Now the Hard Part Making Entreprise AI Work

Vivatech

Transcript

00:00Nous avons le plaisir d'avoir avec nous aujourd'hui

00:03Biljana Kajitović

00:05C'est très bien

00:07C'est plus facile que moi

00:09Vous avez deux décadas d'expérience

00:12Vous avez bridged

00:14le gap entre technologie et business

00:16pour des années

00:17et vous avez démontré

00:19une forte passion pour la transformation

00:21et l'innovatif business model

00:24Vous avez rejoint Engie

00:25en mai 2022

00:27en charge d'executive vice-president

00:29en charge des digitales et informations systèmes

00:31Vous avez la bonne femme

00:33et la bonne place aujourd'hui

00:34Florian Duetot

00:36Vous avez le co-founder

00:38et CEO

00:40de DataEQ

00:41que vous connaissez tous ici

00:42C'est une société d'AI et d'affaires

00:44qui aident à faire

00:46la analyse et la machine learning

00:49plus accessible

00:50Vous avez aussi la bonne femme

00:52et la bonne place aujourd'hui

00:54Jepersot

00:54Vous êtes l'EI Global Emerging Tech

00:58Ecosystem Leader

00:59Vous travaillez avec

01:01l'Emerging AI

01:02et languages

01:03et data

01:04platform technology

01:05companies

01:06et investors

01:07et nous allons commencer

01:08avec vous aujourd'hui

01:10Nous allons essayer

01:11de être très concrets

01:13avec concrets

01:14de implémenter

01:16l'AI

01:17dans les entreprises

01:19nous pouvons

01:20nous parler

01:21avec vous

01:22sur l'EI

01:23Gen AI

01:24parce que vous avez

01:25votre propre

01:26generative AI

01:27adoption

01:28use cases

01:29Merci

01:31Merci

01:32et c'est

01:34c'est ce qu'on voit

01:34à tous

01:35à tous

01:36Et c'est

01:36à tous

01:36la semaine

01:36du week

01:37donc je pense

01:37que les gens

01:38sont tentés

01:39de aller

01:39pour le

01:41$1

01:42à tous

01:42avons

01:43comme tout

01:44le endpoint

01:45nous

01:45nous avons

01:46été

01:46trois

01:46dernières

01:47dernières

01:47ou so

01:48J'ai

01:48dit

01:49il y a

01:50une

01:50d'une

01:51d'une

01:52d'une

01:53d'une

01:54d'une

01:54que nous avons pu faire en sorte que nous avons pu faire en sorte que nous avons construit un nombre

01:58de tools.

01:59Ce sont tous les gens qui sont tous bien. Nous avons les gens qui sont en train de se familiarisent.

02:02Nous avons 400 000 personnes. Nous avons construit ce qu'on appelle l'EYQ.

02:08Ce qu'on a fait, c'est un fabric et un collection de modèles et des modèles et des tools

02:13que nous

02:14nous voulons que les gens se comprennent et qu'ils soient confortables.

02:17Nous utilisons pour faire des concepts de la recherche avec nos clients et d'autres utilisateurs.

02:23Donc, ça a été très bien. Je dirais que nous sommes 10% de l'année là.

02:27Je veux dire, c'est juste que nous avons commencé.

02:29Et maintenant, nous avons travaillé très harde à vraiment être très focussés

02:34sur les opportunités et les casés que vous pouvez pouvoir obtenir des résultats rapidement.

02:38Vous savez, avec le ROI.

02:40Donc, je pense que c'est le stage où nous sommes maintenant pour nous et pour nos clients.

02:45Donc, il y a beaucoup plus à venir, mais c'est très intéressant

02:49quand je pense que ce que nous allons nous embarquer dans les deux ou trois ans.

02:53Et peut-on avoir quelques exemples de casés en healthcare,

02:57banking, sales ou product design, par exemple ?

03:00Absolument.

03:01Je veux dire, il y a beaucoup.

03:03Je mentionne 1,000 proofs-of-concepts que nous avons réalisé.

03:06Donc, il y a une qui meurt très rapidement,

03:08si vous pensez à quelque chose que nous sommes tous familiarisent,

03:12c'est le service, c'est le service,

03:14qui s'applique à presque chaque industrie que a customers.

03:17Donc, quand vous appelez,

03:18c'est pas un très bon expérience,

03:20c'est-à-dire qu'il y a beaucoup d'exploitation.

