Artificial intelligence: the shape of things to come
AI: the shape of things to come
Many financial services industry firms are examining the potential of distributed ledger and artificial intelligence (AI) technologies. Is it too early for any meaningful deployment? Heather McKenzie, editor of Daily News at Sibos, investigates.
If distributed ledger and blockchain was the flavour of the week at last year’s Sibos in Geneva, this year AI looks set to take the technology hype crown. In the wider world, there is plenty of interest in AI, with some warning of a dystopian future in which machines rule the world. In the here and now, AI in financial services still has some way to go and most of the developments are more focused on using AI as a supplement to human activity.
There is a focus on AI now because three elements have come together and created an inflexion point: computing power, sophisticated algorithms and vast amounts of data, says Matthew Davey, managing director, global head of business solutions, Société Générale Securities Services.
However, there is a danger that AI is a “solution looking for a problem”, he says. “Financial institutions need to focus on the business requirements of clients. There is a risk that with all the hype surrounding AI, people will view it as a solution to everything.”
There are areas where AI could be particularly useful, such as fraud detection. AI can identify patterns in vast amounts of data and recognise potential fraudulent transactions more quickly and accurately than human staff. AI can also be applied in customer support, adds Davey, synthesising data across multiple systems internally and producing cross-selling opportunities or client sentiment analysis.
Industry analyst Celent’s August 2017 report, Artificial Intelligence in Banking: Where to Start, advises that relatively few banks have begun production or even full-blown research into AI. “For those who think they’re lagging, the good news is that they’re not – there’s still some time. But make no mistake: lack of progress today doesn’t mean that banks can afford to ignore AI; they must formulate a strategy to deal with the opportunities that it promises to them and to their competitors,” the report states. The principal business justification for AI lies in cost reduction, but it can also mitigate risks and increase revenues (primarily through the indirect route of improving customer experience).
Celent is pragmatic in its definition of AI as technology that makes inferences and decisions that previously required direct human involvement. A series of fundamental and interrelated technologies around machine learning and natural language underpin all AI. Building off the fundamental technologies to apply them in a banking context yields four main AI applications: analytics; bots; robotic process automation; and report generation.
As daunting as AI may seem for many banks, Celent believes that most institutions should be exploring at least the basic forms of the many technologies that fall under its umbrella. The main and most basic benefit is reducing cost, but firms shouldn’t ignore risk mitigation and increased revenue.
AI will have an impact on front, middle and back offices at banks. In the front office, direct customer interactions can either be addressed by AI directly (via chatbots and virtual agents), or by enabling humans to do their job better, or by augmenting their capabilities. While the benefits are tangible, there are risks in exposing AI directly to customers. If the implementation is not flawless, the mistakes may earn the bank public ridicule or damage its reputation among customers. AI that helps front-line employees, but is hidden from the customer via employee mediation, lessens this risk.
In the middle office, AI can be applied to report generation, underwriting and credit decision making, and risk and compliance monitoring. Identifying exceptions (and becoming better at doing so) is one example of AI helping employees become both better and faster at their jobs. The risks of middle office AI are relatively low from a consumer standpoint if outcomes (such as false positives) aren’t degraded. Celent says for any middle office AI-based activity, human intervention is critical. “Internal reports generated via natural language processing, for example, should only be considered first drafts; the responsible analyst should sanity-check them, revise them for tone and voice, and ensure that he or she knows the substance of the report.”
Processing and reconciliation, typical back office functions, can use AI to detect anomalies and exceptions, says Celent. Layered on top of non-AI robotic process automation, they can serve as a back-up and a second set of eyes to ensure that processes are proceeding as they should. The biggest risks in back office implementations lie in over-reliance on AI and in the cost associated with putting initiatives in place.
AI: too early for meaningful deployment?
Joshua Satten, director at Sapient Consulting, says finance and securities industry firms have set up innovation labs and proofs of concept to test distributed ledger and AI projects to “varying degrees of success”. He believes the future success of these technologies will be in utilising both in coordination to deliver optimisation to a firms’ operational model and risk management abilities.
Distributed ledgers have three use case scenarios, he says: interoperability with other companies as a shared peer to peer ledger; internally across subsidiaries and regions as a basis for enterprise architecture; and in conjunction with external permissions and provisioned access to facilitate the release and request of data and other information with counterparties, clients, investors, regulators, and more.
“AI on the other hand is emerging on multiple levels across multiple use cases, including from the basic aspect of replicating manual people-driven activities, to the next step of continued advancement with machine learning, and ultimately to deep learning where giant data sets and differing semantic layers are combined for wholly unique aggregation and predictive analytics abilities,” he says. Natural language processing (NLP), virtual assistants such as chatbots, and robotic process automation (RPA) are beginning to reach some level of maturity within financial services.
