Yannick Even:

Thank you very much and it’s a real pleasure. A big thanks to AOA for inviting me today for the webinar and to my Swiss Re colleagues as well. I want to give a special hats off to our interpreter today that will have the difficult task to translate from a French accent. As you can hear, I’m French, but I’ve been living in Asia for the last 15 years before relocating during COVID last year to Switzerland. As mentioned today, I will present in the next 30, 35 minutes, the result of the application of artificial intelligence across the value chain of insurance by our clients, the insurer across the world, but also highlight some of the key enablers and data analytics strategy that really applied across history, across our group globally in the last five, six years. It have helped us to basically gain many business benefits from a better usage of our data, but also the external data that we have from our data partners so Open Source, leveraging the capabilities that artificial intelligence or advanced analytics brings to our industry.

So I’ve structured the webinar into three parts. First, we will cover a few trends and definition of what I called artificial intelligence and what is going on in this field. The second part will cover a bit more what does it mean for insurance and where do we see the initial artificial intelligence solution emerging and providing value to our industry. And finally I will finish with a few key enablers that basically have help us. So helps Swiss Re group to basically get more value from the advanced analytics capabilities that we have in house but also that we use from our partner. So let’s start with the first part that hopefully give a bit of context of where we are today with artificial intelligence. So when I speak about artificial intelligence, I refer to narrow artificial intelligence. So today there are lots of algorithm and models that translate data into actionable insight using techniques such as machine learning or natural language processing.

And that use data either that exists in our corporate or data that are created by smart device in our smartphone, in our watch, in our cars, in our homes, in all the transport that we use, all the cities we lived in, the factories, the shop all across the supply chain in our laptop, in our hospital, clinics, all the medical device that we use, in our farm, in the food that we eat. There are now a lot of sensors and a lot of data is created every day from all the interaction behavior that we as humans have with all this environment and all this product.

Now all this data is transformed into meaningful action with this narrow artificial intelligence algorithm that are more and more embedded into this device making it smart device. As an example, when I go to my office, I usually use the bus, I never remember the time of the bus. So I will basically real time use my phone and an algorithm will tell me according to where I am and where’s the closest station and what time it is, when is the next bus and what time I will arrive to the office. Similar in my car, an algorithm will calculate for me the quickest bus or the safest bus to take from for me to arrive safely to my destination. So we have now around us a lot of data, a lot of information, a lot of algorithm that transform this data into meaningful insight to help us make better decision, to help us be more efficient but also more and more to help us predict what could happen.

Like traffic, for example, on the way to the office could be predicted very efficiently so that I can change my time of departure or I can change the way that I take. So all of this algorithm have been existing now for many years, but they are built with one function and with only a particular set of data, this is why we call this narrow artificial intelligence, it solved basically one or two problems only and one’s deployed. Basically the model learn only with the data that you have available to solve the particular problem it have been built for. We’re still very far from general artificial intelligence, which is a omniscient machine that fully understand the environment the machine is in and that basically make more than one decision that basically can be close to the human intelligence and can solve a lot of different problems at the same time as it goes through life.

This type of intelligence does not exist and we believe that this will not exist at least in the next decade. However, the narrow artificial intelligence that exists today is less and less narrow and able to solve more and more problem with more and more data. And it’s estimated that this algorithm will contribute to the global GDP up to 15 trillion US dollar globally in the next few years, which to put things into perspective is three times the amounts of money that is collected today by insurance worldwide. So clearly artificial intelligence is here to stay, it’s already in our day-to-day life and it’ll only grow its scope and solve more and more problem including for insurer and for all our ecosystem, for our customer and for society at large. Artificial intelligence have been existing for many, many years. Last century it was mainly techniques that were used by researcher to solve researcher problem.

They were the only one having access to computing power and machines and data that were enabling them to use these techniques and implement them for research purpose. Now since the start of this century and democratisation of access to computing powers through cloud computing, corporate have been using artificial intelligence at scale from using techniques such as machine learning to deep learning to basically better understand the environment they’re in, better understand the business and help them to basically further automate and make better decision. What is interesting to see is that in the last few years, especially with the techniques such as generative AI, such as a Chat GPT for example, as an example of vendors selling algorithm using this technique, AI have been democratised and put in the end of everybody. So now with your mobile you can use this algorithm as I mentioned, to pilot your car, to pilot your health, to pilot your wealth.