03:23C'est-à-dire qu'il y a beaucoup d'exploitation.

03:25C'est-à-dire qu'il y a beaucoup d'exploitation à l'exploitation.

03:36C'est-à-dire qu'il y a beaucoup d'exploitation à l'exploitation.

03:38Tout le temps avant, c'est-à-dire qu'ils ont appris directement avec l'Ai,

03:41ce qui vous permet de avoir des conversations à l'exploitation.

03:47La clientèle de l'exploitation et les centres,

03:50le clientèle de l'exploitation,

03:51et, finalement, la company a beaucoup d'efficacité.

03:55Je peux aller dans la liste, mais nous allons avoir de temps.

03:58Donc, je vais passer là.

04:00OK.

04:01Et nous pouvons revenir à l'exploitation

04:03de l'exploitation.

04:05Bien sûr.

04:05Nous allons avoir d'autres exemples avec vous, Biljana.

04:09Nous pouvons parler de l'utilisation d'Ai.

04:12Je pense que vous avez différents types de business,

04:15et vous avez des parts de votre business qui sont très critiques,

04:20où la sécurité est très importante.

04:23Et, comme vous l'avez dit,

04:25l'Ai n'est pas 100% accurate maintenant,

04:29et peut-être difficile d'améliorer l'Ai dans ces activités.

04:33Mais je suis sûr que c'est différent.

04:35Vous avez parlé de l'Ai.

04:36Vous avez parlé de l'Ai.

04:37Jai, sur le centre de l'exploitation,

04:39sur les services de l'exploitation.

04:41Donc, comment est-ce que l'Ai vous place dans les différents types de business ?

04:47Je suis très heureux de partager.

04:49Donc, c'est très heureux d'être ici,

04:51et merci d'avoir accès.

04:53Je pense que nous sommes en train de voir un vraiment cool temps

04:56dans l'histoire,

04:57dans l'histoire de l'humanité.

04:58Donc, l'Ai, l'Ai, l'Ai est explosée.

05:02Il devient très puissant chaque jour.

05:04Et je pense que ce qu'on voit cette année,

05:05c'est devenu naturel.

05:07Donc, vous parlez de l'Ai,

05:08on parlez de l'Ai,

05:08et on parlez de l'Ai en très naturel.

05:11Dans l'Angie, nous avons une stratégie très claire

05:14à levererage l'Ai,

05:16mais avec les autres types de technologies,

05:19comme l'Ai, l'Ai, l'Ai,

05:21l'Ai, l'Ai,

05:22l'Ai, l'Ai,

05:22l'Ai, l'Ai, l'Ai,

05:23l'Ai, l'Ai, l'Ai,

05:28parce que nous pouvions l'interaction entre l'human et l'amn

05:31à un autre niveau.

05:32Et avec cela, nous avons les insights que nous ne pouvons pas avoir,

05:35ou save costs,

05:37ou increase the satisfaction of our customers, etc.

05:41So, we do quite a number of use cases,

05:43like those that many industries do.

05:45So, call centers, of course,

05:46we equip our employees with co-pilots,

05:49we want to boost their productivity,

05:52we equip all of our programmers

05:54with GitHub co-pilot,

05:55because I think also what many industries,

05:58many companies that have programmers,

06:00do now, nowadays,

06:02is test what is the productivity gain that you can get

06:04if you equip all of your software developers with co-pilots,

06:09this is exactly what we are testing.

06:10But, as we called this panel,

06:13now the hard part.

06:14The hard part is,

06:15how do you truly get value from scaling up these solutions

06:18that are still quite young solutions?

06:21So, this is the area that we are working on,

06:24just like everybody else.

06:25But, you will notice that all of these examples

06:27that I mentioned are not in the industrial sectors.

06:30So, we are an energy company,

06:31we are very highly regulated industry,

06:35for a good reason.

06:36So, the work that happens in our industrial environment

06:38needs to follow very strict guidelines,

06:41very strict procedures.

06:42So, now, imagine you bring a technology

06:44that is very novel,

06:46that sometimes hallucinates,

06:47produces results that are not very accurate always,

06:51or that are not very reliable,

06:52and you have, for the same input,

06:55you can get different output,

06:56and you put that in a high-risk environment

06:58where you need to ensure

07:00that you always have risk-free approach

07:03to ensure that your processes

07:05and that your people are safe.

