Artificial intelligence is revolutionizing the banking sector, forcing a paradigm shift in the way financial institutions operate and manage. This transformation is not without its challenges, primarily that of AI governance revolving around the management of data and algorithms.
In Italy, the monitoring of algorithms and controls represents a constantly evolving reality and requires specific skills to be managed effectively. The figure ofthe Chief Data Officer therefore emerges with renewed importance in the AI era, becoming a key reference point in the definition of business strategies.
But innovation does not stop there: the emergence of technologies such as Generative AI and conversational AI are opening up new horizons for the future of banking, requiring increasingly specialized and advanced skills.
Index of Topics
- Financial institutions and new technologies
- The impact of AI on banking: a paradigm shift
- Algorithm monitoring and controls: the state of the art in the Italian landscape
- The role of the Chief Data Office in the age of AI
- The new skills required by the introduction of AI.
- Generative AI and conversational AI: future prospects
Financial institutions and new technologies
The introduction of new technologies into financial institutions has had disruptive effects and will continue to bring about major changes in the years to come, the credit for which will be mainly due to new AI technologies.
These changes will not only innovate the way banks do business but also profoundly change the way they organize their work. AI tools, in fact, would seem to be able to extend human capabilities through their ability to understand, learn, and reason.
Not for nothing, there is a strong debate going on about the roles AI will play in the workplace in the near future. Some critics of AI speculate that the machine will replace humans, while others believe that AI will, for much longer, flank current jobs in the bank. However, there is a clear understanding on the part of these opposing views: this disruptive new innovation will change the work dynamics and balance in the workplace, including in the financial sector.
To understand what these changes can be substantiated in, it is worth taking a step back and studying the AI approaches of financial institutions.
The impact of AI on banking: a paradigm shift
At the beginning of the AI boom, the approach most used by financial institutions was model-centric. That is, the main goal was to improve the performance of the AI model by optimizing the learning parameters.
The success of a model was perceived and measured by both the design of the algorithm and the sophistication of the actual model. This approach offered few opportunities to systematically and progressively revise and improve data quality.
In a short time, however, the democratization of AI models both within intermediaries and externally through open-source communities has enabled accessibility to advanced models and a multitude of AI providers in the fintech world as well.
Financial institutions then realized that, in this process of democratizing algorithms, the real source of competitive advantage came from the data they were training on the algorithm.
It then began to adopt a data-centric approach based on proactive management of data use issues (let's not forget also the role of regulation, which has been a major driver in this perspective).
Algorithm monitoring and controls: the state of the art in the Italian landscape
The fundamental assumption of this approach is that AI applications are dynamic, as are the datasets that train and monitor them. This results in continuous evolution of the IT architecture on which AI systems are based. The effectiveness of the AI system is then evaluated on the basis of both the performance of the model and the quality of its data. Data are the very foundation of AI systems in general, and their handling requires high levels of responsibility and care because poor choices about the use of data can have very significant ethical and social consequences. In fact, without data governance, the decisions of AI tools could violate privacy, discriminate, manipulate, and misinform. Therefore, the entire data lifecycle and algorithm should be subject to effective governance and monitoring.
Reporting on AI monitoring in the Italian banking and insurance landscape, it can be observed that 75% of institutions have already set up second- and third-level controls, mainly under the Audit area, for algorithm monitoring while only 25% have not yet formalized algorithm control mechanisms. In terms of the frequency of algorithm checks, however, it can be seen that more than 50 percent of institutions decide to carry out checks on an ongoing basis on AI discrimination and performance. In the area of explainability, on the other hand, only 38 percent of institutions adopt a continuous monitoring approach, while an additional 35 percent, who have simpler models, consider it sufficient to conduct checks on an annual basis. The entire financial sector recognizes, however, the great value of ex ante controls, at the algorithm development stage, although only fewer institutions consider these controls as sufficient during the algorithm life cycle.
The role of the Chief Data Office in the age of AI
The Chief Data Office is then also adopting a profound redefinition of roles and skills within it. Alongside Data Governance specialists (54 percent of the workforce in the CDO area), this area increasingly includes Business and Data Translators (10 percent), and Data Scientists (20 percent). The former, are figures with business backgrounds and have an abstract understanding of AI capabilities.
Therefore, they can facilitate the discovery of use cases and act as intermediaries between business functions and AI specialists. In turn, data scientists are employed to support AI solutions. Ethics and Compliance specialists also appear in this area for the governance of ex ante controls and first-level monitoring (1 percent).
Technical teams dedicated to the training and development of AI algorithms also integrate increasingly diverse skills: in addition to data scientists, ML&robot engineers, prompt engineers, computer vision specialists, deep learning & NPL scientists are all technical figures who can find themselves together developing AI algorithms, despite having very different backgrounds.
The new skills required by the introduction of AI.
In addition to the emergence of new figures resulting from the introduction of AI, there will also be a strong need to proceed with upskilling programs of existing employees, including those outside the technical teams. Primarily, there is a need to work on AI awareness so that resources gain an abstract understanding of AI functions. This awareness enables them to conceive of AI as a versatile tool with great potential in their specific context.
For example, employees recognize the importance of high-quality data as a prerequisite for a reliable AI result. Moreover, in a world where employees and technologies will increasingly collaborate closely, digital skills will also become more central. Indeed, the entire financial ecosystem is increasingly moving toward a 'phygital' (physical plus digital) reality , and as banking processes and interactions gain greater levels of digitization, there is a need to acquire more and more digital skills.
Generative AI and conversational AI: future prospects
Finally, the advent of Generative AI and conversational AI are accelerating these changes even more within financial institutions. On the skills front, these technologies are further shifting the boundaries of change, and not only new IT skills will be discussed, but also new interaction and critical-thinking capabilities for all the various areas of the bank. However, given the orders of magnitude, it is still early to dictate predictions. What is certain is that 2024 will be a year of experimentation for some and sustained implementation for others that will require having clear development and governance strategies in place.