This article was co-authored by Dr Mehrzad Mahdavi and Hossein Kazemi, Phd, CFA from FDP Institute, and William J. Kelly from CAIA Association
Digital transformation, with artificial intelligence and machine learning technologies at its core, has created unprecedented challenges and opportunities across all industries. The financial sector will be profoundly affected as demonstrated by numerous use cases including:
- Trading and portfolio management applications include asset allocation; security selection; risk management.
- Operations-focused (or back-office) applications include capital optimization; model risk management; and market impact analysis.
- Customer-focused (or front-office) uses cases include credit scoring; insurance; and client-facing chatbots.
- Regulatory compliance (RegTech) uses by financial firms or by public authorities for supervision (SupTech).
As an example, asset managers using artificial intelligence and machine learning, are found to be growing their revenue 1.5 times more quickly than the rest of financial services. Asset managers use alternative data to gain an “edge” over competitors that use traditional or “old” data sources such as quarterly corporate earnings or low-frequency macroeconomic data. Going forward, systematic strategies deployed by quants such as alternative risk premia, trend following, and equity long/short will increasingly adopt machine learning tools and methods along with alternative datasets.
While alternative data is thought to provide the most value for short-term investors such as hedge funds, long-term institutional investors, by focusing on defensive or longer-duration alpha strategies, can also benefit from its systematic use. For example, ESG oriented portfolios based on machine learning techniques are becoming a vital part of the investment landscape, with the twin goals of medium-term returns and long-term positive impact on the planet and society.
Drivers of Change
Adoption of AI and machine learning in financial services are driven from both the supply side and the demand side. On the supply side, financial market participants have benefitted from the availability of AI and machine learning spurred by tools developed for applications in other fields. These include the availability of computing power owing to faster processor speeds, lower hardware costs, and better access to computing power via cloud services. Similarly, there is cheaper storage, parsing, and analysis of data through the availability of targeted databases, software, and algorithms. Also, we have seen the rapid growth of datasets for learning and prediction owing to increased digitization and the adoption of the Internet of Things (IoT) in industrial and consumer markets.
On the demand side, financial institutions have incentives to use AI and machine learning to improve efficiency and penetrate new markets. Opportunities for cost reduction, risk management gains, and productivity improvements have encouraged the adoption of digitization. In a recent study, financial services firms described priorities for using AI and machine learning as follows: to optimize processes on behalf of clients; to create an ecosystem where applications of AI lead to better decisions; to develop new products and services that address critical needs of clients. There is also demand due to regulatory compliance, which has pushed regulated financial institutions to increase automation and adopt new analytical tools that can include the use of AI and machine learning.
Financial services sector participants increasingly find it necessary to keep up with their competitors’ adoption of AI and machine learning, even if only for reputational reasons (hype). In many cases, these factors may also drive ‘arms races’ in which the ones with better technology win.
Challenges to Successful Deployments
While we have seen many use cases of AI and machine, we are at the early stages of adoption with less than 10% penetration in the financial services sector. Technical and business challenges for sustainable implementations include:
- Algorithm Improvements – Some applications of finance require robust algorithms that are tuned for data with low signal to noise ratios as seen in algorithmic trading. Overfitting is a common pitfall in these regimes and practitioners need to deploy a robust technical and business framework to avoid misleading results.
- Alternative Data Sources – Alpha is getting harder to find, as traditional markets have become highly efficient. As a result, spending on alternative data sources by financial institutions is expected to approach $2B in 2020, up 60% from just a year ago. However, due diligence in selecting the right dataset to ensure compliance and privacy have become front and center issues.
- Explainability – Complex algorithms, along with the quality of ingested datasets, create opacity/explainability questions regarding AI and machine learning techniques. The industry as a whole, and the financial services sector more importantly, need to provide a transparent process for decisions based on AI and machine learning technologies. Investors are demanding “black boxes” be replaced with “glass boxes.”
- Ethics – A robust ethics framework including transparency, accountability, and fairness is required for a trustworthy AI and machine learning in the financial services. Guidelines and regulations are forthcoming, and financial institutions would need to adopt the appropriate ethics framework to strengthen the governance of AI applications and data-analytics-driven decisions.
- Talent – There is an acute shortage of technologists with AI and machine learning expertise in all sectors, including financial services. Besides, the tech-savvy talent pool is not trained as financial analysts and are not necessarily CFA or CAIA charterholders. It is critical for these professionals to train in specific areas such as fiduciary responsibilities with clarity on Client First, ALWAYS. Data science is a team sport with players like data scientists and “translators” that are the primarily financial professionals with data science skills. Other new titles and functions include data engineer, data czar! data hunter, and so on.