07:07So, if you marry the two,

07:08you see that the technology is not yet ready

07:10to be scaled in an industrial environment,

07:13but we have very interesting proofs of value.

07:16Okay.

07:16So, as an example, in our technicians,

07:19so field technicians, for instance,

07:20when they go to field to, I don't know,

07:23do a maintenance on a boiler,

07:25they will go currently,

07:27they go in the field,

07:27and they check what they need to do,

07:29and then they go to a SharePoint site,

07:30and they see hundreds of documents.

07:32They need to find the right document,

07:33open it, find the right information.

07:35It takes them, on average,

07:37maybe 10 minutes to find the information that they need.

07:39Now, if you apply Gen.AI on that,

07:41they can actually get to that information within seconds,

07:43and they get referenced

07:45to what is the document that the information came from,

07:48where exactly they can find it,

07:49and what would be an interesting video to take a look.

07:52So, they can get to the right information within seconds,

07:55but we're still not deploying it at scale,

07:57because we cannot, at the moment,

08:00testify with 100% accuracy

08:02that all the steps are going to be already verified,

08:05and that there will not be one step that might be safety-critical

08:09that actually they might miss if they just follow Gen.AI.

08:11It's very interesting that you talk about accuracy,

08:14and both of you talk about the return on investment,

08:17and McKinsey just released a study

08:19where they said that only 10% of the company

08:22got a 20% improving of growth margin with AI,

08:26which means that you are all very careful about it,

08:30because we have to test to be the accuracy,

08:33and to test also the profitability

08:34and return on investment of those technologies.

08:37I would just say for now,

08:39because I'm convinced that first technology is maturing

08:42at a rate that we've never seen before,

08:43so it's coming, it's going to be more accurate,

08:45and we will have very different,

08:47of course, we will have different ways to ensure

08:49I want more accuracy versus I want more creativity.

08:51You can already do that now very closely

08:53when you work closely on LLM.

08:55So that is to come,

08:57and as well, we need to adapt some of our procedures.

08:59Normally, you get the new technology

09:01and you apply it to the workflows that you have,

09:03and then you change the workflows.

09:05Well, now we're learning,

09:06we need to change the workflows

09:07together with technology that gets applied.

09:10Florian, you are very often in between,

09:12you know, it's true,

09:15and you're working on technology implementing,

09:17but let's talk about people,

09:19because of course it's a revolution in process,

09:23in industrial process, in services process,

09:25but it's also a revolution for people.

09:27Everyone is very concerned,

09:29and I'm sure when you implement,

09:31when you help companies to have their own

09:34AI language, et cetera,

09:38probably one of the first things you have to do

09:40is to train people and to help people

09:42to be able to use those technologies.

09:46Yeah, and why is training important

09:49is because probably after three days

09:53hearing about AI every 20 seconds,

09:56three days in a row,

09:57you might still not know

09:59what is good AI versus bad AI,

10:01especially in the future.

10:02And I don't even mean in the ethical sense,

10:04I'm just meaning in terms of

10:06what are the use cases that would be relevant,

10:08which use cases will be working versus not.

10:10We actually don't know.

10:12We are all discovering it.

10:13And I do firmly believe that

10:15from the perspective of enterprise,

10:17it will be a lot about people

10:20with domain expertise,

10:21knowing a lot about their specific data,

10:24their specific processes,

10:26the safety concerns,

10:28all of the background knowledge.

10:29It will be about those people

10:30being able to step up

10:32and actually build more AI capabilities

10:34by themselves,

10:35or at the least control them

10:36and understand them.

10:38And so that's why training is important

10:39because this transformation for enterprise

10:42is like so huge

10:43because you actually just not,

10:46it's beyond building a proof of concept

10:48of a chatbot.

10:49It's actually about changing business processes.

10:52You need to check everything before and after.

10:54You need to understand

10:55what is your specific data,

10:56why it matters,

10:57if the quality is correct or not.

10:59What are the expectations?

11:01What are the business impacts?

11:02And that does require people to step up

11:04and be trained to understand

11:06and how they can use those technologies.

11:08Okay.

11:10And people,

11:11how do you bring them into the process?

11:14How do you try to make them that fear,

11:17the transformation of their work,

11:19of their mission,

11:21of their raison d'être also?

11:24I may be naive,

11:25but I do believe quite a bit in co-development.

11:27As in, I do believe that you need people

11:30from the domain,

11:31whatever the domain.

11:32Would that be finance?

11:33Would that be energy trading?

