MIRAI - first Eagle app based on generative AI

Before diving into the platform itself, let's take a moment to discuss generative artificial intelligence. This technology enables computers to create new content based on input text, images, or data. It functions as a creative process, allowing for the generation of responses, texts, or recommendations based on an existing database.

The Research Division within the RBI Group is divided into several smaller departments, each focusing on different economic aspects. Together, they form a leading local market knowledge base in Central and Eastern Europe.

DRP – short for Digital Research Platform – is a platform that provides our clients (including other banks, institutional clients, and retail customers), as well as Raiffeisen employees, with access to a vast collection of publications gathered by experts. This contributes to a better understanding of markets, informed decision-making, and a broader awareness of global trends and standards, such as sustainability (ESG).

The main focus of the workshops in Warsaw was to simplify navigation within the platform and enhance the way DRP users can find articles of interest in the growing database. This led to the creation of a solution – the AI Assistant, a new module based on generative artificial intelligence.

Several teams played a key role in bringing this project to life:

• Advanced Analytics – responsible for developing the core response generation engine on the APEX platform.
• Research Team – creators of the Digital Research Hub (DRH), essentially the backend responsible for storing the history of all chats. The DRH forwards user queries to the response engine and then saves them along with the answers in the context of each user's conversation. It also maintains and provides access to the publication database
• Eagle Team – the frontend provider, responsible for designing the graphical user interface for both clients and employees of our group.

During the workshops, we had no idea that our team in Poland would be directly responsible for this particularly exciting project. At that time, we were already deeply engaged in building and developing an application dedicated to another area—Trade Finance for institutional clients.

The opportunity to co-develop a tool based on generative artificial intelligence was initially a challenge, especially in terms of balancing our work across different domains. However, we recognized it as an exciting opportunity, and as soon as we achieved our main goals in eTradeOn IC, we decided to focus on improving DRP.

Stefan Theussl took on the role of business owner for the new initiative. Together, we refined the concept of our AI Assistant, leading to the creation of MIRAI – "Market Intelligence and Research AI."

We agreed that the MVP version of the application would be available for internal users at the beginning of Q3 2024. The initial phase, with a limited set of features, allowed us to take a cautious approach to generative AI, a new addition to our application portfolio.

The project timeline posed a challenge, but we successfully met our deadlines. As a result, starting in July, all Raiffeisen Group employees can request the appropriate role in IDM. Once their request is approved, they gain access to the MIRAI application, which allows them to:

• Submit queries on topics of interest
• Receive text-based responses, including links to relevant articles from DRP
• Regenerate the last response if it does not meet expectations
  

Looking ahead, we want our AI Assistant to continuously improve by learning from user feedback.

It is also worth noting that MIRAI is being developed in alignment with the new design standard – GDS.

Before completing the MVP phase, we had already begun discussions about the future of the project. At the end of May, key stakeholders from Vienna visited us, and together with Dorota Strauch, they helped us define and prioritize the next set of features. During two days in Warsaw, we refined our roadmap and set goals for the coming months.

In the second half of last year, in collaboration with the Research Team, we introduced new functionalities to MIRAI, including:

• Chat history and management (renaming, deleting, and marking favorite conversations)
  
• User feedback on AI-generated responses

In Q1 2025, we plan to conduct interviews with clients to evaluate the accuracy of our AI engine's responses and overall application performance. We expect these insights to help us better understand user needs and guide future improvements.

We are also gradually expanding the MIRAI user base and will soon invite institutional and corporate clients to explore the platform.

MIRAI is an excellent example of how collaboration between different teams – from analysts to developers – can lead to the creation of an innovative solution. By combining knowledge in market research, AI technology, and application development experience, we have created a product that could revolutionize how we use databases at Raiffeisen.

The MIRAI project also demonstrates the importance of a careful introduction of generative artificial intelligence into business applications, with a step-by-step development approach. Focusing on simplicity and usability of the application allows us to collect valuable data on its functionality. Further development, based on user feedback, will enable us to tailor the tool to their needs.

MIRAI is an example of effective teamwork and the combination of technology with business requirements, underscoring the value of investing in innovative solutions.