How to make the latest tech work for you
APPLYING CUTTING-EDGE TECH TO CAPITAL MARKETS
WHITEPAPER
©2021 FIS FIS and the FIS logo are trademarks or registered trademarks of FIS or its subsidiaries in the U.S. and/or other countries. Other parties’ marks are the property of their respective owners. 1320429
linkedin.com/company/fis
twitter.com/fisglobal
www.fisglobal.com
Up until recently, the application of advanced technology to the post-trade portion of capital markets has added up to little more than an incremental efficiency play, meant to reduce costs and streamline back-office processes. With that approach – and in a heavily regulated industry with decades of legacy technology and practices – it’s understandable that the application of advanced tech hasn’t quite lived up to the hype.
But you can’t capture the full promise of advanced tech if you don’t start in the right place. The foundation of advanced technology lies in the collection and use of data – used for everything from automatically improving the user experience to revealing key insights hidden from human view. It’s no wonder that data is called the new “oil” of markets.
However, given the restrictions on the use of some data, the dispersed technological landscape most firms operate in today and the newness of the technology, it can be extremely challenging to take the first steps forward.
Yet, the possibilities are endless if you build a solid data foundation, embrace new ways of developing systems and commit the investment needed.
Tech Under the Microscope
Will the next pandemic happen in another 100 years or just five?
A different way of working
Although there have been several pandemics over the last 100 years, the most similar to COVID-19 is the Spanish Flu of 1918-1920. But for actuaries, there are still major differences.
Obviously, 1918 predates all the computers and technology we have today. And although the U.K. Institute and Faculty of Actuaries was already 70 years old by then, actuarial science was less sophisticated.
The insurance products of the time were much simpler in design, with no complex guarantees, options or direct links to equity and bond markets.
At the most basic level, AI enables a machine to imitate human behavior, learn from it and improve on it, so that humans can stop wasting time chasing the details. Instead, they can make the important calls, and begin to exercise the kind of game-changing creativity that only humans are capable of mastering.
Unlike other technologies that perform only one function within a defined role, AI is malleable. It can be adapted easily to solve a variety of problems, based on the various information you have to feed it. Internal or external data, it doesn’t matter.
As we enter 2021, there’s a lot more
information
for actuaries to navigate – and luckily,
a lot more computing power to manage it all.
If you can get your hands on the data, you can build an intelligent robot to process and report on in any format that suits your need.
As we continue our path through this pandemic, it’s apparent that modeling must adapt—and adapt fast—to assimilate new and increasing data sets. Actuaries will need to not only model the impact of the novel coronavirus over the next five years, but also make sure that their models and solvency plans are effective enough to handle another global health crisis, no matter what form it takes.
Below are just some of the questions that actuaries must address in a new generation of risk models, scenarios and simulations.
Will COVID-19 be with us for decades but be managed effectively alongside the common flu or evolve in unexpected ways?
Begin at the beginning
Tech under the microscope
Innovation working for you
NEXT
As for solvency, it’s critical that insurers look at both external regulatory models and the internal models they use to manage their business. Solvency II makes a solid connection between the two, but not all solvency regimes do the same.
Why are capital market firms investing in artificial intelligence/machine learning over the next 12 months?
Perhaps the most intimidating aspect of using cutting-edge technology is the enormity of it. The word “modernization” alone is enough to freeze you in your tracks.
Rather than boiling the ocean, pick a spot and start there. Focus on one atomic use case that can make a difference. Cash management, securities lending or AML are all good starting points.
Think strategically
all of which are heightened with a remote workforce.
governance
Not only were largely office-based infrastructures under extreme pressure, but the pandemic also raised concerns about:
security
scalability
Regulatory bodies must also review and fine-tune their models. Both insurers and regulators should take care not to overfit their models to COVID-19. There is plenty of data on the current pandemic and also previous pandemics – from the 1918 Spanish Flu to the 1957-1958 H2N2 pandemic and the 1968-1969 H3N2 influenza pandemic – to better protect solvency across multiple stress scenarios and combinations of stresses.
Future pandemics will happen, but they may not be influenza-based, and they could look very different in terms of frequency and severity, as well as who is affected and for how long. Take care to capture second-order effects, too, such as global recession and psychological impacts.
NEXT
Round up the data necessary to solve that problem, determine a success marker based on the system you use today and build a proof-of-concept to test your case. As you refine your model, you will soon evolve an application that supports your business. Then you can begin connecting the dots to the next application in the value chain, leading to a growing platform of cutting-edge solutions.