11:35Would that be safety?

11:37Would that be manufacturing or any type of R&D?

11:40You need people coming from the domain

11:41being able to understand

11:42all the new AI stuff

11:45is actually being implemented.

11:47And for that,

11:48they actually need to participate a bit at least

11:51to some portion of the solution itself.

11:54And in fact,

11:55in the new world,

11:56it is possible to have solution easy enough to use

12:00so that people from the business

12:02that are data savvy,

12:04that are quantitatively smart,

12:06can actually get there,

12:08and actually build AI solutions on their own.

12:10And that's a big, I think,

12:11shift compared to other technological changes,

12:15compared to the internet back 20 years ago,

12:18which was probably all for the geeks

12:20being able to learn how to do PHP as fast as possible.

12:23Today, I think it's actually a revolution

12:25where everyone needs actually to change

12:27and to participate in.

12:29Okay, we go back to you, Jay Persaud,

12:31to see precisely how I talk about your own AI,

12:37generative AI you developed.

12:39So how EY invested in this particular own AI platform

12:44and how it changed the way company work,

12:47or maybe you,

12:48how it changed the way you work in EY.

12:52and maybe also,

12:54let's start with this one.

12:56I have another question.

12:57Yeah, it's a great question.

12:59And so maybe just to contextualize it a little bit.

13:03As I mentioned before,

13:05this first phase,

13:06where I said we were 10% of the way there,

13:08I think there was a big hype machine.

13:11We all remember this, right?

13:13We all had to get on board.

13:14If you didn't get on board,

13:15you'd be left behind.

13:15I think those things are true.

13:17But the reality is within an organization like ours,

13:21or any of yours,

13:22large companies,

13:25every employee and every stakeholder

13:28has to get up to a common level.

13:31What is this?

13:32What does it do?

13:33What does it mean for me?

13:34What does it mean for my company?

13:36What does it mean for society at large?

13:38And I think that's still a job to be done.

13:41What we did with our tool,

13:43as you mentioned,

13:44was that we needed something to get people familiar with it.

13:48Okay.

13:48But we're not there yet,

13:50because we now have to go work with our clients

13:53to make sure that we have that demystification process,

13:56if you like,

13:57that you were mentioning, Florian.

13:58What does it do?

13:59What does it not do?

14:01What's really realistic?

14:02And I think because of all the investment

14:04that's been made the last couple of years,

14:07leadership teams, boards,

14:09are now saying,

14:10we're going to need some payback, right?

14:13And there's a lot of focus on these business cases

14:17and a lot of focus on finding very specific examples

14:20that can be used and solved.

14:22And I think you're going to see that

14:24over the next several weeks.

14:25When I look at EY,

14:28the use of our EYQ that you mentioned,

14:31I mean, there are some really interesting numbers.

14:33you know, our people in technology,

14:37their effort is 25% down when it comes to coding.

14:43Across the board,

14:44we think we're saving hundreds of millions of dollars right now.

14:48We're not too worried about putting that in the bank.

14:50We want people to have easier jobs

14:54and to become very familiar with these tools

14:56and how they can and shouldn't use it.

14:58So that's kind of where we are.

15:00And that's what we're telling our clients.

15:01You know, you've done this first phase,

15:03but let's work together now to really make it real,

15:07to get the kinds of improvements

15:09that we've been talking about the last couple of years.

15:12How do you work with the AI startup?

15:14Because I know you like to cultivate partnership

15:17between the AI startup and bigger companies

15:21to go to market strategies?

15:23Yes, we do.

15:24So my role within the firm,

15:27as you mentioned,

15:29is responsible for our global emerging technology ecosystem.

15:33And today that is almost all AI.

15:35And so while the big players

15:38have sort of cornered the investment side of it,

15:42you know, many of them are here at the conference,

15:44there are lots of opportunities for smaller companies

15:47to solve specific problems.

15:49There are lots of opportunities for other companies

15:51to solve the things that you're going to need

15:53to get the data right,

15:55so you can get the full value out of AI.

15:57So what we're doing is very focused

15:59on a portfolio of those relationships

16:01that we can bring into our organization

16:03as a part of our service offer.

16:05So they're a part of our fabric that I mentioned,

16:08and they're also a part of our client service teams

16:10to the extent that we can find commonality

16:13where we can go to market together.

16:14It's a big win.

16:15I think there's so much to do here

16:18that we're going to need the big and the small

16:21to work effectively together

16:22to get us to where we think we'd like to be.

