In 2004, I was managing a mortgage origination platform, the e-trading system within the Mortgage Capital Division of an investment bank in London, that captured loan applications and provided acceptances in principle to mortgage brokers. The systems were already pattern-matching and making predictions though we were not calling it AI yet.
I started my career at Real Asset Management (now called MRI Software) in London, UK, supporting enterprise accounting and project management software — the kind of mission-critical infrastructure that taught me what makes systems trustworthy in the first place. The e-trading platforms, the AML risk-rating systems I worked on later at Lehman Brothers and Kaupthing Singer & Friedlander were doing what AI systems do now: ingesting structured and semi-structured information at scale, producing decisions, surfacing exceptions, and shaping outcomes for people who would never see the code.
That early experience shaped how I think about AI today. For me, it has always been about more than the surface of the technology, and I am not surprised by what it can do. I am interested in the question that has always mattered: will this system help people do what they do more efficiently? Will it contribute to making society better and safer?