Credit institutions around the world are advancing in digitization with the help of artificial intelligence. In doing so, however, they neglect the majority of their valuable data treasure: unstructured data. Here in particular there are many opportunities to take advantage of.
Data-driven operations are being driven by financial service providers around the world because they promise solid business models. According to one, banks are currently giving Plumber Study It spends more than 10 percent of its income on developing new technologies. However, the artificial intelligence strategy of many financial institutions is not satisfactory. The reason: data prepared for machines is used almost exclusively and therefore easier to read, the so-called structured data.
a Untapped potential It lies in raw data such as documents, presentation slides, emails, or audio recordings, the so-called unstructured data, made available as part of big data. Computer hardware and software are difficult to research and evaluate because they are not prepared for a specific IT context. according to Study by Fintech Futures Almost 80 percent of the current data volume of financial service providers.
However, many financial institutions do not yet take this large amount of data into account in their operations. Only three percent of the unstructured data held in banks is currently analyzed. This is especially due to the fact that many institutes are fully using and integrating more core AI models. However, decision makers must indeed make unstructured data a major strategic priority.
Enrich existing AI models with unstructured data
The most obvious strategy for Use of unstructured data For decision makers, it lies in enriching already produced AI models with insights from unstructured data. Because many types of traditional AI can only generate limited knowledge because the underlying data is of homogeneous origin. Using more discriminating parameters such as motion profile or node data can help to cover boundary cases reliably.
A concrete use case for this is to enrich smart algorithms in fraud detection. In particular, files on payment transactions, information on the accounts of senders and recipients, as well as information on existing fraud schemes are currently used in this area. By using unstructured data in the training process of such models, insights from the meta-level can be obtained. For example, using location data and perhaps even entire traffic profiles can identify fraudulent transactions more accurately.
Identifying trends through comprehensive observation
A special added value is opened by the fact that unstructured data is also comprehensively analyzed. Covering a wide variety of data types from documents to videos, connections can also be made between different data sources.
This is specifically used, among other things, in determining individual customer needs. For example, terminations can be prevented, basic financing preferences can be identified, and financing requests that have arisen as a result of a customer’s individual life situation can be identified. The basis for this is the use of different data sources, which reflect the customer in a differentiated manner across several qualitative levels. For example, by assessing his banking habits, developing his life stage and communicating with his banking advisor.
Using external data that goes beyond the individual customer, financial service providers can also identify other trends. For example, HSBC already has one in 2020 Artificial Intelligence Fund Advanced, which makes purchasing decisions based on an evaluation of unstructured data. With the help of artificial intelligence, this financial product analyzes newspaper articles, Twitter posts, satellite images, and even the CEO’s data tone and uses them to make decisions.
This example shows: Evaluation of unstructured data must also be enhanced through the use of traditional AI methods. In this case, for example, news articles are evaluated using natural language processing (NLP) and audio snippets from CEO data are processed using intelligent speech recognition.
Organization and IT infrastructure are critical pitfalls
However, the fact that only three percent of unstructured data is used in financial institutions has two main reasons. First of all, especially in Germany, financial institutions are not allowed under data protection guidelines and other regulations to use every type of data to improve their AI models. For example, motion profiles or audio recordings may only be processed to a very limited extent. In this context, it is not clear to many organizations what unstructured data banks can use to generate added value. Therefore, the following applies to decision makers: An organizational overview of their data budget is essential in order to be able to develop profitable AI models based on different types of data in the future.
In addition, many financial institutions have insufficient IT infrastructure. Unlike structured data, which can be stored in SQL databases, unstructured data has to be stored in large collections of different data types, the so-called data lakes. Creating a storage structure is difficult for many banks due to the high costs, because, among other things, much more storage space is required for unstructured data.
Unstructured data belongs in every AI strategy
The introduction of smart forms based on structured data is a major concern for banks across the world. It is inconceivable that so many institutes would dare to enter the world of vastly more complex data and enrich their artificial intelligence with unstructured data. Many financial institutions already see developing models using traditional AI as a huge challenge.
However, the following applies to decision makers: Anyone who does not currently have the ability to specifically use unstructured data should consider this strategically, at least in the medium term. For example, through a detailed overview of the internal and external database, and above all through a data strategy tailored to any type of data. Because without preparing for this next evolutionary step of AI, many financial service providers are at risk of losing touch in an increasingly intelligent financial world.