16:25Bien, are you in charge of transformation,

16:27digital transformation?

16:28How do you implement these cultural changes in the company?

16:35and how you helped to create this macro vision for AI in the workplace

16:41and drive adoption across the company?

16:45So I think we've all been driving transformations for a long time.

16:49We called it digital transformation.

16:51Now it's called something with the AI flavor.

16:53It's the flavor of the day.

16:55But the fact is, we are living in a digital world,

16:57whether we like it or not.

16:58If you look at the top five companies by market cap,

17:01they're all digital companies.

17:03And I think many of us in the industries that are not digitally native

17:08or in the companies in industries that have started sometimes 100 years ago

17:12have this challenge of bringing along our business

17:16to open their eyes that the world has changed.

17:19So as part of that, I find one of the most critical things

17:23is to make sure that the top leaders,

17:24regardless of which business line or which function they belong to,

17:29that they have a very strong support and strong belief

17:32that unless a specific company becomes more digital,

17:35more AI-driven, more data-driven, more AI-enabled,

17:38it's going to be very difficult to compete in today's times.

17:42So I'm lucky that in my company, in Engie, that is the case.

17:45So I have a very strong support of our CEO

17:48and the whole executive committee.

17:49I fully agree with you, Jay, when you said it's now about

17:54how do you actually make impact.

17:56We've been doing a lot, we've been piloting a lot,

17:58we have a lot of proofs of concepts, but it's about scaling.

18:01So how do you ensure that you prioritize the right use cases

18:04and then that you can really scale them and get value at scale?

18:08Because we've invested, we want to see that value coming back

18:10and we want to make sure that actually we do change the processes.

18:13And that would be very difficult if it's not for the top leaders' engagement.

18:18and I think it's a critical part.

18:20And what I think is really a great thing about Gen AI,

18:24it's so approachable.

18:25So I've never seen in my career business being so interested in a technology.

18:30Everyone wants to test it.

18:31Everybody wants to do something, they want to have a use case,

18:34they want to test it.

18:35And that's great because business is becoming role model for their staff.

18:39So if staff in a specific business line see that their leader

18:42is very excited about being augmented by Gen AI,

18:45they will want to have a bit of that.

18:47If they see that one of the big bets in their businesses

18:50is related to something to AI, they will want to be part of that.

18:53It doesn't come anymore as just a push from a digital function.

18:56It's actually something that becomes part of the business itself.

19:00And I think that's super critical.

19:01But I think at the same time that coming from digital and IT function,

19:04we do need to continue to educate the people who are not as close to technology

19:10with regards to what technology can do and what it can't do.

19:13Or even better, they don't even need to understand that.

19:16They need to know that we need some other technologies as well

19:19to solve some of the bigger problems.

19:21But what Gen AI brings us is that interface and that ease of use

19:25that really gets people extremely interested in just opening the box.

19:29So that is very nice and then it helps with adoption.

19:32Okay.

19:33But it requires a lot more training than before

19:36and a different type of training because it's easy to use.

19:39But if you don't learn to use it well, the results you are going to get out

19:43are not going to be the optimal results.

19:45So you still do need to actually go through training to learn how to prompt it well

19:49or you need to go through hackathons to really figure out

19:51what kind of use cases can I really unlock only using Gen AI

19:54or I also need machine learning or I need analytics.

19:58And that is the part, because it seems so easy, actually it's a hard part

20:02to get people to understand that they do need to still invest time

20:05to truly get the benefits out of it.

20:07We talk about scaling, we talk about accuracy, we talk about education.

20:13But let's talk about time and speed because maybe there is an urgency.

20:19we see companies adopting very quickly and transforming themselves very quickly

20:25and others who are more hesitating.

20:28So maybe Florio, you can tell us, is it so urgent to adopt the AI process

20:33and is there a risk that we are not doing it?

20:39Yeah, for sure.

20:41I think that we have to acknowledge two things.

20:44the first things to continue on Biljana's comment is that

20:48before there was a big chasm between digital first companies and non-digital companies.

20:53Because digital company, you've got your cookies, your logs,

20:56digital interaction with customers, easy to integrate the data relatively

21:00and you can actually have a quick action system back to your customers or whatever.

21:07Non-digital companies, you've got pilots of docs, real stuff in real life,

21:12you don't know where it is, the only way to know what's happening is to take a picture and so

21:17forth.

