We are happy to let machines sort through important identity matching data, discarding the irrelevant data and highlighting the risky entries. They do a great job of it. What about the post-matching process? Can they do that safely, reliably and intuitively too?
That’s the general idea. The current trend in innovation is moving the post-matching process towards automation, towards artificial intelligence techniques. This is where more research and development activity is needed.
This drive is largely because even at the post-matching stage, there is still a large volume of data that needs to be sifted through. Say an organisation’s customer database is 10 million and the initial identity screening process highlights 1% (a perfectly respectable number) of that database as needing attention, that is still 100,000 alerts that have to be checked manually. That is a lot of hours chewed up by manual investigations and therefore a huge cost to organisations.
If robots could do those checks instead, it would save organisations a lot of time and money. Artificial intelligence techniques are sufficiently mature now to predict that we are already well on our way towards this situation. Within five years, it is likely that robots will have been trained and programmed by human investigators to take over much of the post-matching tasks and be making decisions on more than half of the alerts. By automating the post-matching process and enabling robots to make many decisions, humans will only be required to handle more complex cases.
This substantial cost reduction is very compelling for financial institutions. However, there is a balance to be struck between reducing costs by automating processes as much as possible, whilst still maintaining rigorous standards. There have been recent examples of multi-billion fines because of sanction breaches, so organisations know they have to tread very carefully.
What’s next? The current political landscape is highly volatile. Financial institutions have to be ready to react to fluctuating political demand very quickly. As a result, next week’s post looks at how to do that, honing in on the emergence of dynamic reference data.
Also in our series on identity matching