The use of artificial intelligence in banking processes, the results of the first large-scale implementations and the conditions necessary for the further development of this technology were among the topics discussed by Deloitte experts during the 16th European Financial Congress. Examples presented during the event show that AI-based solutions can reduce the analysis of credit documentation from nearly ten hours to just a few minutes, increase the efficiency of KYC and AML processes by up to five times, and automate the handling of some cases submitted to contact centres. The growing use of such technologies is directing the attention of the financial sector towards issues related to data, infrastructure and technological sovereignty, all of which will influence the pace of further transformation.
Banking is one of the industries adopting tools based on generative artificial intelligence most rapidly. Large language models already support document analysis, credit processes, activities related to anti-money laundering and know-your-customer procedures, customer service and complaint management. As a result, AI is affecting both operational efficiency and the speed of decision-making.
“Three years ago, we were mainly talking about experiments, testing models and looking for use cases for generative artificial intelligence. The next stage involved building strategies and assessing the business value of projects. Today, however, we are talking about solutions implemented in production and about processes in which AI performs tasks under human supervision. This is a very dynamic change — other technologies previously needed more than a decade to reach a similar level of maturity,” says Tomasz Filipek, CE GenAI Investment Programme Lead at Deloitte.
In the introduction to the debate “AI-first — competitive advantage or a new arms race in banking?”, moderated by Tomasz Filipek, specific data and the scale of changes taking place in financial organisations were presented. In one project carried out by Deloitte using AI to support credit processes, document analysis time was reduced from 9.8 hours to between 4 and 10 minutes. In the area of complaints, the use of artificial intelligence reduced workload by around 60%, while maintaining compliance with regulatory requirements. Solutions supporting KYC and AML processes, in turn, increase work efficiency by three to five times.
AI-First Does Not Mean Implementing AI Everywhere
As emphasised during the debate, the source of value lies in an organisation’s ability to identify areas where AI can deliver measurable business benefits.
“The biggest mistake would be to start transformation with technology. The starting point should be the business problem the organisation wants to solve and the value it intends to achieve. Only then should it be decided whether artificial intelligence is the right tool. This approach makes it possible to prioritise projects effectively and focus investments where they can bring the greatest impact for the organisation, employees and customers,” says Paweł Spławski, partner and Financial Crime Leader at Deloitte.
During the discussion, three main areas of AI use in banking were identified: process automation, employee support and solutions designed to improve the customer experience. The second of these areas is developing particularly quickly. Generative AI tools support the preparation of analyses, documents and presentations, and also accelerate information retrieval.
Much attention was devoted to changes in the way work is performed. Artificial intelligence is taking over some administrative and analytical tasks, allowing employees to spend more time on activities that require expert knowledge, situational assessment and decision-making. This also means new competence requirements. The ability to critically assess results generated by models and to ask the right questions is becoming increasingly important.
Progress in AI also poses new challenges for financial institutions in terms of security and oversight. That is why, alongside the principle of “know your customer”, the concept of “know your model” is appearing more frequently. It means the need to understand how the tools being used operate and to continuously monitor the quality of the actions they perform.
Technological Sovereignty as a Priority
The second debate hosted by Deloitte focused on the conditions required for the further development of artificial intelligence. Participants pointed out that AI is becoming a new layer of economic infrastructure.
“The discussion about AI is no longer only about models, agents or further business applications. Questions are coming to the fore about where data are stored, who controls the infrastructure, how to manage the cost of computing power and how to build organisational resilience to growing dependence on global technology providers. Artificial intelligence should be treated as one of the strategic areas not only for the financial sector, but also for the entire economy,” says Michał Pieprzny, partner, CE Technology & Transformation, Consulting Market Leader for Poland and CE Nvidia Alliance Leader.
One of the most important issues raised during the debate “Sovereign AI Without Illusions: Cost, Data and Security Control at Scale” was technological sovereignty. Access to computing infrastructure was at the centre of the discussion. Participants noted that the dynamic development of artificial intelligence will require ever greater energy resources, which presents Poland and Europe with the challenge of ensuring stable, accessible and competitively priced sources of energy. It was stressed that without investment in energy infrastructure, it will be difficult to build long-term competitive advantage and safely scale AI.
The discussion also highlighted the growing role of the financial sector as both a user of AI technology and a potential source of financing for investments necessary to build its foundations. Participants also discussed the risk of excessive dependence on individual technology providers and the need to build a more diversified and resilient digital ecosystem.
