Why 38 Years in Banking Changed How I Think About AI

When I tell people I spent 38 years working inside financial banks before founding a software company, they usually expect me to say technology intimidated me. The opposite is true. Those decades gave me a front-row seat to what technology promises, what it actually delivers, and — most importantly — the gap between the two.

Banks are not slow because of people. They are slow because of systems.

One of the most persistent myths about large financial institutions is that resistance to change comes from people. In my experience, it rarely does. It comes from the systems those people are forced to work within — document-heavy processes, disconnected data sources, approval chains built for a paper world. The people inside these organisations are often far more ready for change than the infrastructure allows.

What AI is actually good for in financial operations

After working with financial workflows for nearly four decades, I can say with confidence that the highest-value applications of AI in this space are not the glamorous ones. They are not predictive analytics dashboards or chatbots that answer customer questions. The real value is in the unglamorous middle layer: document intake, classification, validation, routing, and knowledge retrieval. These are the processes that consume enormous human attention and introduce the most operational risk. AI that handles this well quietly saves hundreds of hours a month and reduces errors that have real consequences.

The mistake most AI vendors make

Most AI vendors approach financial institutions with a solution looking for a problem. They demonstrate impressive capabilities — often genuinely impressive — but they have not spent time understanding the regulatory environment, the audit requirements, the edge cases that emerge at scale, or the political reality of getting a new system approved inside a large institution. This is why so many AI pilots in financial services fail to progress past proof of concept. It is not a technology failure. It is a context failure. When I started v2softech, the thing I was most determined to avoid was building solutions without that institutional context baked in from day one.

What this means for how we build

At v2softech, we build AI systems with auditability, explainability, and operational reality in mind from the start — not as features added at the end when a client asks for them. Every product in our portfolio was designed by someone who has sat on the other side of the table: not the technology vendor's side, but the operator's side. That changes what you build. It changes what you measure. And it changes what you consider done.