“One example of bringing together distributed ledger and AI would be contract review and execution management whereby you would employ some AI using natural language processing to read, reconcile, and accept or reject documents. This would be used in conjunction with an internally distributed ledger allowing for approvers in different groups or entities within a financial firm, including compliance and procurement, to subsequently reject, approve or review the documentation. Also subsequent processes, such as vendor notification, payment release, and notification to internal project management groups, can be automated.” Controls and Chinese walls could be maintained despite automation being embedded within each group and all groups being connected and sharing a single ledger.
“As with other emerging technologies, it will probably be two to three years before we see either come to any disruptive fruition, whereas the application of these technologies in tandem is most likely to be five to ten years out,” he says.
Peter Farley, communications director at AI start-up MarketsFlow, has similar views. Both DLT and AI have the potential to deliver transformational capabilities for the capital markets, both in terms of significant and sustainable reductions in costs as well as strengthening security, reliability and decision-making capabilities, he says. “For the time being the hype that surrounds AI is likely to have both more foundation and deliver earlier tangible results, partly because many parts of the industry are more familiar with its use and as it has become more readily accessible.
There are already many examples of AI and machine learning being deployed across the transaction lifecycle, from intelligent automation in the post-trade environment to augmenting pre-trade decision-making and investment strategies. In the back office, AI can accelerate clearing and settlement, prevent and correct errors and significantly enhance compliance activity by more accurately implementing regulatory mandates. In the UK, the Financial Conduct Authority (FCA) has acknowledged this by launching a discussion on issuing new rules in digital AI formats so they can be directly embedded into bank IT systems. “This could be a serious game changer if there is wider adoption,” says Farley.
AI, or quant- and algo-led, trading strategies have found significant traction among hedge funds and other asset managers. “To date these have been mostly confined to equities and much of it using big data techniques and sentiment analysis. However, there are indications that this, too, is widening to include market sensitivity capabilities as well as inroads into other asset classes. There are sceptics that say this has delivered some success in recent years due to low market volatility and benign rate environments. Time will tell as they become more tested,” he says.
Securities industry technology firm Broadridge has made significant investments in and has been experimenting with distributed ledger for some time, says Vijay Mayadas, head of global strategy and fixed income at Broadridge Financial Solutions. “We have evolved from testing of the technology to developing minimum viable products that can be integrated to production systems. For the most part, throughout most of our use cases, we concluded that distributed ledger has the potential to deliver a step change in process improvement.” In some cases, such as the emergence of digital securities native to the blockchain, the potential impact of the technology could be “even more dramatic” than the improvement of current process and lead to the re-imagination of the entire ecosystem, he adds.
Earlier this year, Broadridge successfully executed a blockchain pilot focused on proxy voting in partnership with JP Morgan, Northern Trust and Banco Santander. “As it relates to proxy voting, we proved during our first phase of the pilot that DLT has the potential to drive significant benefits for all participants in the process, including shareholders, boards of directors, corporate issuers, regulators, and broker dealers,” he says. These benefits included increased efficiency by reducing the complexity of the reconciliation process; enhanced security via encryption and increased transparency around vote confirmation. Additionally, for corporations, DLT can provide data and analytics about key drivers of voting behaviour, which can have a significant impact on the corporate governance landscape.
“One of the keys to unlock full potential of DLT and drive wide adoption, not just in proxy voting but in all areas of capital markets, is the development of industry standards to ensure harmonisation of the new processes envisioned via this new technology. The main driving force behind the setting of these standards will be the key participants in the particular industry, such as Broadridge in the case of proxy voting,” he says.
Rob Palatnick, chief technology architect, DTCC, says DLT’s promise of a distributed and shared ledger with built-in integrity, security and resiliency would be “an enormous step forward from the complex financial ecosystem that exists today, which requires continuous reconciliation and re-validation of financial transactions”. But it will take time for the technology to mature, for the range of needed standards to emerge and for the industry to adopt them, he warns. “DTCC is working with the industry to move forward on those objectives and the progress to date is probably commensurate with similar technology innovations such as database technologies in the mid-1970s, the internet and network standards in the mid-1980s and mobile technology in the 1990s. The timetable for those technologies to move to widespread use and adoption is likely a pattern that will be repeated here.”
In terms of AI, DTCC has been working on initiatives to collect sufficient data that can be leveraged for machine learning and AI tools. There are opportunities with repetitive operations in support of daily processing of millions of trades, and to support risk management functions, but it is still too early, says Palatnick, for DTCC to determine how it can leverage those tools to provide benefits to its owners in the financial industry.
This article is also featured in the Daily News at Sibos 2017 – Day 1 edition.
Click here to read the issue online or pick up a print copy if you are at the conference!
The digital and print editions are free.
