On September 3rd 2019, the AI For Finance event, organized by Startup Inside, took place at Palais Brongniart.
Adina Grigoriu, CEO and Co-founder of Active Asset Allocation, was part of the great panel “Risk management use cases” together with Sophie Elkrief, CIO of MAIF, and Anne Lamotte, My Future Ecosystem Leader at ALLIANZ.
Watch the full panel, moderated by Emma Sezen, Head of AI For Finance Cluster & Partnerships at Startup Inside:
To read the full panel in English:
Announcer: We are now talking about Risk management use cases. For this panel, I’d like to welcome on stage Adina Grigoriu, CEO of Active Asset Allocation, Sophie Elkrief, CIO from MAIF, Anne Lamotte, “My future” ecosystem leader from Allianz and Emma Sezen, Head of AI for Finance from Startup Inside.
Emma SEZEN: Hello everyone! We will talk about the impact of AI in risk management. We have the chance to have two representatives from major Groups: MAIF and Allianz, each of whom is working on a project with Active Asset Allocation, represented by the CEO here. Before starting, Adina, you are CEO of Active Asset Allocation, which raised about 4 million euros this year. Can you tell us a little more?
Adina GRIGORIU: With pleasure! Hello everyone. Active Asset Allocation is a financial engineering company that I co-founded 9 years ago with Olivier Hiezely. We make asset allocation models according to a proprietary methodology that we deploy for institutional investors. We have also created a digital platform that allows us to interact with our customers in B2B, and now in B2B2C. Our great advantage, I would say, compared to what can be seen (we can often be associated with a white label Robo-advisor), is that we can accompany our customers from one end to the other of the value chain. So really, the creation of tailor-made algorithms that can then make innovative products, the fact of being able to deploy them in B2B for our customers, for example the asset managers, so that they can interact with the model by putting their market views. The B2B2C part allows our own customers to interact with their end clients by doing all the classic profiling part, but also the definition of a savings project and then especially access to a simulator that allows the client and his advisor, to see in which measures each decision he can take will affect the chances of success of his project. By doing this, we come to individualize the algorithms around the client. And then of course the last part which is the follow-up, the rebalancing and the individualized reporting, which is very important.
Emma SEZEN: Thank you Adina. So, Sophie precisely, can you tell us a little more about the projects you are working on together and the impact it has on risk management in particular?
Sophie ELKRIEF: Hi everyone. We have two projects that are very important for MAIF with Active Asset Allocation. First of all, a project that I would call institutional. What is this project? Very simply, she – Adina – explained it, they have a proprietary model of asset allocation. My job at MAIF is to allocate assets with insurance constraints, and given the rate environment that everyone knows, it becomes more and more complicated to have financial returns, so we diversify. And one of the methods of diversification that we are testing with the AAA teams is to say: my job is asset allocator, so I will entrust part of my outstandings to people like AAA who will allocate them differently from me – my team and I to be more precise – with constraints that we set obviously because it is the insurance asset, and at that time we’ll diversify, we’ll have a different management. So, it’s very interesting in terms of feedback for us. And related to this project, there is all their technicality, their experience in terms of Big Data, Adina will speak much better than me about the intervention and the impact of Artificial Intelligence in its models, but all of this is obviously behind the concepts that I expose. The second project, which is also very important for MAIF, are the digital tools that Adina spoke about, and a platform from one end of the chain to the other, including the interface between the projects of the clients, the decision-making tools made available to the advisers, and ultimately, the establishment of a management. All these tools, we test them, we find them interesting, there is this possibility to offer a personalized management and also to offer a management quite different from the euro funds, to the greatest number. So here we touch the heart of the DNA of the MAIF, which is that in the current environment (you will say she’s painful with her interest rates but for an insurer it is quite important), we are a bit in the euthanasia of the annuitant. So, if you’re an annuitant, you get euthanized but a little less than the little patrimonies. So, our vision is that if we have the opportunity to offer a cheap management, we will say personalized, and effective to maximize the chances of success of a project for all, we will do.
Emma SEZEN: So, Adina, can you tell us more about the technical part, and what are the specificities of your model finally? And I want to be a little tricky, everyone is talking about AI and doing AI, how is AI there?
Adina GRIGORIU: Yes, I’ll explain that to you. What Sophie was saying above all is that today we are in a specific environment – because AI has existed for a long time, like quantitative finance – and it is really today this combination of having both the power computation, the Cloud and actually AI that has become a bit generalized and allows us to release this kind of offers. But we start at the base of financial engineering, I will give you a concrete example of how we use AI: we have this methodology that applies in general to the asset allocation to manage the maximum capital loss, that’s really our approach. To calibrate a model like this, if we have 10 asset classes, we have more than 10 power 10 parameter combinations that could be used to meet one or the other requirement. So, where the AI comes in is really about finding those optimal settings for the client’s project, for what the client is trying to do. And when we did things by hand, 10 years ago, when we started modeling, we needed about 8 weeks to create a custom allocation model. 8 weeks is long! And if you want to individualize and go to the end of the final customer, it must be done in a few seconds. So, it’s AI that helps us do that, especially genetic algorithms. This will allow us, by successive changes to converge fairly quickly to the result. So, it’s a use that is very practical. Another use is that we are still in the financial markets, and one of the most important things is not to lose. So, we developed advanced stress indicators in equity markets, and there is another type of AI that applies, they are not self-learning algorithms, but it is supervised learning. And what the algorithm learns is the weight to be given to each of the indicators it uses to ultimately maximize its success rate. And when we encapsulate all this with digital, in the Cloud, we realize that it is an incredible war machine because we manage to deploy and actually go to the end and really to the last point which is finally the most complicated: the individualized reporting (we will talk about it with Anne). It is almost easier to individualize algorithms than to individualize a report, because in reporting you have to speak with words, and it must talk to the client about his project. So only AI can do that.
