The great promise of AI in financial services is usually expressed in terms of what it can remove. Less administration, less duplication, fewer manual checks and ultimately less time spent doing things that technology can do perfectly well for us.
All of that is welcome, but there is another side to the argument which deserves rather more attention. If we remove much of the routine work from banking, what else might disappear with it?
Consider underwriting. Experienced underwriters rarely acquired their judgement from a textbook. They accumulated it over years of looking at cases, checking documents, finding inconsistencies and gradually recognising the difference between something unusual and something genuinely concerning. They have seen the perfectly straightforward application that became complicated and the apparently complicated customer whose circumstances made complete sense once somebody took the time to understand them.
Somewhere along the way, administration became experience and experience became judgement.
Technology is now changing that journey. We are increasingly able to collect information automatically, reconcile different sources, identify discrepancies and present people with a much clearer picture upon which to make a decision.
At Ohpen, this thinking runs through the technology we are developing, including Collect & Conclude, but it is also part of a much bigger change taking place across financial services.
The objective is surely right. We should not employ talented people to spend their days chasing paper, rekeying information or checking things that machines can check more quickly and consistently but consistency and judgement are different things.
A machine can identify that two pieces of information do not correspond. It can apply a rule consistently and increasingly it can recognise patterns across quantities of information that no individual could reasonably process.
What it cannot necessarily understand is why an anomaly exists, whether the circumstances surrounding it are reasonable or whether an exception is appropriate.
That is where human judgement becomes more valuable, not less. Yet there is a paradox here. At precisely the moment technology allows our most experienced people to spend more time exercising judgement, we risk removing some of the work through which their successors learn how to exercise it.
I think of this as the disappearance of the administrative apprenticeship. Now, that does not mean preserving inefficient processes for educational purposes but we do need to think much more deliberately about how expertise is developed in an increasingly automated institution.
Perhaps the answer is that apprenticeship changes too. Instead of learning through the volume of ordinary cases, tomorrow’s underwriter might learn through greater exposure to exceptional ones. Technology could identify why a case requires human consideration and allow less experienced colleagues to examine the evidence alongside senior decision makers.
They could see not merely what decision was reached but why it was reached, what information mattered and where judgement altered an otherwise predictable outcome.
Done well, this might create better underwriters rather than simply faster ones. Banking has traditionally developed expertise through people spending years relatively close to processes before progressing into positions where they design, manage and govern them.
As technology absorbs more of those processes, institutions will need to consider where tomorrow’s institutional knowledge comes from.
That makes AI a leadership question as much as a technology one. Efficiency is relatively easy to measure. We can count hours saved, processes automated and reductions in handling time.
Wisdom is harder to put on a dashboard, but financial institutions depend upon it nonetheless.
The next stage of technological change should therefore be about more than removing work. It should be about redesigning work around the things we continue to need people to do and creating new ways for them to become good at doing them.
