The Value of Human Judgement
Wed, 15 Mar 2017 05:41:16 GMT
Current waves of digital technology innovation are offering financial services firms multiple opportunities to automate and reengineer processes, thereby generating new value propositions or improving the customer experience through lower cost or greater efficiency. Artificial intelligence (AI) is the next wave of technology innovation that is beginning to revolutionise other industries, and is being explored for new applications across the finance sector.
Through its ability to make better decisions and recommendations based on the continuous consumption and analysis of structured and unstructured data (increasingly in response to natural language queries), AI is credited with the potential to improve efficiency and performance in financial services, in areas from investment advice to regulatory compliance to information security.
The scalable exploitation of AI requires finance sector firms to re-evaluate the mix of skills and capabilities needed to complete the series of tasks integral to the efficient execution of a process, or delivery of a service. An efficient, multilateral decision-making process, for example, could be said to be composed of three distinct elements: an infrastructure or network to manage the required information and connect the participating entities; a mechanism for sharing and / or structuring the relevant information; and the ability to exercise judgement based upon information assembled or exchanged. Automation has an increasing role to play in the first two elements, through its superior ability to process data. But can artificial intelligence replace human intelligence in the final element of the decision-making process? The jury remains out.
Let’s take a worked example. In the back-office operations of banks, previous waves of technology innovation have partially automated and standardised post-trade tasks such as clearing, settling, confirming and reconciling transactions. But we are still a long way from full straight-through processing. Humans are regularly called upon to fix breaks, exceptions and trade fails in the transaction chain, even for relatively vanilla financial instruments. Moreover, those back-office staff must wade into the realm of the machines before they can utilise that unique human attribute: judgement. In the first instance, the underlying network is often faulty, incomplete and out of date, leaving staff with no clue how to reach the appropriate counterparty at another institution in order to find the vital missing information. And even if they could be identified, there may still be delays and mistakes in sourcing the information needed to resolve the problem, exposing both parties to potential financial, operational and reputational damage. Indeed, such is the effort required to align all the inputs needed to resolve a failed transaction or process that precious little time is left to assess the overall situation carefully and make the right call.
That’s not to say AI has no role in achieving new back-office efficiencies – far from it! At Taskize, we have developed a solution that aims to ease the workload of back-office staff by supporting accurate, timely decision-making, thus enabling more efficient exception resolution. Taskize provides an easy-to-use platform for collaboration, problem-solving and dispute resolution between counterparties and multiple financial institutions. Users benefit from the network effect of having access to contact listings by job title, but our algorithms also learn from experience, rating and ranking the appropriateness of contacts for future tasks according to their participation in past cases. Taskize plays a key role in ensuring back-office staff have all the necessary data to hand in a single, shared environment, but then leaves the final leg of the process to the judgement of the experienced human who has been given responsibility by management for overseeing the process.
After decades of unfulfilled potential, AI and machine learning (ML) programmes have made progress in leaps and bounds in recent years, driven by advances in computing power, big data analytics and natural language processing capabilities. But there are many sensible reasons why the finance sector should be circumspect about its application. After all, it is one thing if an algorithm on Netflix recommends a film that leaves you cold, it would be quite another if an algorithm settled a high value transaction to the wrong account. Quite apart from the increasing regulatory focus on investment advisors’ fiduciary responsibility to their end-clients – as represented by MiFID II in Europe and the Department of Labor’s fiduciary rule in the US – a key constraint is that the decision-making logic of most AI/ML programmes is hard - if not impossible - to identify, isolate and explain. As such, AI/ML is typically used to flag potential issues, solutions or problems to humans rather than replacing experienced staff. In the cyber-security realm, for example, this means AI/ML is used to bring unusual IP addresses or data traffic flows to the attention of human information security experts who will decide whether and how to escalate the matter.
As technology innovation continues, the potential for AI/ML to streamline processes and uncover new insights will undoubtedly be realised. Moreover, the scope for ‘automaton error’ will likely reduce in comparison with that of manual error. For the foreseeable future, that ever-decreasing potential for error will nevertheless demand that the human skills of judgment and negotiation retain a critical place in the decision-making process. At a time when finality of settlement is under threat of reversal in cases of latterly uncovered fraud, and when the systemic stability of financial markets is threatened by the unpredictable consequences of malfunctioning trading algorithms, we still need the human touch to override machines when necessary. Systems can’t tell when they’re being gamed – for now – but humans can use their judgement to weigh up the evidence they provide with increasing efficiency. As such, both humans and machines remain essential to the ongoing transformation of our financial services.