- Regulations – One challenge for the development and use of AI in financial services is to understand the extent to which regulation applies. Some regulators are technology agnostic: They regulate to improve the industry and protect consumers irrespective of technology or means of delivery. So, pre-existing regulation applies to AI as it does to any other product, service, or technology used in a regulated business. In practice, this means that firms must carefully consider what technology they are using and how, and then assess how regulations will apply to it.
- Process– The ultimate benefit of AI and machine learning will be realized once they are integrated within the business processes. Therefore, the term “Systematizing” for scale is often used by financial managers, referring to technical and business processes to quickly deploy AI systems in their organization. AI and machine learning in the future will become an “enterprise application.” Financial services firms, depending on their size and applications, need to develop a framework and “repeatable” processes to deal with data due diligence, due quality, models, regimes of operation, etc.
Financial Data Professional Charter
To address the above challenges encountered by the financial services sector, there is a need for educational programs that would help financial professionals acquire the knowledge required to utilize these new technologies. Financial professionals with AI and data science skills, can bridge the gap between data scientists and business managers as well as explaining the results to stakeholders, and helping in the deployment of new products and services.
Financial Data Professional Institute (FDPI) provides a curriculum and a level of discipline and professionalism coincident with the alternative data explosion. It sets the foundation for financial analysts to develop specific skills such as working with data sets, managing a team of data scientists, communicating results to various stakeholders. In addition to helping financial professionals develop new skill sets, the FDPI curriculum elevates the role of privacy and ethics in the development and execution of any financial data science projects.
Skilled and certified Financial Data Professional (FDP) is a practitioner that possesses the same rigorous standards set by other global designations such as CAIA, CFA, and FRM. Organizations that employ them, realize critical value propositions including:
- Competitive advantage – The future direction of finance is driven by data-informed decisions. FDP charterholders, enable deployment of AI and machine learning tools and methods resulting in finding investment ideas faster than the competition. The predictive capabilities brought about by machine learning techniques can provide a significant edge over the traditional analysis techniques.
- Trustworthy AI and machine learning tools and outcomes – FDP charterholders help create the frameworks governing ethics, transparency, and risk management that are essential for the success of AI and machine learning in financial services firms. FDP charterholders inherently understand the fiduciary responsibility to investor and ensure the interest of clients are front and center.
- Multidisciplinary teams – Sustainable operational success of AI and machine learning projects, require teams that include data scientists, data engineers, and translators. FDP charterholders provide the “translator” function bridging the gap between scientists, technical staff and the subject matter expertise of financial professionals. In this role, they ensure that outcomes are within the objective framework of the organization.
- Regulation and compliance – A combination of existing regulations and digital transformation of financial services creates a complex compliance environment. Compliance officers and regulators realize that AI and machine learning tools can increase compliance while reducing the cost of compliance. FDP charterholders can help organizations deploy their AI and machine learning assets to develop better compliance processes.
- Standards and best practices – AI and data science field is changing rapidly with new techniques, methods and improved algorithms. FDP charterholders are the natural gate keepers for introduction of new AI technologies VS the requirements of their financial firms.
Summary
AI and machine learning technologies are at the core of digital transformation in the financial services sector. The value of AI and machine learning has been demonstrated across several use cases with varying levels of maturity. Drivers for the adoption include both supply side and demand side. However, there are challenges to successful deployment of AI and machine learning amongst which talent is a common thread. There is a significant shortage of skilled analysts to carry out data science projects. Financial Data Professional charterholders address a critical segment of the talent pool and provide the skill set to understand the core concepts; to explain the results to stakeholders; and to help in deployment of products and services based on AI and machine learning.
References
Mehrzad Mahdavi and Hossein Kazemi (2020), It’s All About Data: How to Make Good Decisions in a World Awash with Information. The Journal of Financial Data Science Spring 2020, jfds.2020.1.025; DOI
Financial Stability Board. (2017). Artificial Intelligence and Machine Learning, Market developments and financial stability implications
Provost, F. and T. Fawcett. (2013). Data Science for Business. Sebastopol, CA: O’Reilly Media Inc., Chapter 9 & 10.
Chi Chan et al (2019). Artificial Intelligence Applications in Financial Services.