You can also use Chat GPT now to ask any kind of questions, summarise a book, summarise a text, give a PDF or ask some question. Make sure that you have not forget any point for your presentation like I did today. So you have a lot of AI tools available at your personal level that enable you to solve personal problem and that’s generate a lot of new opportunity to basically solve more and more problem in your personal life or in your corporate life. But this also bring a lot of more new risk that needs to be mitigated by corporate and this is why it’s really important for the corporate that wants to leverage data and artificial intelligence that they do it the right way, the resilient way with a proper governance and a proper data analytics strategy so that it’s not becoming messy and it’s well managed and the value that AI can bring is basically brought at scale for the organisation.

So speaking about AI today, it’s difficult not to speak about generative AI. So in the last two, three years, the public and the media have become aware of the capabilities that generative AI technology can bring. So one example is Chat GPT, which is a tech generator. I’m sure that a lot of you have already used the tool where basically you can prompt, ask any question or give even a PDF or tell Chat GPT to summarise a book for you. And basically the chatbot will generate text leveraging all kind of data that it have been trained on, which is trillion of parameters and data points that have been used making the response usually quite good in term of output. However, it make mistakes, it’s not yet perfect and that’s why it could be dangerous to use it straight away to basically engage with your customer or automate further your decision process.

So it have to be managed this wrong output, what we call as hallucination needs to be properly understood and managed before we can bring this solution at scale. But it can already solve quite a lot of problems. But there are all the generative AI tools, some that generate image such as Delhi to name again open AI tool. There are some that can generate small videos. You can now on meta basically use these tools to create some small videos that will be fully generated by AI. But what I also find super interesting is the capability for generative AI to create some piece of code in any kind of language. It can also translate a code from a very old language that only a few coders still use such as Cobalt for example, for those who remember those times, there’s still a lot of Cobalt code running in production in the insurance backend, but less and less talent that understand how to maintain this code.

So this is also an interesting area where AI can be powerful to actually translate a piece of code from one language to another to validate code that you have been created to create test scenario and automate also a few technical piece, et cetera. Of course all of this technology is still very early days. So when you look to the Gartner hype cycle, generative AI is at the top of the hype. So according to me and my experience whenever it’s there, it needs a couple of years usually to basically get standardised, get understood, for the vendors to clarify their solution. So definitely everything around generative AI and responsible AI, which are the top of the hype are things that should be now experiments but not necessarily be used at scale yet as there will be a few months, a few years before the industry and the expert can standardise the offering so that it can be used at scale limiting, mitigating the risk around these technologies.

However, there are a lot of tech AI technologies that now are ready for scale that actually passed this hype a few years ago and are now on the, what Gartner called the plateau of productivity, deep learning, machine learning, computer vision, speak recognition, and our techniques that have been used for many years by data scientist and data engineer and that leverage at scale to solve data challenge that the industry have. And this is also very true for the world of insurance. So at Swiss Re we’ve been using this techniques for many years and my team, data science globally have been helping our business to implement these techniques across more than 1,500 project. So it’s important to understand that there are a lot of techniques that are now mature, which most of your AI investment and AI implementation should go into, but then you should still keep a bit of energy and money and talent on experimenting with the new techniques such as generative AI and developed your governance around it.

So all this is great, but it’s a real challenge. The insurance industry have looked to becoming more data driven in nature and look to adopting more artificial intelligence and advanced analytics for many, many years now. And it’s still a challenge despite the focus. Most of insurer are having problem to get the required quality from the data that they used in their business and across their decision process. Most of insurer and talent with an insurer spend a lot of time in searching for data and curating data. Across insurance value chain, there is a lack of standardisation, tax taxonomy definition, different kind of people across the industry can have different kind of definition for the same data that will also put orders in the use of AI.

And that’s why we see for the years to come the investment into artificial intelligence, advanced analytics, data acquisition, data processing, data engineering will stay quite high as there is still a lot of problem to be solved by the industry, but also these techniques are starting to bring finally a lot of value to the insurance industry. So let’s move to the second part where we will cover a bit more. Where do we see artificial intelligence techniques and capabilities bringing value to the world of insurance? Working with insurer as a ranchers Swiss, we have the privilege to actually be involved in a lot of the implementation project of the AI across the world and that give us a good perspective of how AI is used by our clients and by the industry all across the value chain.

