Financial services firms are deeply entranced with artificial intelligence (AI), yet the revolution is under pressure as the industry continues to become more educated and selective about it. Recent research data from WatersTechnology and SmartStream seeks to separate the reality from the magic, presenting new perspectives on the extent of AI’s adoption, its potential benefits, and its prevailing direction in the back office.
Every technological innovation has its time, when perceptions move from initial awe onto questions about how it actually works. Sometimes these little moments can be trivial—think of a virtual assistant misinterpreting a common idiom. Other times they are born of irritation, as when your GPS tries to reroute you—and everyone else—around a traffic jam, causing another one in the process. Often we come to these questions well after the fact. We have all posted to social media only to be bewildered by an ad derived from it immediately afterwards, with the realization that these networks were designed more for oversharing and data monetization than anything else. These inflection points aren’t innately negative; on the contrary, they help us understand technology and frame users’ relationships with it.
In 2019, we are approaching a similar moment with artificial intelligence (AI). From curing disease to combating climate change and solving conundrums around driverless cars, there is a limitless allure to AI. It encompasses a broad umbrella of different techniques and applications, and many among them—from deep learning to robotic processing automation (RPA) bots—are attractive to financial services, where proponents point to cost and head count reduction as well as computational and efficiency gains. Now, several years into AI adoption and with the development of evermore sophisticated AI, firms are asking new questions: “How much competitive advantage does it really offer?”, “Is it suitable?” and “Can its methods, biases and use of data be explained under scrutiny?” Measuring these risks is no longer optional, as clients, board members and regulators become more familiar with the flavors of AI, and concepts such as explainability.
Partnering with SmartStream, WatersTechnology gauged the industry on a range of important questions related to AI, with several salient findings about AI’s growing presence in the back office. The study sought to determine the extent to which perspective on AI has changed, why financial services companies are ramping up its adoption, and what kinds of help—and in which areas—it can provide. Above all, the results found that both interest and expectations are on the rise. Whereas caution was once afforded because of AI’s ineffability, today’s decision-making turns upon almost the opposite approach: whether it can measure up.
Room for Growth
Ideas and concerns about social networks, geospatial services or virtual assistants generally grow out of market penetration. In 2019, these things are either indisputably ubiquitous, or seemingly so, and they are steadily doing the same for AI. Still, there are a couple of critical distinctions to draw.
First and most obvious, digitally native industries have far less work to do in integrating AI into core business processes. Global banks aspire to that description. But, being hundreds of years old, with tangles of legacy systems and employing hundreds of thousands of people makes AI adoption harder. Second, and equally important, tech firms are not nearly as tightly regulated as financial services—and rightly so. Relying on AI to surface your next Netflix suggestion invites far less risk than relying on it to rebalance a multi-billion-dollar institutional portfolio or initiate an exchange circuit-breaker after trading aberrations. The stakes are undeniably higher. Even if the enthusiasm for AI is spilling over, those stakes demand that financial services undertake more development and testing, ongoing monitoring and organizational transition—all of which require time and investment.
For those reasons, it is not shocking that finance’s most vocal, all-in adopters of AI are small-shop hedge funds, while the story for banks and investment managers with far larger back-office operations is more mixed. Industry respondents were neatly split when answering a basic query about AI adoption in the back office: just over one-quarter (26.3%) said AI is live in their operations, and a slightly higher number (27.6%) said they are trialing AI at a proof-of-concept (POC) stage. Another one in five (19.7%) are considering a POC, while the final 26.3% said they have no plans to use AI at all.
Much can be read into those numbers. That 46% of respondents—nearly half overall—are only in early-stage considerations or forgoing AI altogether is telling. Whatever the genuine opportunity to benefit from AI, many firms are holding back. Meanwhile, the live environment result broadly aligns with previous research. For instance, a wider study by consultants McKinsey & Company in late 2018 found that AI adoption is around 21% across industries, though the study also noted growing traction in financial services—along with high tech and telecoms. Indeed, that strikes at the most important result: the highest overall share of respondents took up AI more recently, and are currently in POC. Many firms did their homework on AI, instead of jumping in for a first-mover advantage, meaning live environment AI will likely spike significantly by 2021.
Thirsting for Performance
Next, given adoption splits, it is reasonable to ask how AI can potentially impact the back office. Will it be truly transformative, broadly upending the back-office concept altogether? Will it drive toward back-office optimization as we currently know it? Or will it take on a lesser role, simply doing the “dirty work” incrementally faster and cheaper? These expectations matter. They not only define firms’ engagement with AI, but could also explain why such a sizable minority of institutions aren’t yet actively engaged at all. One of the most common challenges with implementing AI adoption is selecting the right technique or application for its given purpose. It might therefore be less about a categorical, “up or down” question on AI, and more a matter of the options available and what they can and cannot specifically achieve.
