Background, Principles, and Specific Use Cases.
Background, Principles, and Specific Use Cases.
An interview about the background, principles, and specific use cases for artificial intelligence at Contovista.
With advanced data analytics technologies and user-centric solutions, Contovista is able to unleash the potential of artificial intelligence to deliver an attractive value proposition for financial institutions.
The topic, however, is quite complex. How does AI specifically help banks and their customers? What principles need to be considered to ensure that the use of AI is not only productive and innovative but also robust, ethical, and secure? And what is the deal with the new Contovista AI principles?
We discussed these questions with two Contovista AI experts, Martin Biehler (Senior Data Scientist) and Ilario Giordanelli (Data Scientist).
The term “artificial intelligence” covers a broad spectrum. As experts at Contovista, what do you understand by this and what are the benefits?
Ilario: At Contovista, we understand AI as the ability of machines and software to perform tasks that usually require human intelligence. This includes learning from data, pattern recognition, understanding natural language, and decision-making. For example, we use AI to accurately identify merchants and categories based on transaction data. We then transform this data into meaningful, data-driven insights for end customers, allowing them to have better oversight and control over their finances. Banks have access to around 15 analytics features through our five AI-based client analytics modules.
Martin: In the context of AI, it’s important to also speak about machine learning, which is a subset of AI that has been used for a long time. While machine learning has proven effective in pattern recognition within datasets, approaches such as generative artificial intelligence (GenAI) provide context-sensitive results even in areas where they have not been specifically trained. In addition, we need to distinguish this from general artificial intelligence (GAI), which is still a thing of the future.
ML approaches sometimes suffer from issues like the generalisation gap, where a model fails to generalise. Other common weaknesses include overfitting (over-adapting to training data), bias (prejudices in the training data), and data quality problems. Machine learning has made impressive strides and enabled many practical applications, but an innovative use of AI can achieve entirely different results in terms of contextualisation and flexibility.
Why is it necessary to define AI principles for companies? For example, Contovista has recently adopted explicit, binding AI principles for its own solutions.
Martin: These principles help us develop and apply AI technologies in our company in an ethical, transparent, and responsible manner. A top priority for Contovista is to build trust with customers, ensure data protection and security, and make sure our AI systems operate fairly and without bias. To allow banks and customers to understand this, we have consolidated and transparently published these principles.
The AI principles give us clear guidelines and enable us to develop sustainable, trustworthy AI solutions in the long term. By defining and adhering to these principles, we ensure that our technologically advanced products are always ethical, secure, and customer-centric.
Image: Martin Biehler (Senior Data Scientist) and Ilario Giordanelli (Data Scientist) in interview.
And where exactly is this pioneering technology being used at Contovista? Can you give us specific use cases? You mentioned the analysis of transaction data earlier, are there any other examples?
Ilario: There are quite a few, actually. Analysing transaction data is just the beginning. We also use GenAI for the precise identification and verification of merchant names, descriptions, URLs, and contact information. In addition, GenAI helps us with address extraction. We also employ embedding techniques, such as for determining the counterparty category
Martin: Supervised ML methods and gradient boosting models are also used. These techniques are used, for example, to predict the next holiday of customers or to rank customers based on purchase probabilities. Decision trees ensure that the models used are always transparent and explainable. Unsupervised ML approaches also support customer segmentation (spectral clustering, k-means). Probabilistic methods, such as kernel density estimation, are used to detect subscriptions.
This sounds like a wide range of applications that are already up and running today. What hurdles are there when implementing and using AI solutions in banks? And how does the cloud aspect play into this?
Martin: One of the biggest challenges in implementing AI solutions in banks is the decision between on-premise systems and cloud environments. On-premise solutions offer greater control and can better meet stringent data protection requirements. At the same time, however, they are often more expensive and less flexible in terms of scalability. They often require the bank to make additional investments in infrastructure, such as new servers and GPUs, which also need to be maintained.
On the other hand, cloud solutions offer cost efficiency and scalability but also bring challenges in data security and sovereignty. Added to this are requirements such as regulations, IT infrastructure, and skills. As a proven technology partner for financial institutions, we ensure that these aspects are always considered in our products, e.g. through the aforementioned AI principles.
What is your outlook for the future? What use cases do you foresee?
Ilario: At the minute, our customers primarily use AI in areas such as personalised financial advice, risk management, fraud detection, and customer service automation. In the future, we would also expect a greater use of AI for predictive analytics to better understand customer behaviour and then offer tailored products and services. There is significant potential for dynamic financial institutions: if they build the necessary capabilities using our solutions quickly, flexibly, and securely, they can achieve a long-term competitive edge.
With our principles for the development and application of innovative AI solutions, we ensure that this pioneering technology is used with maximum customer focus and the highest security standards. These principles take into account all important aspects discussed in the global debate surrounding AI, including the various planned regulations of different countries and institutions.
Image: Contovista’s AI Principles
