Machine learning (ML) has become a core component of financial institutions’ business strategy, accelerated by Covid-19 and driven by use cases across risk, trading and investment. Rising numbers of data scientists embedded in the business lead the way, working with increasingly sophisticated tools and moving into deep learning. The challenge to successful ML projects is data quality and availability.
The level of adoption of ML and AI, its use cases, challenges and potential, are reported by Refinitiv in a 2020 report, The Rise of the Data Scientist – Machine learning models for the future. The report is based on a survey including over 400 telephone interviews with data scientists, quants, technology and data decision-makers at sell-side and buy-side firms across the Americas, EMEA and Asia-Pacific.
Key findings suggest firms are scaling AI/ML capability across multiple business areas, progressing from proofs of concept to developing mature ML models, and deploying more data scientists, who are themselves evolving from a supportive function to drive strategy. Data strategy is more important than technology strategy, Natural Language Processing (NLP) is unlocking unstructured data, and many firms are turning to deep learning, although it does have implications for hardware, cost optimisation and AI/ML explainability.
Considering Covid-19, the report notes market volatility left 72% of firms saying their ML models were negatively impacted and 12% declaring their models obsolete. Lack of agility to respond to rapid change was the main problem, leading firms to update and build new models that are more dynamic. The pandemic also accelerated implementation of ML projects for decision making, although investment levels differed.
David Craig, CEO at Refinitiv, comments in the report: “Covid-19 is the arch accelerator. In under six months we have experienced a level of technological change in financial markets that would otherwise have taken a decade to play out. This is a revolutionary moment in financial technology and one that is quickly widening the gap between the haves and have nots. Those businesses that can best harness data and emerging data science techniques – and deploy them at scale – are stretching their advantage.”
Amanda West, global head of Refinitiv Labs, adds: “Tackling the fallout from Covid-19 has only made implementing data-driven strategies, which give current and accurate insights we can trust, even more important. If firms are going to genuinely benefit from the speed, agility and value of an ‘AI-first’ vision, they need high-quality, trustworthy data that can be easily accessed, ingested and manipulated for the variety of financial use cases they are progressing.”
Looking forward, financial institutions intend to use ML to extract more value from data, extract better quality information and stay ahead of competition over the next year or two. From a technology perspective, the next target is ML operations (MLOps) that will scale AI/ML for the enterprise and help data scientists drive change by operationalising AI/ML models and replacing manual steps for data preparation and model evaluation with automatic processes. Financial data scientists and ML engineers will drive change, and 2021 could see the emergence of a new role, the MLOps engineer.