21:18You can't actually put it into a computer system.

21:21Generative AI, at the end of the day, bridge that gap between the digital and the non-digital world.

21:26Because all of those pilots of documents, all of your assets,

21:29you can analyze pictures, you can analyze real life stuff.

21:32And so everything non-digital can actually become digital very, very quickly now.

21:36Because literally, very soon, you could analyze in real time every video in every of your physical assets.

21:43That is becoming realistic.

21:45So non-digital is becoming digital, creating an urgency actually for non-digital companies.

21:50The other aspect from my perspective for all companies is that Gen.AI is indeed not a sure game.

21:59Some of those projects, some of those use cases will work, some of them won't work.

22:04But still, changing organization and changing culture takes years.

22:09And the cost of not starting now is the cost of having your company not understanding in the future

22:17the changes needing to be done.

22:19Because you kind of need to test it now to anticipate the change in your work,

22:24the change in your organization.

22:25And it will take five to ten years for most of us to actually get there.

22:31Maybe you work with all kinds of companies everywhere in the world.

22:35Do you see differences between the US, Europe and Asia in the scale of adopting generative AI?

22:45Yeah, there are differences.

22:47I think this conference we're having, the fact that we're talking about AI so much,

22:52is you can't avoid it.

22:54So there's that commonality that is all over the world, right?

22:58At a pace I think that's not been seen before, in terms of the people wanting to be a part

23:05of it.

23:06So that first question that you just raised, it's not whether you should or shouldn't use it,

23:11you have to use it because it is the future.

23:13The question then becomes, how do I use it effectively in my particular situation?

23:20And I think there are differences between North America, Europe and Asia,

23:24that will determine the pace at which they're used.

23:27You can see it right now, right?

23:29The US traditionally has been a country where people just jump right in.

23:33And let's see what happens, then we'll worry about what happens afterwards.

23:36Europe maybe is a bit more cautious, try to plan a bit more, so that is definitely there.

23:40But the one thing I wanted to highlight is,

23:43it's really important in your organization to think about this.

23:47It's a powerful technology, it's a technology that democratizes a lot of things,

23:51puts it in the hands of people.

23:53But you have to think about your processes and workflows.

23:57Because what's going to happen is that it will change all of them.

24:01It may eliminate many of them because it can go from beginning to end.

24:05And there are lots of people that work in there, right?

24:07So this process that we have to go through of managing that change within organizations,

24:12I think we're only now coming to grips with that.

24:15And that becomes important for companies, it becomes important for societies.

24:20Because if you think about whether we actually realize the efficiencies that people are talking about with Gen AI,

24:29that's a lot of people that are going to have to do something else.

24:32And I know this has been talked about a lot in this seminar or this conference,

24:36but I think that's an important thing for us to focus on together with all of the stakeholders.

24:41There are going to be different things to do, I have no doubt about it, right?

24:45But we've got a plan for that.

24:46And that's kind of what I wanted to close with here today from my perspective, Christophe,

24:50is that the people aspect of this as we go forward I think is important,

24:55but it's going to become even more important.

24:57Okay, we are almost running out of time.

24:59Maybe one more you want to add to what we just said?

25:02Yeah, I think maybe what we didn't touch upon that much is around data readiness.

25:07So just maybe advice to people in the audience who are working in small, big, whatever companies,

25:13use Gen AI as a bit of a trojan horse to get into the business and to help them to

25:18clean their data faster

25:20and to raise the quality of the data.

25:22Because whenever you are ready for the game, you will use the data that is now has been prepared for

25:27it.

25:27So data readiness is super important.

25:29We have, for instance, developed a tool to help our business to just assess where they are,

25:34depending on the use case they have, so we can see whether we actually spend time on that POC

25:39or we don't because the data is ready or not.

25:41Because we've seen through what we've been doing in the last year that a lot of course will depend on

25:46data

25:46and when people say at the beginning our data is great, well actually it might not be.

25:50So do spend your time to understand the data and then to clean it before you dive into some of

25:56the big POCs

25:57because it's always going to come back and use Gen AI as a nice way to do that.

26:02Florian, one more recommendation for those who still are hesitating.

26:06As you said, clean your data because ChatGPT won't do it for you.

26:09Exactly.

26:11Okay, thank you, all of you. Thank you very much.

26:14Thank you.

26:16Thank you.

Catégorie

Transcription

Commentaires

Recommandations