The adoption of artificial intelligence in the investment management space is happening at pace. But firms need to balance risk and reward, and be aware of impending regulations, writes Rhodri Preece of the CFA Institute.
The use of artificial intelligence (AI) in investment management is increasing rapidly. Tools such as natural language processing, image and voice recognition software, and machine learning algorithms are being adopted across the investment management spectrum: in portfolio management, risk management, trading and investment advice.
This is causing profound disruption to traditional business models and investment processes that have shaped the industry for so long, and there is increasing demand to adopt these technologies at greater scale. As a result, the global market for the use of AI in asset management is expected to be worth $13.4bn by 2027, with a predicted compound annual growth rate of 37% between 2020 and 2027.
The demand for greater AI adoption is coming from the next generation of asset managers and investors alike. Our recent report assessed the future of skills in the investment industry and found that 64% of professionals are currently pursuing, or plan to pursue, skills development in AI and machine learning, rising to 71% among young finance professionals. This indicates how much young millennials and Gen Z-ers anticipate AI will change the investing landscape, and the scope for growth in this field.
We also calculate that for every professional currently pursuing AI expertise, there are three more who intend to do so. This not only indicates the breadth of the future potential talent pool, but also underscores the importance for firms to develop a strategic framework for AI encompassing culture, governance and teamwork to best capitalise on these skills.
Risk, reward and regulation
AI adoption has the potential to deliver improvements across the investment value chain, but also carries several risks that have the potential to undermine the trust and confidence of investors if not kept in check. These risks include how data is sourced and processed by AI tools where issues of data integrity and possible biases exist, and how transparency and accountability can be enforced when there is limited ability to observe or explain the decision-making process of an AI application to clients or supervisors.
Businesses should also be aware of several plans afoot from authorities in major jurisdictions to establish frameworks for the use of AI in financial services. For example, the EU’s General Data Protection Regulation, which provided a comprehensive framework for data management, has been followed up by the proposal of an Artificial Intelligence Act.
But regulation alone cannot address all AI risks. Because AI algorithms do not intrinsically possess fundamental ethical attributes of honesty, fairness, loyalty, and respect for others, they must be imbued as design principles by the professionals responsible for their development and use. A focus on ethics and professional standards is therefore central to the development of responsible, client-centric AI practices.
Four principles for ethical AI
We have developed a decision-making framework for businesses to guide the ethical design, responsible development and deployment of AI tools. It is guided by the following four principles, which must be considered at each step of the AI workflow.
- Data integrity: datasets are subject to biases, so they need to be cleansed to ensure they are fit for purpose in an AI model. Devolving decision-making from a human to a machine does not eliminate bias, but professionals working with AI tools must be cognisant of this fact and take appropriate steps to mitigate these sources of potential bias in AI decision-making processes. Sampling techniques should be representative, with fair and accurate data labels, and adhere to local data privacy laws.
- Accuracy over complexity: AI applications need to be reliable and perform as intended in a live environment. Models are often overcome by excessive complexity at the expense of accuracy and understanding, meaning AI processes cannot be effectively challenged.
- Transparency and explainability: investment professionals working with data scientists who build AI models should evaluate potential trade-offs between model accuracy and interpretability, so they can effectively explain outcomes to external audiences. Investment professionals need to be able to understand the key features of any AI program that informs investment decisions, including the parameters used in models and quantitative research that are incorporated into investment recommendations.
- Accountability: given the complexity of AI projects and the need for business functions to collaborate in their development, leadership accountability, ethical culture, and collective ownership of IT deployment are essential for success. This begins with senior leadership establishing a strategic vision and ethical culture for AI development within their organisation.
These principles create the basis for a decision-making framework to guide the ethical design of AI tools. The onus is on investment professionals – those with a duty to clients – to provide ethical leadership and ensure such considerations are passed on to the teams developing AI-driven solutions.
For more guidance on the development of ethical AI practices in investment management, read the report: Ethics and Artificial Intelligence in Investment Management: A Framework for Professionals.
Rhodri Preece is a chartered financial analyst (CFA) and senior head of research at the CFA Institute.
