Interview with Adina Grigoriu, Founding President, and Hala Najmeddine, Director of Research and PhD in AI at Active Asset Allocation (AAA)

 

Interview conducted by Pascale Caron

Active Asset Allocation (AAA) is a financial engineering firm. It specializes in designing custom asset allocation strategies, with a particular focus on risk management and active capital loss limitation. The company has developed a patented methodology based on Maximum Drawdown, which is automated through artificial intelligence algorithms.

First, tell us about your business and what prompted you to integrate artificial intelligence into your company?

I founded the company in 2010. At the time, we were developing investment algorithms for institutions, with an approach different from what was being done then. Our risk management was based on maximum capital loss, and our models were dynamic, designed for a forward-looking universe. This allowed us to design strategies focused on the future, rather than simply analyzing the past.

Another distinctive feature of our approach was placing the client’s objective at the heart of the algorithm, which was rare almost 15 years ago. At the time, few people asked what objective an algorithm or financial product should achieve. We persisted with this approach, which brought us great success. Today, we advise over 22 billion euros in assets under management for major institutions, such as pension funds and mutual insurance companies. We work both for their proprietary accounts and in creating distributed products, particularly through life insurance contracts. This truly represents the DNA of Active Asset Allocation.

We then consolidated our financial engineering expertise into a digital platform, with the goal of serving savings industry players within a digital ecosystem. This is how we launched our tool called PMS. It is aimed at asset management companies practicing piloted or delegated management, but also at insurers who integrate it into their life insurance contracts. It allows portfolio construction, simulations, investment optimization, and report generation. It meets the growing needs for digital and efficient management in the savings sector.

A third aspect of our business is dedicated to savings distribution. For this, we have developed simulators for distribution players, which integrate sales support tools for customized, often complex solutions. They allow clients to concretely visualize the impact of monthly savings over several years, investing in a specific product or strategy, thus facilitating their decision-making.

Current regulations, particularly the Green Industry Act, push us to move in this direction. This requires us to meet strict standards, and this is where our agility as a Fintech gives us an advantage.

 

Our systems are based on recent technologies like cloud native, and all our calculations are performed via APIs. This approach is much simpler than for companies that must deal with technological layers accumulated over decades.

How did you think about integrating AI?

Hala: When I joined AAA six years ago, we were in full transition. At the time, we were mainly conducting studies for our clients. We encapsulated our expertise into digital methods to meet our own needs and those of our clients.

We then sought to address different use cases, structuring and standardizing our approaches to make deployment simpler. We used the latest technologies, paying particular attention to data security, a crucial aspect for managing sensitive information. These considerations allowed us, over the years, to develop our digital platform.

We launched the first version in 2018, which was built externally at the time. During the Covid period, we decided to rebuild it, particularly during July and August, and delivered to our first major clients. For example, 1,200 MAIF advisors used our platform.

Today, we have a team of about 20 people. What’s remarkable is that most have been with us for six years, and over time, we have built a team with unique expertise. Recently, we have also brought in new recruits.

As for AI integration, it’s part of our DNA as a Fintech. What has changed is the interconnectivity and advances that make AI more accessible and relevant. Since our beginnings, we used simulation and learning models, well before AI was at the center of discussions.

Our algorithms do not fall under big data or generative AI, but rather a branch focused on specific calculations. This allows us, as solution providers, to bring more speed and precision to our calculations, as well as more convergence and individualization. We also eliminate errors and automate repetitive tasks. Back then, it took us a week of work for two or three clients. With the increased demand from our many clients, AI helps us make these processes much more efficient.

 

Do you use LLMs?

Today, we do not directly use large language models (LLM) in our core business, but it is a path we are exploring.

However, we do use tools like Notebook LM occasionally, primarily to assist us in content writing for marketing and communication, among other things. We have established strict rules for their use. For example, we favor paid versions with subscriptions managed by our CTO, and we remain very vigilant about data management. Additionally, we ensure we never include sensitive documents and only handle public data when it is intended to be visible to everyone, such as data contributing to AAA’s reputation.

We have seen the immense potential of this technology and the successes it has already enabled. It would be a shame not to take advantage of it! That’s why we have started exploring more advanced solutions, with ongoing projects to include LLMs in our business.

As an example, we are currently conducting workshops to integrate small LLM-based models. These models will be trained on an appropriate dataset to meet our specific reporting needs. Their performance will be optimized through fine-tuning and continuous adjustments, based on the scoring obtained and the confidence level of the results provided by the models.

 

Have you noticed positive changes among your clients since integrating AI?

Our clients often request detailed justifications about our models and their behavior at different stages. We strive to make this information accessible and understandable, to explain precisely how they work. Everything is transparent, there is no black box, which strengthens client trust. Some were hesitant at first, but once they understood how it worked, they became much more open to AI.

Historically, we were more oriented toward institutional clients, but AI has allowed us to move closer to the quasi-mass market, although we do not work directly with end clients. For example, we provide tools to financial advisors so they can offer customized investment solutions. Our software simultaneously takes into account a large number of parameters, such as the investment universe, management rules, risk level, ESG criteria, and taxonomies, to generate a portfolio in seconds. This portfolio is then compared to the client’s objective to offer an appropriate solution.

This allows advisors to offer individualized solutions, even to two clients with similar economic situations. Each will receive a different portfolio, perfectly tailored to their specific objectives.

 

What advice would you give to SMEs or other companies hesitant to adopt AI?

It is essential to take an interest in AI, as it will be unavoidable in the future. Familiarize yourself with the technologies upstream and clearly identify the use case you want to solve. Understanding the challenges and risks is crucial before getting started. Consulting experts can be wise, particularly to avoid pitfalls related to infrastructure costs, cloud, and data security. A poor technological choice can quickly become a budget headache. Adopt a pragmatic approach: evaluate the expected benefits against the costs, and progress in stages, without expecting immediate results on revenue.