you know,
15:54in order to balance the power of the technology,
15:58the automation and the human connection.
16:00So one example I will give you,
16:01in 2023 when we decided that Club Med will go aggressive
16:05with the deployment AI
16:06and we want to be benchmark setters
16:09and not smart adopters or followers,
16:12we had a dichotomy.
16:13And the dichotomy was that Club Med is a love brand.
16:16So in France, people wear T-shirt written
16:19with the logo of Club Med.
16:21You know, people go as far as creating tattoos of Club Med.
16:24So we are actually not a holiday brand.
16:26We are a lifestyle brand.
16:27And we want to be the most desirable lifestyle brand
16:30in the world.
16:31This is a very big moonshot ambition.
16:33Ambition.
16:34So we said how, there's a dichotomy,
16:35even though as experts we understand the close synergy,
16:38but how do we do?
16:39So the approach was ethics.
16:41You know, so ethics has to be
16:42at the core of everything we do.
16:45And that's when we created our ethics committee,
16:47chaired by Professor Jean-Gabriel Ganasia,
16:50who's doing AI for the last 40 years,
16:52when the term AI was outcasted.
16:54You could not use the term AI, right?
16:56And he's not only an expert in AI,
16:58but he's also an expert in digital ethics
17:00and a philosopher, okay?
17:03So we have a quarterly committee
17:05with Professor Ganasia,
17:07with the CTO, with the CISO,
17:09with the HR,
17:10all the stakeholders.
17:11And then we take a list of our initiatives
17:14or our ambitions with AI,
17:15and we say,
17:17is ethically,
17:18are these initiatives aligned
17:20with the values of Club Med?
17:22And the values of Club Med are many.
17:24So you have kindness,
17:25you have multiculturalism, etc.
17:28Okay, so we are a very progressive brand.
17:31You know, so,
17:33and ironically,
17:34you know,
17:34many people told me
17:35that this ethics concept
17:36will actually slow you down
17:38and block all your projects.
17:39But the outcome has been totally opposite.
17:42It has become a business enabler,
17:43you know,
17:44because it allows you to think
17:45a little differently.
17:46I'll give an example.
17:47We still have 12 minutes left,
17:48so I can give some examples.
17:49I know everyone loves examples,
17:50then a concept.
17:54Recruitment is a core topic for Club Med.
17:56Why?
17:57Because our GEOs
17:59are the core
18:01of the value proposition
18:02and the client delight, okay?
18:05So that means that
18:06we need to recruit
18:07the best GEOs
18:08in France,
18:09but also all over the world
18:10and then move them around
18:12in the world.
18:13So recruitment
18:14is actually
18:15a piece of business
18:16as important as sales
18:17and marketing, okay?
18:19Now, in order to...
18:20And because the brand
18:22attracts
18:24GEOs,
18:24so we have more than
18:252,000 CVs
18:26and applications
18:27which are sent to us
18:28every year.
18:30Now, humanly,
18:31it's not possible
18:31to read 200,000 CVs, okay?
18:35And there are statistics
18:35which show
18:36that only 30% of CVs
18:38are actually processed, you know?
18:40So in this case,
18:41human is more biased
18:42than an AI.
18:44All right.
18:44So we decided
18:45we are going to use AI
18:46and automate
18:48at least this first phase
18:49of screening
18:49of the CVs.
18:51Then we went
18:52to the ethics committee
18:53and the ethics committee,
18:54the professor told us,
18:56what you're trying to do
18:57is really good,
18:57but you can't use
19:01large language models
19:02to analyze those CVs
19:03because if the large
19:04language model
19:04does a ranking
19:05of the CVs
19:07and there is some
19:09anomaly in the ranking,
19:10you can't explain
19:12why
19:12because you haven't
19:13built that model.
19:14It's Sam Altman
19:15who has built that model,
19:16right?
19:17It's a small tip,
19:18but ethically
19:20you can do it, okay?
19:21So what we did is
19:22we went back
19:23to our technical teams
19:24and we took
19:25an open source
19:25much smaller model
19:26which works very good
19:27in CV screening
19:28and we applied it
19:30and we can see
19:30the source code
19:32and we are now
19:33in the verge
19:34of deploying it
19:35to screen the CVs
19:36which is just one step
19:37of the whole process
19:38of recruitment,
19:41but you see
19:42how an ethics approach
19:43can be an accelerator
19:44otherwise we would have
19:45gone with an LLM
19:46based approach
19:47and after some time
19:48there could have been
19:48an audit,
19:49external audit
19:50of a shareholder
19:51or of auditors
19:52like the European Commission
19:53and we might have
19:55had to roll back.
19:55So there are many decisions
19:57like these small decisions
19:59that we are taking
20:00thanks to an ethical approach
20:02which is allowing us
20:03to go faster.
20:05So trust is everything
20:06in hospitality.
20:08What does it take
20:08for a traveler
20:10to actually trust
20:11an algorithm
20:12and how do you
20:13build that in?
20:16So travelers,
20:18travelers already,
20:20you know,
20:21the research shows,
20:23again,
20:23booking.com
20:24is referenced in the market,
20:2660% of our customers,
20:28travelers around the world
20:30are using AI
20:31to plan their vacation
20:33or their trips
20:35and 61% are using it
20:37during their trips.