AI surrounds us now more than ever. Asking more questions of AI will raise its profile among financial institutions for the better, and though the back office lacks flash, it is the most logical space for investment banks and asset managers to begin the journey
Survey respondents made clear that they are thirsting for stronger performance, as nearly two-thirds (65.8%) noted more accurate processing, fewer errors and greater transparency as top-line impacts for AI. A smaller though still significant number (57.9%) expected reduced processing times, while exactly half believed AI will help redeploy personnel to higher-value tasks. The lowest responses, meanwhile, were more transformative in nature: greater straight-through processing (STP) at 46.1% and stronger support for AI-based applications elsewhere in the enterprise, only 19.7%.
What does this say about expectations, and what kinds of AI are most in demand? Both answers take on a Goldilocks quality. Above all, respondents want AI to be smarter, executing tasks with a lower error rate and greater process insight than legacy tech or human eyes. To a slightly lesser extent, they expect AI to do these tasks faster and with added organizational benefits. But they also view back-office AI as separate and contained, with fewer aspirations at broad STP initiatives—despite STP essentially being a back-office Holy Grail—or linkages to AI being deployed in the front office or by risk managers.
Instead, firms today appear to view AI as doing better as a “black box” at a solution level, rather than revolutionizing the enterprise in a structural way—which introduces more explainability questions and liability—or merely replacing older methods for rudimentary task completion. This reflects the fact that it is still early days for many firms’ AI posture. Accordingly, most back-office AI implementations often target a middle technological ground as well, with RPA (bots) and machine learning—such as natural language processing—in the mix, if not a synthetic combination of both.
Two Sure Places to Invest
After having answered questions of “whether” and “why”, the final piece of the puzzle is figuring out the “where”. The back office hosts a rich stew of functions that are notoriously inefficient or just plain hard to solve, such as corporate actions. But that does not make all of them good candidates for AI—at least not yet. So many in financial services are picking their battles, with survey responses showing that two areas in particular are leading the way: reconciliations and compliance.
From curing disease to combating climate change and solving conundrums around self-driving cars, there is a limitless allure to AI. It encompasses a broad umbrella of different techniques and applications, and many among them … are attractive to financial services
Both of these share a commonality of scale, with firms justifying the cost of the AI investment rather than going after the most historically sticky problems. When asked for areas of potential AI benefit, reconciliations led the way with a survey response rate of 75%, and compliance wasn’t far behind at 73.7%. While each of these garnered an impressive majority at around three-quarters of participants’ responses, the next pair—accounting (51.3%) and cost and expense management (50%) functions—landed only roughly half, while the final sets were still further behind: corporate actions with 39.5%, and collateral management at 36.8%.
In this instance, neither top-voted option is surprising. Reconciliations represent a multifaceted challenge of data volume and processing strength, and legacy systems in this area are often siloed, fickle and inflexible. Much needs to be done post hoc to scrub data or map reconciliation output to other internal platforms, and ultimately to report it. Furthermore, firms today are increasingly looking to incorporate unstructured data from off-exchange illiquid instruments, such as securities finance or collateralized debt, onto their master ledgers. AI can reasonably sit at any—or all—of these pain points, and generate significant improvement.
Much the same can be said for compliance, the breadth of which has exploded in recent years. Here the question is not only around interpreting unstructured data such as names and legal entities, and aligning these to lists or analyzing their activities for patterns, but to do so at speed and while documenting the process. It is an old tale that great compliance provides little competitive advantage. But deploying AI to run these checks faster may do just that. For similar business benefit reasons, collateral management is certainly a favorite to rise from its spot at the bottom in the coming years, while corporate actions is less so.
Going into Battle
AI surrounds us now more than ever. Asking more questions of AI will raise its profile among financial institutions for the better, and though the back office lacks flash, it is the most logical space for investment banks and asset managers to begin the journey. As WatersTechnology’s research has shown, significant opportunities remain for chief technologists and technology providers to convince the uninitiated of AI’s benefits. Many institutions, likewise, are more carefully calibrating AI projects according to practical purposes—reaching beyond small efficiency gains toward greater impact while still exerting proper institutional governance and control. Finally, they have overwhelmingly identified areas ripe for AI progress that currently cost the industry billions of dollars in operational spend every year.
As AI continues to proliferate and financial services face new sources of potential disruption, firms will ultimately look for AI that can help them win in battle: tools that are cost-reductive, right-sized, armed with the apropriate techniques and capable of generating value where human eyes and toiling cannot.
Download the SmartStream whitepaper AI Use-Cases Emerge Across the Back Office