The vast potential of artificial intelligence, especially generative AI, has taken center stage at business and technology conferences. The financial sector is increasingly drawn to the opportunities this technology offers to automate routine tasks typically performed by banking professionals. By doing so, it promises to unlock employees’ capacity, boost efficiency, and free up valuable resources that can be redirected toward building meaningful client relationships and delivering expert, personalized advice.
Inflating the hype balloon can easily lead to the trap of unrealistic expectations. The market is flooded with bold promises to buyers about potential returns on investment and benefits in areas such as process optimization, enhanced customer experience, and increased employee productivity. At the same time, solution providers face immense pressure to stay at the forefront of technological advancement, contend with ever-alert competition, expand their portfolios in line with market best practices and AI-related regulatory frameworks, and deliver services that genuinely address the evolving needs of increasingly demanding clients.
Although there are no signs that the development of AI is slowing down, decision-makers in companies planning to implement or expand existing AI-powered tools still have many questions and concerns, both about practical applications and choosing the right implementation strategy. In the following article, we’ll take a closer look at these issues, addressing the most common concerns within the banking context.
Table of Contents
- The Shift from “Cowboy AI” to the Concept of “Trusted AI”
- Regulatory Challenges and the Impact of the EU AI Act
- The Importance of Transparency and Accountability in GenAI Models
- What Is an “AI Winter”—and How Can It Be Prevented?
- The Role of Metadata and Its Use in Conversational Contexts
The Shift from “Cowboy AI” to the Concept of “Trusted AI”
Stephen Brobst, Chief Technology Officer at Ab Initio Software, referred to the phase of AI development—when developers were experimenting with new AI tools without fully controlling their outputs—as the era of “cowboy artificial intelligence.” The term playfully nods to fans of Western aesthetics, which celebrate a Manichaean worldview of good versus evil and a seductive vision of almost unrestrained freedom, famously captured in the iconic films of John Ford and Clint Eastwood.
According to Brobst, the market is currently in a transitional phase between “cowboy artificial intelligence”—led by developers thrilled by emerging technologies, much like gunslingers exploring the untamed landscape of the Wild West—and the era of trusted, responsible AI, characterized by proven and validated real-life implementations on a large scale. Brobst emphasizes that this moment is truly a turning point for the industry. It applies not only to case studies showcasing the potential of generative AI but to artificial intelligence as a whole.
Regulatory Challenges and the Impact of the EU AI Act
When it comes to trust and responsibility, simply using the technology is not enough—it’s essential to fully understand the associated risks. The introduction of new regulations, such as the EU Artificial Intelligence Act, compels banks to exercise even greater caution. They will be required to demonstrate responsible data management and maintain transparency toward both regulators and customers.
At the start of 2025, the first provisions of the EU Artificial Intelligence Act came into effect. This landmark regulation aims to enhance public safety and strengthen the capabilities of IT professionals working with AI systems. It establishes a comprehensive framework for the development, deployment, and use of artificial intelligence across the European Union. Given the complexity of the legislation, its provisions will be rolled out gradually over time.
For the tech industry, it marks a significant shift. Professionals involved in implementing AI-powered solutions must not only understand the technical aspects of deployment but also grasp the broader risks and responsibilities that come with managing AI, ensuring compliance, accountability, and ethical oversight at every stage.
The Importance of Transparency and Accountability in GenAI Models
As new obligations are imposed on software solution providers, several important questions arise. For example, can we explain the origin of the data used to develop AI tools? What transformations have been applied to the data included in the training set? Are the datasets used for training free from bias?
As IT experts, it is our responsibility to monitor these aspects, ensure compliance with regulatory requirements, and, above all, be accountable to our clients. It is not sufficient to measure the performance of artificial intelligence solely in terms of true positives (TP), false positives (FP), and other metrics; responsible development demands a much deeper level of transparency and oversight.
What Is an “AI Winter”—and How Can It Be Prevented?
One of the key challenges companies face when implementing solutions based on generative artificial intelligence is the market’s tendency to inflate expectations about the value that can be achieved. As Brobst points out, the banking sector often encounters overly optimistic forecasts regarding AI’s impact on productivity. Vendors frequently promise spectacular results—claims of a 40–50% increase in efficiency is achievable—whereas actual outcomes, while still encouraging, tend to fall within the 10–20% range.
The problem arises when exaggerated promises fail to materialize. This can lead to customer disappointment, a loss of trust in the technology, and, in extreme cases, to what is known as an “AI winter”—a period of stagnation driven by unmet expectations. Rather than building a narrative around the prospect of massive gains, a more effective strategy is to focus on small, successful projects that gradually deliver tangible benefits. These initiatives help prove the value of AI, build trust in the technology, and demonstrate that it is more than just a buzzword.
The Role of Metadata and Its Use in Conversational Contexts
Brobst, however, emphasizes that it is worthwhile to aspire to these promising results and encourages banks to invest in pilot projects. He also adds that the team at Ab Initio is closely examining market pilot solutions, including those being implemented for internal use within organizations, aimed at increasing programming productivity and generating sequential code. Developers use AI assistants or other tools to help them solve problems related to a given project, and in return, they receive a code sequence.
When Brobst discusses code, he refers to charts that enable access to data and the collection of information from various sources, but also—importantly—the ability to engage in conversations with AI, not just with data, but also with metadata.
At this year’s Gartner Data & Analytics Summit in Orlando, keynote speakers emphasized that, in order to make informed, data-driven decisions, companies must prioritize metadata. Metadata provides essential qualitative context for data by describing the specific resources. This is particularly important because organizations need to engage in conversation not only using data but also with metadata.
The need to set realistic expectations regarding return on investment following the implementation of GenAI solutions is a broad and crucial topic in the context of business planning. For this reason, we will dedicate a separate article to this series to explore it in depth.
In our next article, you’ll find out more about:
- The value of evolutionary AI implementation in finance
- The complexities of transitioning from the “AI hype” cycle to “AI winter”
- How to approach AI implementation realistically within an organization
- Why it’s better to focus on a “proof of value” strategy rather than a “proof of concept”
Do you want to start harnessing the potential of GenAI in your bank?
Learn more