Remember, AI is flexible. The models you build today can be reused and refined in other areas, so any investment you make in the process is a long-term benefit to the enterprise.
Should there be flat additions for mortality, or should older age groups and impaired lives be affected differently?
Could excess deaths be spread out over a number of years?
And ultimately, should there be a stronger correlation between mortality and the economic fallout of a pandemic?
This simple model needs to be revisited
for internal risk management purposes.
NEXT
At the time of writing, early results of several vaccine trials have shown high effective rates, in some cases over 90 percent, and some countries have already approved these for use with rollouts planned before the year end or in early 2021. Trials are also ongoing to see whether the BCG vaccination, developed in 1921 to protect against tuberculosis, offers any protection against COVID-19.
Whatever the direct outcome, research into vaccines and treatments will teach us more about how the immune system works and, like all medical advancements, should make a generally positive contribution to lowering mortality rates. Other aspects of the pandemic may also have a favorable impact on overall health – from an increased focus on personal hygiene to less traffic-related air pollution and lower stress from flexible working as more people work from home.
The first task in dealing with AI is to examine your data value chain. With data trapped in so many places in and around the organization, you’ll need to determine where your most valuable information is stored and how you can readily access it. Fortunately, modern cloud storage and enterprise infrastructure have matured to a point where accessing vast pools of data has become routine. Vendors can assist in the effort, providing linkages to support your AI model and feed it with essential inputs.
What’s important to consider as you move forward is the economic benefit you hope to derive from your efforts. Most AI applications in the capital markets space will be augmenting or replacing rules-based systems, such as anti-money laundering (AML) or credit reserve calculators. If AI can improve these systems by just 20%, it’s an economic gain.
Start by determining your baseline. Remember, your data can come from a variety of sources – transactional histories, commodity prices, SEC reports, sensory data, weather bureau information, even leadership changes reported in the news. Find a way to bridge the gap between that data and your AI, and develop the applications you need to achieve your desired outcomes.
Are your terms and conditions clear?
Begin at the beginning
Even now, the pandemic’s impact on mortality rates isn’t well understood. In the first half of 2020, many experts believed that the majority of COVID-19 deaths were linked to underlying health conditions and seen essentially as bringing forward deaths from future years. As a result, it was assumed that future populations would then be slightly healthier, causing the average mortality rate to fall in subsequent years.
That school of thought is already becoming dated. Now, discussions center more on the effects of social isolation on mortality, delays in routine screening for cancer and early cancer diagnosis, so-called “long COVID” and the socio-economic consequences of a lockdown-induced recession. These impacts are not well understood yet but have the potential to increase mortality rates.
What’s the outlook on excess mortality?
Review the resilience of your solvency platform and framework
With infinitely more compute power, data and resources at their disposal, underpinned by public cloud infrastructure, today’s actuaries are better equipped than their predecessors to model both white and black swan events.
Considering all of the current pressures on insurance risk management, I would recommend taking following steps to modernize your approach to modeling.
By definition, however, the COVID-19 outbreak is actually a white-swan event: not the first of its kind and therefore predictable and actually likely to occur at some point.
“Any sufficiently
advanced technology
is indistinguishable
from magic.”
Any change in your business process will demand a change in your workforce mindset. This is where change management comes in. When people suddenly go from reactive mode to predictive, new skillsets must be learned. The fear of losing your job to a robot is real. But if people are engaged in the process, they can weigh in on what parts of their job might be improved through AI and see first-hand the value of it in terms of personal productivity and outcomes.
People tend to think of AI as a black-box computer, working behind a veil, spitting forth outputs from thin air. Magic that can’t be explained is not necessarily good in an industry that is so regulated and routinely audited. Firms must be able to explain how their systems arrived at their conclusions and demonstrate the linkages behind the technology that produced the results.
Explainable AI (XAI) provides the insight needed. XAI is artificial intelligence that can be understood by humans. In a regulated environment, XAI is essential for humans to have confidence in
the results.
The algorithms that go into it must be explicable. The data lineage, feature lineage and model lineage must all be clearly understood. Indeed, the system’s transparency, interpretability and explicability must be drawn from an end-user perspective if the information produced is to be trusted.