Emma SEZEN: It’s awesome! Bravo anyway for this project that you lead together. Anne, you are Ecosystem leader, can you explain exactly what ecosystem means?
Anne LAMOTTE: At Allianz we have implemented a new mode of operation that is actually called ecosystems and the word operation is extremely important since it is not a new organization. It’s just having a job together, collaborative and agile, with collaborators who have different skills – actuaries, marketers, people of finance – and who aim to work co-located, together on the same open space, on projects but also sometimes on operational actions in order to meet the needs of customers, and therefore customer-oriented. So, the customer has been in the center for a long time, but here we all share, with the shared visions that we can have by our own job, what the customer wants. The goal is to be able to respond as quickly as possible with the skills. To do that, today we have 6 ecosystems that are in place at Allianz, the youngest is 1 year old, the oldest is 2, so it’s a real change of corporate culture, it takes time, it is based on the agile method, the scrum method that we are used to seeing in the IT world and that applies here to the technical world, insurance business. I am in charge of the ecosystem My Future, whose objective is to anticipate and optimize the future revenues of our customers, so we are right in the topic.
Emma SEZEN: That’s great! And then what are the projects that you are currently conducting within this ecosystem?
Anne LAMOTTE: We have a lot of projects that are related, and everyone will identify with it, to retirement: it’s still the strong news around savings. But we also have a project that brings us closer to Adina and AAA, which is a customer-oriented project, and in fact, to allow a client to track the performance of their savings contracts. So, it sounds like simple, but in fact we all know that in the life of a savings contract there are a lot of events, be it events of the customer by itself by payments or buyouts, or market-related events, Sophie spoke very well about that. So, our concern today, and the project we’re running and going very strongly with Adina, is to allow the client to be able to see through individualized reporting, what is happening, what is his performance. Finally, when we defined with him what risk class he wanted to take, he could take, follow it, adjust it, with today a proximity of sales consultants. But tomorrow we know very well that customers want more and more autonomy, decisions they can take, and we see the value of a tool that allows to look at what is the result that I can reach with a choice of allocation to actually measure if I do it or I do not, me as a client. We are on this project for a few months. We will launch a pilot in a very short time. It’s also eagerly awaited by sales people, and I want to emphasize this because we see it on the client and risk management side, but a sales representative who can also anticipate a question from a client because he has seen the result obtained by the investments that have been chosen and advised, is extremely important knowing that we are usually today with a target of customers, for this tool, which has not only euros but a lot of diversification on different types of funds so we see the benefit of being able to have this type of tool at their disposal.
Emma SEZEN: That’s great! It’s really interesting to have both points of view: customers and salespeople. Thank you for this return. Adina, could you tell us a bit more about the technical part and how exactly it materializes?
Adina GRIGORIU: The technical part is a little bit the same as that of earlier, the bases are the same. Earlier, we talked more about model-driven management, and we also developed decision support tools when the client wants to be alone and have a strategic allocation. That’s something pretty new too, because the strategic allocation is as old as the world, so it feels like it’s easy to do, and most people do a kind of efficient frontier min variance. In fact, when we think about it, volatility is not a measure of risk since it means nothing to anyone, not even to the institutional, so we have a little trouble with that. We will take the right measure of risk and the capital loss that the customer is willing to bear, we will of course take into account his SRRI profile as required by the legislation and then we will especially be able, again thanks to the IA, to propose a strategic allocation from 10, 12, 15 assets classes by optimizing them, but here when we look, the efficient frontier does not look like a border at all since when we have 3 assets we already have bumps and depending on the starting point we can end up on a local minimum so ultimately it’s much more technical than we think and the fact of really looking at the financial technique allows finally to offer not only a product that will be a little “state of the art” but also a way to interface with the customer in a fun, friendly way, with a particular digital experience.
Emma SEZEN: Before we leave perhaps Anne, Sophie, can you give us a vision of the Groups and the evolution of this impact of the AI and the projects you are doing right now?
Sophie ELKRIEF: For us if I had to keep highlights, the first essential point is going to be scaling up. If we take MAIF which is a medium-sized mutual, we have a little more than 3 million members who are not all members of the life insurance but well we’ll say that it’s the potential, it makes a lot of people! So, we talk about personalized management, digital tools and then project simulations, there are sometimes several projects per client that evolve over time. All these elements complicate the project and I think that the real risk, the real challenge in the short / medium term is to achieve this industrialization. On the institutional part we are working on, we are not really on a scaling up; it’s more, for us and thanks to people like AAA, opening up to the world and being always “in”. Products evolve, management evolves, in what we did 15 years ago there are things we still do and there are things we do not do at all today. We need to manage more and more important assets, so tomorrow we must be able to manage them with the modern means available to us.
Anne LAMOTTE: For Allianz Group today, Artificial Intelligence is essential in the projects that can be carried out, regardless of the areas. We are truly focused on the client, you have understood with the project that I lead with Active Asset Allocation, but in a more general way on the customer data, on what we know, how we can bring him hyper personalized responses and really contribute to risk management for us regarding the customer. To give another example than that which we develop, we work a lot on the issues of fraud, which are really risks for us insurer, and on which the Artificial Intelligence can bring us enormously in terms of fluidification and acceleration. What we believe a lot is that our challenge for tomorrow is to be able to go as fast as possible because the world is changing, the customer is changing and the more we can integrate these technologies through partnerships as we have with AAA or on our own, the more it will be a differentiating factor tomorrow from our position in the market.
Emma SEZEN: Thanks a lot for your answers!
Thanks a lot to Startup Inside and Emma Sezen for this beautiful event, and a big Thank You also to Sophie and Anne for joining Adina on stage for this panel!