Generative AI went mainstream when OpenAI’s ChatGPT was launched in November 2022. What followed was a tidal wave of experimentation, exploration, and excitement—especially in industries such as financial services. But what’s the reality behind the AI hype and what kind of new skills are required?
One of the big game changers, according to Stephan Paxmann, Chief Innovation Officer, Landesbank Baden-Württemberg (LBBW), is that data is now compatible and accessible. Paxmann was speaking at the recent SAP and SAP Fioneer Forum for Financial Services event presented by TAC Insights in Munich.
Providing context
“Data comes in many shapes—PDFs, spreadsheets, audio recordings, videos, handwritten notes, and so on,” he explained. “Traditionally, systems struggled to integrate and process such diverse formats making data compatibility one of the historic challenges of AI in finance.”
Transformer-based models changed that. Suddenly, comparing apples and pears—or PDFs and audio files—became possible. Whether summarizing a 100-page ESG report or extracting highlights from a 90-minute Stanford lecture, generative AI can process and distill complex, multiformat data with ease.
This shift isn’t just about compatibility—it’s also about accessibility. We can now talk to models, and no coding is required. Prompting is the new skill, and context is everything.
“If you just tell your teenager to clean up his room, he probably won’t do it,” said Paxmann. “But if you explain that otherwise he’ll miss out on his pocket money, he’ll understand the context and respond accordingly.”
The key to effective prompting is to provide clear and specific instructions. Vague commands don’t work; contextual ones do. You don’t need to become a prompt engineer or a hardware expert, because prompting will be a baseline feature, a standard requirement.
“It’s very easy. If you don’t know how to prompt, just ask ChatGPT,” said the innovation expert.
Whatever you do, don’t put it off. The next phase is already well under way. Unlike current AI, which is specialized for specific tasks, artificial general intelligence (AGI) refers to machines possessing human-level cognitive abilities, capable of performing any intellectual task that a human can. AGI is able to learn, adapt, and solve problems across various domains, even those it wasn’t explicitly programmed for.
Paxmann cited Sam Altman’s essay titled “The Gentle Singularity,” which maintains that AGI won’t erupt. It is already seeping steadily into our world, gently expanding capacities, reshaping society, and requiring careful design and shared governance to ensure it uplifts humanity. Rather than a sudden “event,” the singularity will unfold gradually, delivering a smooth, incremental transition to superintelligence.
AI in fintech is already pervasively integrated, quietly reshaping our lives. It’s transforming financial services, not with sudden disruption, but through a gradual, pragmatic shift that empowers people, augments human tasks, and requires thoughtful, responsible integration—especially around data, usability, and trust.
Delivering true value
Paxmann’s take is that generative AI is not magic but the result of decades of infrastructure and progress. What makes it truly transformative is its usability for nontechnical users. AI acts as a copilot, enhancing efficiency without replacing humans, especially in regulated sectors like banking.
Public large language models (LLMs) such as ChatGPT are trained on public data. But real value in financial services lies in internal company data—which is rich, regulated, and private.
“That’s why many banks, including ours, are building internal GenAI systems with secure access to internal documents, CRM systems, and proprietary knowledge,” said Paxmann. “The bank is using retrieval-augmented generation (RAG) to retrieve internal data which is fed to the model in real time to improve the accuracy and relevance of responses.”
Paxmann considers it a very promising method. It summarizes documents, generates reports, comments on documents, and even creates bullet points from complex text.
“Strangely enough, the core payment transaction itself isn’t touched much by GenAI,” said Paxmann. “It’s too structured and already automated. But the ecosystem around payments—that’s where AI is thriving.”
He gave two use cases as examples. The first—payments—is the AI side of the story. Each of the touch points, from marketing campaigns to customer onboarding, documentation, reconciliation, and post-transaction services—can benefit from generative AI. AI is already being used in many ways, such as chatbots for handling queries 24×7 in multiple languages, document automation for onboarding, and risk and fraud detection using pattern recognition.
The second use case is about the credit application process, which is infamously time-consuming. For corporate clients, it can take 25 hours to complete a credit application.
“We’re piloting a GenAI-based credit documentation process, targeting a 60% to 65% reduction in time,” Paxmann explained. “AI does the heavy lifting from extracting and analyzing financial history to generating draft proposals while humans still make the decisions.”
Facing the future
The next frontier is autonomous agents—generative AI entities that take actions without human input, each performing parts of a specific process to complete a workflow. Paxmann asked the audience to imagine that each agent owns a wallet, makes decisions, initiates payments (using stablecoins, for instance), and also communicates with other agents.
This scenario is not just imaginary. It’s already happening. Last year, thanks to a combination of agents, generative AI, and blockchain, the transaction volume of stablecoins exceeded Mastercard’s global payments, driving a truly autonomous, scalable financial world.
To be clear, AI still hallucinates, meaning human oversight remains crucial—especially in regulated environments like banking.
Paxmann concluded with some encouraging advice. To prevail in the new world of AI in financials, remember:
- Prompting is power. Learn to provide good context. It’s the new literacy.
- Combine internal data with AI. This is where real value is delivered.
- Agents are coming. Think beyond chatbots—toward autonomous, intelligent workflows.
- To apply generative AI in financial services, break down complex processes into smaller steps—and apply the right tech, at the right time, with the right oversight.
“This is not magic,” he said. “It’s strategic transformation.”