Address the concerns
Regulation has traditionally led insurers’ robust approach to solvency, with Solvency II driving stronger governance and control of modeling results. Now, critically, insurers must look deeper and consider all of the aspects of their solvency models that haven’t worked as well.
You must also ask yourself serious questions about the changing nature of the workplace in the pandemic. Will what started off as emergency lockdown measures become a more permanent shift and keep many staff working from home or remotely more often? Should cloud computing and SaaS-based platforms form part of your protection against future challenges to business continuity?
WHAT IF YOU DIDN’T HAVE TO CHOOSE BETWEEN COST AND INNOVATION?
Email us to find out more
Let’s solve innovation today.
When it comes to capital management, you need to know that your reserves can withstand the very worst-case scenarios: extreme, black-swan events that would otherwise cause insolvency.
"That’s a huge payback that you
won’t want to miss."
John Omahen
Head of Product Management for Securities Processing, FIS
previous
NEXT
previous
previous
John Omahen, Head of Product Management for Securities Processing, FIS
But many of the assumptions that make up risk models
are best estimates.
Cutting-edge technology in capital markets has become much more than technologists taking solutions to market. It involves leadership, compliance, risk management, IT and operations working together to change how organizations work, think, innovate, manage and interact.
In 2020, FIS® began a journey in partnership with C3 AI, a leading enterprise AI software provider to help accelerate the adoption of machine learning technology within FIS. This partnership, along with an investment in dedicated data science professionals has allowed FIS to make significant advances in applying machine learning and AI techniques to its capital markets solutions.
Look at the stresses and combinations of stresses that you are running on your models. How did any pandemic modeling you carried out compare to actual COVID-19 results and would it have protected you in previous pandemics? Or should the stresses be higher and more explicitly targeted?
In terms of solvency, it’s important to gauge how well your stochastic and stress tests on models actually protected your company. What’s more, could you have survived any additional shocks after the initial market falls in March? By focusing on reverse stress testing, you can uncover any hidden risks and vulnerabilities that may also cause insolvency.
Assess stress testing
Consider layering on several mortality scenarios, but be prudent and make sure you underestimate for annuitant mortality and overestimate for life assurance mortality.
Some basic assumptions should also be made in the stresses about the pattern of mortality either worsening due to increased deaths from COVID-19 or the impact of long-COVID improving because of improved public health, hygiene and resistance to future viruses or flus.
However, remember that at this stage there is still a danger of over-modeling, overthinking and spurious accuracy, with not enough data or knowledge to support more complex models.
Rethink mortality assumptions
Tying back to your solvency platform and framework, actuaries need to be able to answer senior managers’ most fundamental questions, fast. At the start of the pandemic, C-suite executives were desperate to know what could happen next, the impact on their solvency ratio and whether risk teams could model all this remotely.
To consistently provide a rapid response in these circumstances, you need a risk platform that is well governed, scalable and accessible in the cloud, with the flexibility to model changes quickly.
Deliver timely information
Solvency modernization is now a reality for insurers around the world, with many countries echoing the spirit and format of Solvency II and the upcoming Insurance Capital Standard.
Beyond compliance, these supervisory guidelines are helping create a stronger insurance sector and improving protection for policyholders and the overall management of insurance companies. What better reasons to modernize solvency through both your technology and your risk modeling framework?
Push ahead to meet new solvency requirements
To adapt to a new world of post-pandemic risk, firms must cast a critical eye over their current modeling practices and accelerate solvency modernization.
In the century between the Spanish Flu and COVID-19, actuarial science, modeling and technology have come a long way. By renewing their approaches and taking advantage of modern innovation, today’s insurers can face the challenges, crises and extreme events of the next 100 years with greater confidence.
However, there are still clear lessons
to learn and much room for risk management to improve going forward.
previous
Think strategically
Solvency modernization is now a reality for insurers around the world, with many countries echoing the spirit and format of Solvency II and the upcoming Insurance Capital Standard.
Beyond compliance, these supervisory guidelines are helping create a stronger insurance sector and improving protection for policyholders and the overall management of insurance companies. What better reasons to modernize solvency through both your technology and your risk modeling framework?
Push ahead to meet new solvency requirements
Tying back to your solvency platform and framework, actuaries need to be able to answer senior managers’ most fundamental questions, fast. At the start of the pandemic, C-suite executives were desperate to know what could happen next, the impact on their solvency ratio and whether risk teams could model all this remotely.
