Our latest media interview with James Haxwell of FinTech Global kicked off with our views on the ‘mystique’ that still surrounds machine learning and why we believe the conversation will move away from the technology towards simply finding the best way to solve for business challenges.
We work very closely with financial services firms and partners to get an in-depth view of how machine learning can best be built and deployed to support the processes involved in risk investigations. It is vital that the ‘hype’ surrounding AI and Machine Learning doesn’t detract from the focus of improving processes for financial services firms.
At Caspian, we focused on the business needs and outcomes of financial services firms and specialise in risk investigation solutions. We spent over 24 months with banking partners and financial crime experts to focus on those needs by observing, capturing and testing how expert investigators gather evidence, judge and evaluate evidence, make risk decisions and then explain those decisions. The insights are a a blueprint that led to the development of our Financial Investigation Platform. You can read more about that process in the article but we’ve summarised here what that process helped us to learn:
The first thing that helps to take away the ‘mystique’ around machine learning is explainable outputs. Any financial services firm that is considering, testing or implementing such technology to help provide analysis and judgements for processes like risk investigations will expect that the machine can explain itself. Making sure that this is human readable and evidences all stages of a decision gives invaluable insight to maintain and optimise the technology whilst providing constant learning to expert analysts.
Beyond explanations, the technology needs rigour to validate the complex model risk management of an integrated AI architecture that features multiple models acting in concert. There is limited guidance from regulators to support this process and many banks and AI vendors themselves do not have the expert experience necessary to assess model effectiveness in line with risk policy. As learning develops, this stage will play a critical role in helping Financial Services firms to focus on whether or not technology meets their business and compliance needs.
Even if Banks can make the progress required to validate complex models, there is often the cold realisation that these systems require sophisticated levels of ongoing model maintenance to deliver against an ever-changing financial crime compliance environment. Satisfying external regulators and internal auditors is a given but risk teams can also learn extensively from such maintenance processes. Our Financial Investigation Platform includes variance, resolution, experiments, priors and simulations as part of its capabilities and this is the kind of standard all financial services firms should be looking for as a minimum.
Our mission is to support financial services with the world’s most powerful investigation technologies and part of that is building knowledge and education across the sector. That’s why opportunities to share our insights with titles like FinTech Global are so important.