Introduction
David Keller spent much of his working life writing credit notes by hand. Not poetry, not bureaucracy for its own sake: reasons. Why a loan was approved, why not, which uncertainty remained. When he saw automatic scoring systems enter the debate, he looked first for the reasons.
David lives in Zurich and now reviews financial AI from a consumer-protection perspective. Retirement did not remove his interest in decisions that change people’s lives.
Story of the Path into AI
After leaving the bank, David read about AI accelerating credit decisions. He wondered whether younger teams still knew how a rejection feels to a self-employed person, a family or someone with thin documentation. He did not want to sound nostalgic, so he learned the modern language: model risk, explainability, bias, appeal process.
His first project was a checklist that connects automatic credit assessments with documented human reasoning. The early version was too long and nobody used it. David reduced it to a few questions: What data mattered? What alternative explanation exists? Can the customer contest the decision? What would a human need to see?
One test case changed his emphasis. A young entrepreneur was marked risky because she had little history. David argued that missing history is not the same as bad behaviour.
Current Work
Today David advises smaller financial service providers and volunteers in consumer projects. He helps teams build processes for borderline cases, complaints and transparent explanations. He is less interested in stopping automation than in preventing unexplained decisions from becoming normal.
The result is better documentation, especially where the model is uncertain. David knows that no checklist eliminates bias. But it can force people to say out loud what they otherwise leave inside a score.
Personal Advice
“A fast decision is only good if it remains explainable,” David says. He advises financial teams to treat every automated rejection as a letter someone may have to read at a kitchen table.
Key Facts
Age and place: 70, Zurich.
Background: banking, retirement, consumer protection.
Entry into AI: checklist connecting automatic credit scores with human reasons.
Focus today: financial AI and transparency.
Typical tools: model risk review, explainability methods, credit processes.
Werkstattnotiz
David keeps old anonymized credit-note templates beside new model reports. The old forms were imperfect, but they demanded sentences. He is testing whether modern systems can recover that discipline without returning to slow paper rituals.