20:39So let's keep it in,
20:41let's be very clear,
20:42I want to be very clear,
20:43our travelers
20:44and consumers in general
20:46are already AI native.
20:48My father is over here,
20:50he's 72 years old,
20:51but he's using AI every day.
20:53So everybody is using AI,
20:55okay,
20:55so there is not a problem
20:56of trust.
20:57There is not a problem
20:58of trust,
20:59but what they want is,
21:00they want precision.
21:02They want precision
21:03and they want fast information.
21:0551% of consumers,
21:06they want fast information
21:07and more than 60%
21:09want precise information,
21:11you know.
21:11So what we have to really
21:12be careful about
21:13is hallucinations
21:14and algorithms,
21:16okay,
21:17and that's where
21:17we have
21:20very structured methodology
21:21where we do not
21:22deploy anything
21:24if it is not
21:25rigorously tested
21:27either manually
21:28or automatically
21:30or semi-automatically,
21:31you know.
21:32So you have to control
21:33the hallucinations,
21:34you need to ensure
21:35that the AI,
21:35you always keep the human
21:36in the loop
21:37to make sure
21:37the final decision
21:38is being taken
21:39by the human
21:39and you're testing constantly
21:41because if you start
21:42giving imprecise information,
21:44then you lose the trust.
21:45This is one of the reasons
21:46we have decided
21:47not to create
21:49an application
21:49on chat GPD
21:50even though the application
21:52is ready.
21:53It's in our sandbox.
21:54We have done it
21:55but we are not going
21:56to release it
21:57because when we tested
21:58some of the applications
21:59that is existing,
22:01frankly,
22:01at times,
22:02the answer is embarrassing
22:02and you can't control
22:04the answers
22:05that the GPD is doing.
22:08So control
22:08is prime most
22:10and the only way
22:11to keep the trust
22:12of your consumer
22:13who is already AI native
22:14is to ensure
22:15that you have
22:16that control
22:16on the output
22:17of the AI.
22:19I know a lot
22:19of companies
22:20partner with LLMs.
22:21Is Club Med
22:22partnering with some
22:23of these
22:23or are they all
22:24built in in-house?
22:27So we have
22:29the best of breed approach
22:30and we have
22:32a strong belief
22:32that the core
22:35artificial intelligence
22:38and the infrastructure
22:39should be part
22:40of the company
22:41infrastructure
22:42and core assets
22:43because this will be
22:44part of the operating
22:45model and coming
22:46back to what Andrew
22:47said,
22:48how can AI drive
22:49the operating model
22:50of the company?
22:51So this has to be
22:52internal.
22:53But on different
22:54targeted topics,
22:55let's say legal,
22:56where we do not
22:58have the data
22:58to train that AI,
23:00there might be
23:00SaaS providers
23:01who already have
23:02that data
23:04that they have
23:05been building
23:05over the last
23:0610-15 years
23:08and they already
23:09have trained
23:10an agent
23:10which can do
23:11the task
23:11and you can
23:12win in time
23:13to market.
23:13So all of these
23:15niche use cases
23:16where data
23:17is extremely important,
23:18we work with
23:18partners.
23:19But of course,
23:19we have many
23:21partners.
23:21AI is a team
23:22sport which
23:24requires your
23:25employees,
23:26it requires
23:27your management,
23:28it requires
23:28your shareholders
23:29and it also
23:30requires your
23:31partners.
23:32So we work
23:32with Microsoft,
23:33we work with
23:34all cloud providers,
23:35Google,
23:35Alibaba in China,
23:36and we work
23:38with startups,
23:39we work with
23:40high quality
23:42consulting firms
23:43who provide us
23:44talents which
23:46work hand in hand
23:46with our internal
23:47teams,
23:48squads to build
23:49AI products.
23:51So this afternoon's
23:52theme is
23:53sovereignty and
23:54ethics.
23:54What should
23:55remain uniquely
23:57human and
23:58how do you make
23:59sure it does
23:59in practice?
24:00There must be
24:01areas where you
24:01say, okay,
24:02let's not even
24:04bring AI into
24:05that or
24:06what are your
24:07boundaries?
24:15I can stay
24:16like this for
24:16the next one
24:16minute.
24:18It's a tough
24:18question.
24:21You know,
24:22there is no
24:22boundary as
24:23such.
24:23There is no
24:24boundary as
24:24such.
24:25I have given
24:25the blueprint,
24:27you know,
24:27accuracy,
24:29precision,
24:30you need to be
24:30able to have
24:31control.
24:32What you can't
24:33leave is,
24:36there is no
24:37boundary where
24:38AI, even
24:40not today,
24:41in three
24:42months, six
24:42months, one
24:43year, five
24:43years, ten
24:44years that AI
24:44would not be
24:45able to do.
24:46We can't even
24:46think of it.
24:48I'm working
24:48in AI for 15
24:49years.
24:50I could never
24:50think when I
24:51started doing my
24:53deep learning
24:53lessons that
24:54there would be
24:55large language
24:56models who can
24:56answer any
24:57question around
24:57the sun.
