00:00Hello everybody, great to have you.
00:02You might wonder what an insurance company has to transform.
00:07And I mean, you might have seen the title, right?
00:11Transform Tomorrow Today.
00:14So this sounds like an urgency.
00:16And insurance is not known to be a very fast-paced industry
00:19because it typically doesn't evolve so quickly.
00:23But I can tell you why Gen.E.I. for us is really a means
00:28and also an opportunity, but if we badly manage it, a threat not to speed up.
00:34And this is, first of all, we have seen a democratization
00:37through generative AI for all AI use cases
00:41in a speed that we have never had before.
00:43So definitely a reason to act.
00:45Secondly, and equally important,
00:50insurance is just essentially data and people.
00:53There is much more. We don't have a physical product, right?
00:56I mean, out there there is Peugeot and Renault.
00:57You can touch their product, an insurance product you cannot touch.
01:01We sell protection. We sell certainty.
01:04And we do it with using data.
01:06We have done that all the time.
01:08And that is why for us, generative AI or dealing with data is not really new.
01:14We have started our AI journey for a very long time,
01:17already since eight or ten years.
01:19And that is why AI is really at the core of our strategy.
01:22And this is why we have worked already for the last couple of years
01:26to build up roughly 400 use cases of AI in all our entities across the globe.
01:32And they are already contributing a lot to our value,
01:35which I believe makes us probably to one of the more innovative insurance companies
01:39and that we want to stay at the forefront of this industry
01:42requires that we embrace generative AI fast, quickly, robustly, and responsibly.
01:51And this is now asking you the question, how are we going to do this?
01:56I mean, what especially is it that we as a large insurance company are doing?
02:01So it's basically three things, technically speaking.
02:06So to speak, first, if AI is everywhere in Gen AI and it can easily be used,
02:13we need to make it accessible for all our people.
02:16So we have built already, almost a year ago, a secure version of JetGPT
02:22that we rolled out to 150,000 people within AXA
02:25so that everybody can start learning it, that everybody gets wet,
02:29that everybody can start doing prompting
02:31so that it doesn't come as a surprise.
02:34And the funny thing is, at the day when we launched it last year,
02:38it was, I think, the 28th of July,
02:40another large financial institution just said that it shuts down
02:44the access to JetGPT for all their employees.
02:48And I thought, well done.
02:51So two days later, my CEO publicly said,
02:56AI is the center, at the center of our strategy.
02:59And then I began to sweat.
03:02Because if your CEO is saying that,
03:04you can be sure it's going to happen and it's going to happen big time.
03:07So, now, secure GPT for everybody, I think, was the first step for us,
03:12you know, to have everybody get on the journey.
03:15The second topic is, as many probably have thought,
03:19I mean, now we're building these large language models
03:22and you can buy one and train it with your data
03:24for whatever, 50 million or what have you,
03:27huge amounts of money.
03:29Thank God we didn't do it.
03:31What we did instead is we started to use co-pilots
03:35wherever this is feasible and possible.
03:37Now, we all know co-pilots is still not at the end of the journey
03:41and we have still some developments to take.
03:43However, this is really a very promising field
03:46for just a user like us
03:48if we are able to implement it reasonably in our business processes
03:52and we fully understand the value that this is getting.
03:55And this is, again, a lot about piloting,
03:58a lot about trying, a lot about using
04:01and a lot about being mindful of where it makes sense
04:04and where it doesn't make sense.
04:05So, that's the second big step.
04:07And then there is a third big step
04:09and that is around
04:12how can we bring generative AI
04:14into the core processes of our business?
04:18And what are the core processes of our business?
04:20I mean, you might have experienced that as a customer
04:22but it starts basically with, okay, you buy an insurance.
04:26You don't do that very often in your life
04:29but at some specific points when your life changes,
04:33you buy an insurance
04:33and you want to get the best advice
04:35and you want to get the best product that fits best to you.
04:38And this is, of course,
04:40where we can leverage all the data from the past
04:43and support our sales agents
04:45and support our service people
04:48in giving the best advice to you as customers.
04:51So, that's what we call hyper-personalization.
04:54That's one area where we instill our own data
04:57and we can talk about the technology,
04:59how we're going to do it as well.
05:01But, I mean, I'll leave that aside for the moment.
05:02But that is where we bring in
05:04our own access-specific information
05:07to be better for our clients
05:10and also to be more effective
05:12and more efficient for ourselves.
05:13Because sales time is very precious time.