AI is now leverage in better understanding the customer needs but also societal needs at large as there are more and more open source data from governments that help us understand the coverage of insurance today, but also where are the gaps. And we can leverage AI to basically transform this data into meaningful insight about our customer needs, about protection needs, about trends that is happening in the world of healthcare, in the world of property and casualty so that we can basically commodify new existing risk pool and create better product that basically fill this need and fill this gap with the help of AI. And here particular think about what we are doing around health and wellness. So creating product that are moving more into prevention with digital wellness data, for example, as an AI engagement such as chatbots to basically coach the customer to better understand the wellness level and prevent some of the risks to happen so that you can keep the people healthy longer.

We have also developed a lot of new products, especially parametric product, fully data driven with AI models especially around the world of property and casualty, from motor to natural catastrophes risk that or travel that we have implemented across the world with the help of algorithm and better understanding of data points that can be used to understand the needs, price, the risk and manage the claims. Then moving across the value chain as a insurer we also use AI and help our client to use AI to further streamline the underwriting with productive modeling. Moving into personalised pricing where the regulation allows it, where we basically use the data from insurance and reach it with external data points and use artificial intelligence such as supervised, unsupervised machine learning to basically better risk assess, ask less question during underwriting as we have more data and more understanding of the risk that enable us to streamline the underwriting decision.

The industry also use AI, of course, to sell and distribute the product to better identified the customer that have needs for particular product that have the capacity to buy, identify the customer before the labs to retain them, propensity to buying model. We use it as well to identify the trend in all the different risk pool to basically be a bit in advance in adjusting our product definition, our risk models so that we keep updating our risk understanding with all the changing environments that we are living in today. A lot of our clients use also AI and some of our AI models on the claim side to basically fast track the claims decision, identify the fraud, automate and save cost. Better understand the outlier and basically manage the claims process decision much better. We also finally use AI as well for managing our asset and better understand as well the wordings that we have in our contract to automatically understand our coverage and simulate as well what could go wrong so that we can manage our risk and portfolio better.

So these are just a few example of how AI is used across the value chain today. To bring value not only to better understand customer needs, but also to engage better with customer, to better understand the risk that we already have in our portfolio or new risk so that we can price and end the right and manage it better. And finally there is a big piece that AI bring on further automate the decision process and to be much more efficient in doing our business and providing our services so that we can provide more affordable protection to our customer. Now I wanted to finish the presentation with sharing a few enablers using the Swiss Re experience that have helped us to deliver AI at scale to our organisation. The first one is that four, five years ago we have actually, from the top, our group and Steve committee have validated our corporate level data analytics strategy that is fully integrated in the business strategy.

And this is based on five key enablers. One is around people, so data culture, so upskilling all our talents across all the organisation for them to understand how to work with AI, how to work with the machine, how to work with data. Then we have data foundation looking at the right data architecture to increase data centricity decision process. Next pillar is data governance. So looking not only at the data quality but also compliance to data and AI regulatory across the world, but also compliance to our own data and AI principle. Then data analytic, my team that basically continued to develop our advanced analytic capabilities and build best practice and tools to help our business to implement AI. And finally the data tech that looks strategically at leveraging value from our tech partners and tech platform. Now, one important things that we looked at, especially our data culture is that the media used to always speak about AI like machine versus human, but that’s actually the very wrong way to look at it.

We know now that the company that combined machine and human together outperform the company that only use machine or the company that only use human. However, even the company that understand that have the tendency to under invest in the human components. So it’s very easy to spend a lot of energy, effort and money on developing your machine capabilities, AI governance, AI capabilities, responsible AI, but it’s very easy to under invest the human element. So that’s basically a combination of both that create the magic and it’s really important to develop your augmented human as a business strategy.

So where are the areas where machine is better than human and you can let the machine do the work. Where are the areas where human are still very much needed and how do you help to upscale all your expert, all your talent to make sure that they’re focused on these areas and they know how to work with the machine, they understand the biases that the machine could have, they understand the issues, they mitigate the risk that the machine can have, especially around risk management, IP legal, et cetera.

I’m very optimist and I’m on the side thinking that human have always shown a unique capability to learn and adapt when given the right environment and tools. So it’s up to the corporate to actually create that and make sure that your talents have access to all the upskilling tools to help them navigate in this new world. One profession that is really ready for this is the actuaries like data scientist. They’re very data driven and they’re here to solve problem with data and statistic. However, actuaries and data scientists work a bit differently.