24:58I could not
24:59imagine, so
24:59what will
25:00happen in
25:0010 years, I
25:00can't imagine.
25:02But where we
25:04as leaders of
25:06enterprises need
25:07to be careful as
25:08to how much
25:09autonomy we want
25:10to give to
25:11these agents.
25:12And always
25:13ensure that the
25:13human is in
25:14the loop.
25:14So the
25:15boundary will
25:15be set by
25:16the level of
25:17autonomy that
25:18we provide to
25:20the AI agents
25:22in the new
25:22operating model.
25:23We can make a
25:24choice for this
25:26invisible back-end
25:27processes where
25:28the stakes are
25:29not so high.
25:30You can have
25:30some margin of
25:31errors.
25:32You can give
25:33full autonomy to
25:33that agent to
25:34execute the
25:34task end-to-end.
25:36But there might
25:37be other areas.
25:40If you compare
25:40sectors in the
25:42medical field, we
25:43cannot give full
25:45autonomy.
25:45You always need
25:46a radiologist,
25:48even though the
25:48AI will help him
25:49to make the
25:50diagnosis, but the
25:51final stamp would
25:51be the human.
25:53So the cursor
25:53would depend on
25:54that, and that's
25:54where you have to
25:55be very careful
25:56as to how much
25:56autonomy we give
25:57to these tools.
25:59All right.
26:00We have a couple
26:01minutes left.
26:02So the final
26:04question I have
26:04for you is,
26:05what's your
26:05advice to other
26:07enterprises that
26:08are building
26:09combination human
26:10and agent
26:11organizations?
26:13And then we
26:14want to know
26:14what's next for
26:15Club Med.
26:23So, you know,
26:24I think the key
26:25messages we
26:26shared is, you
26:29know, it's
26:29already, I
26:29think it's a
26:30good base.
26:31But there's one
26:32thing that I
26:33would like to
26:33emphasize is
26:34that companies
26:37must not consider
26:39AI as a
26:40technological topic.
26:42Okay?
26:42The moment we
26:43think of AI as
26:44tech is finished.
26:46We can't go
26:47forward.
26:47We have to
26:48think of AI as
26:49a human topic.
26:51Okay?
26:52And when I say
26:52human topic, it's
26:54because it is
26:56touching the way
26:56we are going to
26:57live, behave,
26:59work, and it
27:01touches every
27:03employee in the
27:03company.
27:04It touches the
27:05operating model.
27:06Right?
27:06So, not to
27:09forget that the
27:10consequences of
27:12anything you will
27:12do with AI will
27:13have, will be
27:14pretty large, and
27:15many, many people
27:16will ask questions,
27:18will be involved.
27:18Okay?
27:20So, that's the
27:20first thing.
27:21Second is, because
27:23it's a human topic,
27:24you need to
27:25convince, empower,
27:28and embark your
27:30colleagues and the
27:32humans in this
27:33adventure.
27:35The way to do is
27:37not only through
27:38communication, because
27:40just communication is
27:42futile.
27:43You need to do
27:44communication via
27:44demonstration.
27:46So, you need to
27:47show real value.
27:48And I would
27:51advise the same
27:52approach that I have
27:53been following for
27:54Club Med, but also
27:55previously, is don't
27:57hesitate to start
27:57small.
27:59Big banks don't
28:00work.
28:00This is a technology
28:01which is complex.
28:02It can hallucinate.
28:04Okay?
28:04So, take small pain
28:05points, try and do as
28:06much bottom-up as
28:07possible, and then go
28:09end-to-end in order to
28:10redesign that process
28:11that you're trying to
28:13impact, and show the
28:15value as fast as
28:16possible.
28:17In my team, we have
28:17a rule.
28:18We don't do any
28:19project which takes
28:20more than three
28:20months.
28:22Within three months,
28:23from ideation to
28:24production with real
28:26value, maximum four
28:27to five months.
28:29For that, if the
28:30project needs to be
28:30cut into multiple
28:31pieces, that's all
28:32right.
28:32Okay?
28:33But we need to see
28:33something, because you
28:34can't be in this loop
28:35of construction,
28:37construction, research,
28:38and development,
28:38innovation, and
28:39proof of concepts.
28:40And you can't do
28:41all of that, because
28:42there's a learning
28:42path, there's a
28:43learning curve.
28:44You need to
28:44re-skill, re-train,
28:46re-train.
28:47So, start small,
28:49demonstrate, and once
28:49you've demonstrated,
28:51communicate.
28:51Go and explain to
28:52every employee as
28:53many people as
28:54possible, smartly,
28:55under low budget,
28:57so that it's exciting,
28:58and then people
28:59themselves will come
29:00and say, okay, we
29:01want to do more of it,
29:01more of it, and we
29:02want to pass to scale.
29:05All right, Siddharth.
29:06Thank you so much for
29:07your insight and
29:09knowledge.
29:09This was very
29:11enlightening, and we
29:12appreciate you, and
29:13we'll see you around
29:14this week here at
29:14Viva Tech.
29:15Thank you so much.
29:16Thank you so much.
29:17Thank you so much.