To consistently provide a rapid response in these circumstances, you need a risk platform that is well governed, scalable and accessible in the cloud, with the flexibility to model changes quickly.
Deliver timely information
Consider layering on several mortality scenarios, but be prudent and make sure you underestimate for annuitant mortality and overestimate for life assurance mortality.
Some basic assumptions should also be made in the stresses about the pattern of mortality either worsening due to increased deaths from COVID-19 or the impact of long-COVID improving because of improved public health, hygiene and resistance to future viruses or flus.
However, remember that at this stage there is still a danger of over-modeling, overthinking and spurious accuracy, with not enough data or knowledge to support more complex models.
Rethink mortality assumptions
Look at the stresses and combinations of stresses that you are running on your models. How did any pandemic modeling you carried out compare to actual COVID-19 results and would it have protected you in previous pandemics? Or should the stresses be higher and more explicitly targeted?
In terms of solvency, it’s important to gauge how well your stochastic and stress tests on models actually protected your company. What’s more, could you have survived any additional shocks after the initial market falls in March? By focusing on reverse stress testing, you can uncover any hidden risks and vulnerabilities that may also cause insolvency.
Assess stress testing
Regulation has traditionally led insurers’ robust approach to solvency, with Solvency II driving stronger governance and control of modeling results. Now, critically, insurers must look deeper and consider all of the aspects of their solvency models that haven’t worked as well.
You must also ask yourself serious questions about the changing nature of the workplace in the pandemic. Will what started off as emergency lockdown measures become a more permanent shift and keep many staff working from home or remotely more often? Should cloud computing and SaaS-based platforms form part of your protection against future challenges to business continuity?
Review the resilience of your solvency platform and framework
By documenting the mechanics behind a model, AI developers at financial firms can create models that are reliable and defensible – not only to auditors, but also to analysts who might eschew the results as “magic” and carry on in traditional fashion – only to arrive at the same conclusion.
So, the discussion comes full circle to change management, wherein users must be engaged in the process upfront. IT can no longer be the sole proprietor of AI solutions. They must be developed under the guidance of a cross-functional team, including IT, data, automation, business and operational principals. Then the resulting applications can be explained and users can sleep at night knowing that their jobs are not only secure, but greatly enhanced by a machine that can do the dirty work 10 times faster and cheaper, while they perform more creative work.
Analyze large data sets
35%
Build new data-driven products and services
29%
Address the concerns
Imagine you have $2 billion in inventory and it improves by
20%
with AI
Combine alternative data sets and artificial intelligence
31%
Enhance process automation and operational efficiency
33%
Provide more meaningful information clients
29%
Strengthen compliance and risk management
37%
Arthur C. Clarke
The Securities Processing Suite provides a global, real-time view of your operations, with AI-driven risk analysis, digital remote access and simplified integrations. With FIS, you can achieve operational efficiencies, improve risk management and better serve your customers. We’ve taken our best-of-class products into the digital age with operational dashboards, advisor portals and mobile capabilities.
With constant investment in our solutions,
FIS makes innovation work for you.
An incremental approach is always best. Find the use cases that stand out in your organization, build models to address them, and take care to monitor and adjust those models over time. AI is anything but static. Through continued course corrections, you can continue to evolve your applications to a high degree of performance.
“With great power comes
great responsibility.”
Peter Parker - Spiderman
It all comes down to explicability. If your counsel agrees that you are within the limits of ethical behavior and your cross-functional development team has constructed the application around the XAI principles of transparency, interpretability and explicability, you should be able to connect the dots and justify your case. As long as you can explain why the model came to its conclusion, you should have confidence in your final outcome.
AI certainly has the power to drastically change the way financial services are conducted. Because of this, it should be treated with great respect. The ethical issues that arise when you start talking about using personal financial information to manipulate outcomes will cause regulators and society to take note.
So, be aware of the risks. If you are using personal information to populate onboarding documents, for example, the risk is probably low that you will cross an ethical boundary. But if you are citing fraud or using personal data to make hiring decisions, gray areas begin
to arise.
The ethical perspective
NEXT
previous
The ethical perspective
Provide more meaningful information clients
29%
Build new data-driven products and services
29%
Combine alternative data sets and artificial intelligence
31%
Enhance process automation and operational efficiency
33%
Analyze large data sets
35%
Strengthen compliance and risk management
37%
*The 2021 FIS Readiness Report