So what we’ve seen is that bringing the two world together actually create the optimum output for an insurance company where actuaries will think about risk pulling, data scientist will think about individual score. Where actuaries will reach the data gap with their insurance and actual expertise, the data scientist will be able to bring unstructured data, unlabeled data in the decision process. So bringing the two together usually enable organisation to solve in a robust way your data-driven challenge. So this is how it looks at Swiss Re with all this enabling pillars in place, we have been able in the last few years to basically develop a very robust architecture of solution that’s enabled us to deploy AI at scale. And now we have more than one third of our workforce that use our data platform as we have been able to build one data layer, bring all our internal and external data source together.

We still across this journey, but we have more and more users as we have more and more data in the system that’s enabled also us to rationalise and optimised the implementation of artificial intelligence capabilities to create more and more value as we have access to better and better data all across our organisation. And this enable us as well to provide a much better insight and solution to our client. The last things I wanted to cover before we go into Q and A is that another enabler is around responsible AI. So it’s very easy to deep dive into the AI techniques, machine learning, deep learning, but regulators are standing to also look into the way the industry use these techniques to ensure that it is done the right way, it is not bringing additional discrimination for example, and it’s done fairly, that everything stay transparent and that we can explain output, that humans stay in control.

So these are some of the principles that regulator all across the world are looking at when deploying AI. And from principle a few years ago, regulator all across the world, I’ve moved to guidance regulation proposal. We even have now laws that are starting to be voted like a few weeks ago in Europe. The European Commission have actually voted the data act. So they are following basically the trends as AI is getting more and more deployed across industries, the regulator are also starting to regulate, which bring more certainty on what can be done and what cannot be done. We have worked very closely with the regulator in Singapore around responsible AI and we have a series of white paper all free. I put the QR code here at the end of the slide that basically give you a very good checklist of all the question that you should ask yourself across the lifecycle of AI project as you implement AI projects.

So around fairness, ethics, accountability, transparency, there are quite a lot of question that business should ask the data scientist and data scientist should ask the business across the development and deployment of such project in order to make sure it’s done properly. So this checklist is free and I really encourage you to have a look to that. We have leveraged this to develop ourself across our group, a very powerful governance around the responsible use of data and AI starting from principle to targeted startups. We have also catalogs where our employees can check all the data that is available before they actually look around. So they leverage all the data that we already have. We have also modeled catalogs so that we know all the model that we have and we deployed and we can govern them and dedicated governance team, tools and processes to ensure that data is used properly.

And that models algorithm that we deployed and that we built are done according to our responsible AI principle and AI best practice. We have also continuously built seven key AI capabilities that we deployed now with our team all across the world to help our business unit to implement AI project, but also to govern and to also upscale themself. And we have more and more data engineered data scientists in the business unit themself, working very closely with business expert. So on this I want to conclude there are many orders in the insurance industry today for getting value from AI, but I show you a few things that could be done to actually enable AI value and bring it at scale.

But there are also a new AI risk that needs to be mitigated. I put here a list that is non-exhaustive around algorithm risk, around anonymisation around IP copyrights if you use models from external vendors, compliance to imaging regulation that are not very clear yet, there might be social backlash as well across the different countries, AI is not viewed the same way. And there are some countries where the population is very keen to use AI on a daily basis, where some orders less keen. So this needs to be taken into consideration and this is why it’s really complex to actually build a AI model globally and then deploy it locally. It doesn’t really work like that.

You can have a global governance approach, but you need to have a local implementation approach because of the local regulation, because of the local societal environment that you are in and local regulation, especially around responsible use of data and AI, you will have to basically implement this locally.

On that I would like to close now and wishing you a great journey in getting value out of algorithms and it’s up to us to decide as an industry what we want to do with it and how far we want to go. But most importantly, we really need now to upscale our people to understand how to work in this new AI normal and to upgrade our corporate data native strategy, governance, actuarial model to mitigate this new human AI risk.

The above text has been produced by machine transcription from the webinar recording. ICMIF has made every effort to ensure that transcriptions are as accurate as possible, however, in some cases some text may be incomplete or inaccurate due to inaudible passages or transcription errors. Listening to or watching the webinar recording will allow you to hear the full text as delivered during the webinar but this is available in English only. Our transcriptions are provided to enable members to select the language of their choosing using the dropdown menu above.