05:17These people know how to deal with customers
05:22and to overburden them with admin stuff is a bad idea.
05:26Second, and this is the second big process
05:29and I hope you never had that too many of them,
05:31is a claim.
05:32If you have a claim, you go to your insurance
05:34and you want to get paid.
05:37So, as it happens,
05:38we have spent already quite a lot of money in the past
05:41to optimize our system
05:42so that this is going to go straight through
05:44and that it's going to be automated.
05:47But, the more you automate,
05:50the more complex cases are being left.
05:53And these complex cases,
05:55whoever works kind of in processes,
05:58knows that very well,
06:00they are really cumbersome.
06:02And this is where we can leverage,
06:04of course, unstructured data a lot,
06:05be it our standard operating procedures or whatever,
06:08you know, to augment our service people
06:10to be even faster in serving our clients
06:14and serving our customers
06:15and even, you know, fully automate claims payouts.
06:19Or, for that matter, also underwriting decisions.
06:22We are, outside of US and China,
06:26potentially the largest commercial lines insurer in the world.
06:29So, we insure a lot of companies.
06:32Companies are typically more complex than individuals.
06:35So, the same applies.
06:36We need to understand all the risk information
06:39and we need to compile those risk information
06:42from various sources
06:43to really understand and support our clients
06:45to manage their risk,
06:46be it in their factories,
06:48be it liability of products,
06:49be it their supply chain, whatever.
06:51What helps us here
06:52is we can use pictures.
06:57If you want to assess
06:58whether a factory has evolved over time
07:01or is uncertain or unsecure
07:03or is going to be flooded or whatever,
07:04take a picture.
07:07And combining generative AI
07:09in a multimodal mode
07:10with geospatial, aerospatial picture
07:12helps us a lot to be much better
07:14in assessing those risks.
07:16So, here you see
07:17that we can bring a lot of those information
07:20into the core of our insurance
07:22and leverage Gen AI
07:23for better decision making.
07:25Augmented,
07:26and augmenting and supporting our people.
07:28And this has real potential
07:33to transform a large part
07:35of the way we do our business.
07:38But, and this is then
07:39the third part of the story,
07:42this is a huge transformation.
07:45So, how do we run this transformation?
07:47So, first of all,
07:48this whole transformation
07:49is as much about people
07:51as it is about technology,
07:53probably even more.
07:54Because the speed of change
07:55is always driven by the speed
07:57people change their behavior.
07:58So, that is why
07:59we are investing massively
08:02into upskilling our people,
08:04into training our people.
08:05And if you want to put it
08:06in one tagline,
08:07it's training, training, training.
08:10That's what we need to do.
08:11We need to educate our people.
08:13And we need to educate them
08:14in a way that they only can handle
08:17the technology,
08:18but we can only do that
08:20by giving them guidance.
08:22And this is the second part of it,
08:24what I call
08:25we need as an organization,
08:26or we need to strengthen
08:28as an organization,
08:29our moral compass.
08:31Who decides what is correct
08:35in a world where Gen AI
08:36is giving you
08:37a probabilistic answer?
08:39If we are getting more productive,
08:41who is deciding
08:42what we are going to do
08:43with that additional value?
08:45Are we giving it back
08:45to customers,
08:46to society,
08:47to our people,
08:47to our shareholders?
08:48How do we ensure
08:50that the decision
08:51that an AI
08:53or a Gen AI
08:54is proposing
08:54is responsible,
08:56free of biases,
08:57and actually coherent
08:58with our values?
08:59And how do we make sure
09:01that we're not misusing it
09:03because we empower
09:04our people a lot
09:05because they can basically
09:06take even more decision
09:08decentrally
09:09than ever before?
09:11Which is, by the way,
09:12true for all large corporates
09:13and all large organizations
09:15in an AI
09:16and data-driven world.
09:18And these are the questions
09:19where we have to be
09:20very strong on values,
09:21very strong on education,
09:23very strong on risk management,
09:24very strong on sharpening
09:26the few of all of our people
09:29on being responsible
09:30using a technology
09:31that has a huge potential
09:33to transform our organization,
09:36a huge potential
09:38to transform our industry.
09:39And again, as I said,
09:41we want to be at the forefront
09:42of this movement
09:43and not lagging behind.
09:45And this is where I can only say,
09:48referring to the title
09:49of the presentation today,
09:53if you start this transformation
09:55a year later,
09:57you will always be a year late.
10:00Thank you very much
10:01for your attention.
10:02Thank you.
Commentaires