Introduction
Lena Wiegand noticed the trembling hand before she noticed the transfer amount. An older customer at the counter insisted everything was fine, but her story kept changing. Someone on the phone had told her not to mention the call.
Lena lives in Berlin and works on fraud detection while still close enough to branch work to remember what a risky payment feels like from the other side of the glass.
Story of the Path into AI
As a trainee, Lena had little status. When she asked whether unusual pressure patterns could be detected earlier, some colleagues saw her as too young, others saw AI in banking as too dangerous. Lena did not want a system that blocked people blindly. She wanted prompts that helped staff ask better questions in critical moments.
She learned data analysis, interviewed experienced employees and mapped common fraud trajectories: sudden recipient changes, urgency, secrecy, unusual amounts, emotional pressure. Her first dashboard marked suspicious combinations. The early version overreacted to perfectly normal family transfers before holidays. Lena corrected the rules and added a human check before any escalation.
Current Work
Today Lena works in a project group for secure customer communication. The system does not automatically stop transfers. It proposes questions: Does the customer know the recipient? Was secrecy requested? Has there been pressure to act today? Staff can document the conversation and decide how to proceed.
In pilot branches, several risky transfers were reviewed in time without slowing ordinary payments on a broad scale. Lena treats this as the right kind of success: a better pause, not a digital guard dog snapping at every movement.
Personal Advice
“Good AI asks better questions before it makes hard decisions,” Lena says. For her, fraud prevention is not about mistrusting customers. It is about recognizing when fear or pressure is being smuggled into a financial action.
Key Facts
Age and place: 24, Berlin.
Background: bank training, young age, responsibility without hierarchy.
Entry into AI: dashboard for suspicious combinations of amount, recipient and pressure.
Focus today: financial safety.
Typical tools: anomaly detection, dashboards, conversation guides.
Werkstattnotiz
Lena keeps anonymized phrases from fraud conversations, not the numbers. “Please do not tell anyone” appears more often than any model feature she expected. She is now testing how much language from real pressure situations can be built into training without exposing vulnerable customers.