Last week, we enjoyed the opportunity to engage with FinTech Circle and its ecosystem regarding the complex issues that arise at the intersection of artificial intelligence (AI), data, and scenario analysis. The FinTech Circle ecosystem consists of some of the most forward-looking companies on the planet when it comes to financial technology. These are early adopters and pioneers. (You can access the webinar recording HERE).
So we were curious to know how many participants currently use AI to conduct scenario analysis. Evangelists for artificial intelligence often make it sound as if everyone is currently using AI for any and every data challenge. The reality is more nuanced. As the recent AI Journal survey indicated, around 60% of their survey respondents indicated that as of summer 2020, their companies were already using AI. A further 52% were planning deployments and/or expanded usage of AI utilities internally.
AI can be used for a broad range of purposes, from chat bots and marketing to risk pricing and risk management. The AI Journal did not ask survey participants how AI was being deployed internally. Their focus instead was to identify whether and to what extent the pandemic may have impacted AI adoption.
AI and Scenario Analysis: A Natural Experimentation Juncture
Our webinar focused on a subset of risk management: scenario analysis. Going in to the webinar, the expectation was that AI deployment regarding scenario analysis would be fairly high for two reasons.
First, scenario analysis is a great place for financial firms to experiment. Scenario analysis outputs rarely, if ever, generate direct changes to product prices, risk profiles, or regulatory capital. Scenario outputs are evaluated by humans.
Second, scenario analysis incorporates a large number of sometime disparate data sets and multiple complex calculations that often include causal chains in order to identify second-order, third order, and sometimes indirect effects.
Scenario analysis paired with AI – or at the very least deep learning – can thus deliver real efficiency gains as well as superior insight generation. Risk managers can experiment with AI systems while generating relatively little operational risk to the firm. It is a natural entry point for experimentation.
So we kicked off our webinar by asking how many participants currently use AI in their scenario analysis. The answer surprised us: only 12% currently use it.
Why would the current usage rate be so low? We did not have time during the webinar to explore this question. We wanted to cover other topics regarding data.
Our working hypothesis is that risk managers tend to be risk-averse. Risk managers working within regulated financial institutions….and fintech firms serving as contractors to regulated financial institutions…must adjust their pace at the innovation frontier to meet regulatory expectations.
Financial firms and their regulators will want a high degree of comfort that the automated analysis provided by AI is both objective and transparent. They will take great care to ensure both the models themselves and the data used to train the models do not inadvertently import subtle or hidden bias that can skew outcomes.
We are fairly certain the answer will be very different when we ask the question next year.
Despite the hype, it is still very early days for AI adoption. The pandemic intensifies the need to move methodically and thoughtfully along the innovation frontier. The high amount of interest in the data/AI/scenario analysis topic from last week’s webinar tells us that firms are actively exploring potential use cases for AI within the risk management process. We look forward to helping our colleagues explore that frontier by contributing our alternative data to the risk management landscape.
BCMstrategy, Inc. helps risk managers and other knowledge professionals manage their exposure to public policy risk by measuring the language of the public policy process. Our alternative data, generated by 9+ layers of patented analytical automation, is the next frontier for risk management and scenario analysis. For more information on how our data can help you and your team manage exposure to public policy risk, please visit us at: www.policyscope.io