Managing the risks of AI in finance: data governance and management are crucial

Regulators are aware of the disruptive impacts and security threats posed by weak data governance (DG) and data management (DM) practices within the investment industry. Many investment firms are failing to develop comprehensive DG and DM frameworks that may keep pace with their ambitious plans to leverage latest technologies akin to machine learning and artificial intelligence (AI). The industry must define the legal and ethical use of information and AI tools. A multidisciplinary dialogue between regulators and the financial industry at national and international levels is required to determine legal and ethical standards.

Steps to data efficiency and effectiveness

First, create diverse and concrete Goals short, medium and long run. Next, you create a primary Timeline which reflects the hassle in manageable Phases: for instance, just a few small pilot initiatives to start out with. Without clear goals and deadlines, you will soon be back to your day by day work, with the outdated refrain from the business side: “The data management and administration thing is IT’s job, right?”

It is amazingly necessary to start out with a transparent vision that features milestones with set dates. You can take into consideration How to fulfill deadlines along the way in which. When defining and establishing the DG and DM processes, it is best to take into consideration Future security Systems, processes and results. Are a particular data definition, approach and decision-making policy linked to an overall business strategy? Do you may have management commitment, team and customer involvement?

As I identified in my first post on this topic, the organizations which have essentially the most success with their DG and DM initiatives are people who T-shaped team approach. That is, a business-focused, interdisciplinary partnership supported by a technology team that features data scientists. Setting realistic expectations and demonstrating successes shall be essential disciplines, as DG and DM frameworks can’t be created overnight.

Why are DG and DM necessary in financial services?

It is more necessary than ever for investment professionals to remodel data into complete, accurate, forward-looking and actionable insights.

Ultimately, information asymmetry is a big source of profit for financial services. In many cases, AI-powered pattern recognition capabilities make it possible to extract insights from esoteric data. In the past, data was mainly structured and quantitative. Today, well-developed natural language processing (NLP) models also process descriptive or alphanumeric data. Data and analytics are also necessary to make sure regulatory compliance within the financial industry, one of the crucial heavily regulated business sectors on this planet.

Ultimately, irrespective of how sophisticated your data and AI models are, “human sense” can significantly affect users’ perceptions of the usefulness of the information and models, whatever the objective results actually observed. The usefulness of the information and techniques that usually are not based on “human-understandable” logic is unlikely to be properly assessed by users and management teams. When intelligent people see correlations without cause-effect relationships identified as patterns by AI-based models, they consider the outcomes to be biased and avoid making incorrect decisions based on the final result.

Data and AI-driven initiatives in financial services

As financial services grow to be more data and AI driven, many plans, projects and even problems arise. This is where DG and DM come into play.

Defining the issue and the target is crucial, as not all problems are suitable for AI approaches. In addition, the shortage of transparency, interpretability and accountability could lead on to potential procyclicality and systemic risks in financial markets. It could also result in incompatibilities with existing financial supervision, internal governance and control, and risk management frameworks, laws and regulations, and policy-making that promote financial stability, market integrity and healthy competition while protecting customers of economic services which have traditionally been based on technology-neutral approaches.

Investment professionals often make decisions based on data that isn’t available to the model and don’t actually have a sixth sense based on their knowledge and experience. Therefore, strong feature capture in AI modeling and human-in-the-loop design, i.e. human oversight from product design and throughout the lifecycle of the information and AI products as a safeguard, is important.

Financial service providers and regulators should have the technical ability to operate, audit and, if essential, intervene in data and AI-based systems. Human involvement is important for explainability, interpretability, auditability, traceability and repeatability.

The growing risks

To capitalize on the opportunities and mitigate the risks posed by increasing volumes and diverse data types, in addition to newly available AI-powered data analytics and visualization, organizations must develop their DG and DM frameworks and concentrate on improving controls and the legal and ethical use of information and AI-powered tools.

The use of massive data and AI techniques isn’t only reserved for larger asset managers, banks and brokerage firms which have the capability and resources to take a position heavily in vast amounts of information and complex technologies. In fact, smaller firms have access to a limited number of information aggregators and distributors that supply data access at reasonable prices, in addition to just a few dominant cloud service providers that make common AI models accessible at low price.

As with traditional non-AI algo trading and portfolio management models, the usage of the identical data and similar AI models by many financial services providers could potentially trigger herding behavior and one-way markets, which in turn may pose risks to the liquidity and stability of the economic system, especially in times of stress.

Worse still, the dynamic adaptability of self-learning (e.g. reinforcement learning) AI models can discover interdependencies and adapt to the behavior and actions of other market participants. This can result in unintended collusive outcomes without human intervention and the user may even concentrate on it. Lack of convergence also increases the danger of illegal and unethical trading and banking practices. Using similar or similar data and AI models increases the associated risks, as AI models can learn and dynamically adapt to changing conditions completely autonomously.

Given the difficulties in explaining and reproducing the decision-making mechanism of AI models using big data, it’s difficult to mitigate these risks. Given today’s complexity and interconnectedness between regions and asset classes and even between captured aspects/characteristics, the usage of big data and AI requires special care and a focus. DG and DM frameworks shall be an integral a part of this.

The limited transparency, explainability, interpretability, auditability, traceability and repeatability of massive data and AI-based models are key policy issues that remain to be resolved. The lack of those is inconsistent with existing laws and regulations, internal governance and risk management and control frameworks of economic services providers. It limits users’ ability to grasp how their models interact with markets and contributes to potential market shocks. It can amplify systemic risks related to procyclicality, convergence, reduced liquidity and increased market volatility attributable to simultaneous buying and selling in large volumes, especially when most market participants use standardized data and third-party AI models.

Importantly, users’ inability to adapt their strategies in stressful situations can result in a much worse situation during times of acute stress and exacerbate flash crash events.

Big data-powered AI in financial services is a technology that augments human capabilities. We live in countries where the rule of law prevails, and only humans can take protective measures, make decisions, and take responsibility for the outcomes.

References