Artificial Intelligence and Digital Transformation in the banking sector

Application of Artificial Intelligence (AI) solutions in the banking sector has transformed the customer experience, allowing for 24/7 customer interaction.
It facilitates identification and authentication of clients, provides support and customer care services through chatbots and voice assistants, helping strengthen banks' relationships with their clients and providing personalized tips and additional information.

As well as front-end applications, artificial intelligence is being implemented by banks within central office functions in order to detect and prevent payment fraud and improve anti-money laundering (AML) and know-your-customer (KYC) processes.
But today, AI is, in particular, taking on the need to automate activities and workflows in the back office of financial institutions: many processes require routine, repetitive tasks, and often have many people processing a single client request. This high level of manual processing is slow and costly, and can lead to inconsistent results and high error rates. Artificial intelligence (AI) offers solutions which can free these back-office procedures from unnecessary costs and errors.

This is the context in which Lutech won the tender for the three-year digital transformation program for one of the biggest European banking groups, based on the use of Artificial Intelligence technologies.

The goal of this project is to optimize and automate as much as possible, thanks to the introduction of cognitive capacities, both the integration of certain company IT systems and the client's back-end processes, making the work of certain teams more efficient in carrying out repetitive and time-consuming activities, consequently improving customer satisfaction of the consumer and business targets.

The requirement was to generate a significant quantity of data in a limited timeframe and to implement a scalable architecture in order to develop cognitive functionalities and incremental use cases over time, without giving up on high levels of precision

This project involved the design and implementation of “cognitive robot” software solutions, based on Artificial Intelligence technologies, and the corresponding application maintenance activities. The first phase involved the identification of three robots which were implemented to respond to three different use cases:

• Automated recognition, classification and sorting of emails (complaints, fraud, disputes)
• Advanced Customer Analytics, for analysis and monitoring of the customer lifecycle
• Automation of transaction requests, such as filling out forms for payments (e.g. transfers, tax payments) from various systems

Underlying the solutions is the cognitive technology of our partner Loop AI Labs and other AI libraries and algorithms. At the architectural level, the key elements are an API Gateway, to isolate the internal components from the external ones, and the implementation of microservices managed by a Kubernetes installation, which allows the system to be made scalable and able to manage resources for the addition of new functions in a flexible and modular manner.

From the point of view of the artificial intelligence techniques used, the main areas of reference are Natural Language Processing (NLP) and Understanding (NLU), in which text is analyzed in order to extract the most relevant text from a semantic point of view and the specific “use case”. These include, for example, techniques involving the removal of irrelevant words (conjunctions, prepositions etc.), filtering with methodologies based on the frequency and relevance of the words in the text, as well as the creation of models which try to represent the semantics of the content in order to better create significant correlations.

NLP is fundamental for AI as it offers people an intuitive way to communicate their intentions to a program. People's unstructured words are thus analyzed and converted into instructions which can be understood by the machine. Once the program has understood and processed the instructions, it responds. The response can be in an intuitive format (for example simply displaying a list of transactions if the query is "Show me the last three transactions on my credit card"), a graph, or spoken/written words in simple language.

Artificial intelligence can help banks become more efficient and effective by reducing costs, mitigating risks and increasing income by applying new technological capabilities to analysis, robots, RPA and report generation, thanks to new use cases which can be implemented by combining business requirements and a team of experts.