A Google engineer has put artificial intelligence (AI) under the spotlight in recent weeks with claims that a computer chatbot had become “sentient”, capable of reasoning and thinking at the level of a human child. While the company quickly set the record straight with an official statement to the contrary, this episode is the latest example of how far AI computing has come – and how poorly it is understood.
Chatbots, or text generation programmes, are certainly among the most visible and easily understandable applications of AI. But these – and the debate around computing sentience – are little more than a sideshow to the true potential of this game-changing technology.
At its core, AI allows more complex processes and decisions to be outsourced from human to machine, enabling industries to operate at a new level of efficiency and – crucially – far greater speed.
Nowhere is this more important than in the financial services industry, where billions of dollars are at risk every second. Augmenting the unseen function of risk management may not sound as exciting as designing a language programme that can hold a real conversation, but it may well be AI’s biggest contribution to society so far.
Calculating value
The global financial crisis, as well as a host of subsequent misconduct and scandals, radically changed banking’s regulatory environment. Financial firms have paid fines totalling hundreds of billions of dollars since the collapse of Lehman Brothers. And one of the industry’s responses has been to hire more compliance staff. By 2019, the Economist reckoned these employees accounted for 10% or more of headcount at large banks, about double the pre-crisis proportion. Compliance costs increased accordingly.
Largely, those investments paid off as intended, cutting the amount of fines banks shouldered for anti-money laundering (AML) and data privacy lapses in half for 2021. Still, the structural shift also drove up costs, taking capital away from other uses, contributing to inflation and reducing returns – as well as risks – for shareholders.
To thrive in this environment, financial institutions must remain focused on improving operational efficiency without compromising on risk management.
Enter AI. Three in four financial institutions expect it to become an essential part of their business in the near term, according to a World Economic Forum study. And the business consultancy McKinsey has highlighted more than 25 use cases for AI in financial institutions. The firm estimates that the technology could deliver an additional US$1 trillion of value for global banks each year.
This is still only the start of things to come. While about 85% of financial services companies may have started using AI in some form, only one in four have adopted AI across four or more business domains. The often-cited barriers are a lack of access to well-organised data and a perceived immaturity of AI applications.
These hurdles are quickly coming down as the technology develops and innovators widen access to AI infrastructure, platforms and data sets.
As a result, AI is being used right now to help banks manage a broad spectrum of risks. The technology streamlines everything from know-your-client (KYC) checks to identifying fraudulent behaviour among millions of transactions a day.
In this industry, analysing vast amounts of data quickly and without emotion is key. Just as AI is already crunching millions of online ad purchases globally without human oversight, why not put that same power to use for something more economically productive than determining who sees an ad for a new brand of toothpaste?
A real-time challenge
Complying with regulations for basic KYC requirements, AML, counter-terrorist financing (CTF), or sanctions screening is a data challenge for any financial institution. And for multinational firms, it’s multiplied by the number of jurisdictions they cover. Risk mitigation requires monitoring government watchlists and cross-referencing consumer-credit or business-rating databases, which may have structured data to work with, but not in a consistent way.
Such challenges slow down humans and render basic computer automation ineffective. But AI can be trained to focus on significance rather than structure. AI can already verify documents through computer vision and Optical Character Recognition (OCR). It can process complex documents and data sources and arrive at a lending decision, much like an online ad buy for example, in the time it takes to read this sentence.
Modern and robust risk management today means being able to rapidly ingest and analyse reams of unstructured data. This includes traditional and social media sources, which are out in the open but not always accessible for analysis, that can tip off fraud or other misconduct. Criminals often leave unwitting patterns that AI can be trained to spot as it combs online resources and databases for information in real time. AI’s outputs create insights that help people make informed decisions.
AI and its subsets like Machine Learning (ML) are still emerging technologies. But the largest cloud providers, including Amazon Web Services, Google Cloud Platform and Microsoft Azure are continuing to improve the breadth and quality of AI services on their platforms. As this trend continues, it will become easier for financial institutions of all sizes to start adopting AI solutions across their businesses.
I expect AI investment and adoption within the financial services industry to rise as benefits increase and costs and barriers fall. Broad adoption across business lines should feed through into more efficient operations, emotionless analysis and – all things being equal – superior shareholder returns.
It’s not sentient, but the AI we don’t see (or chat with) is already augmenting risk mitigation and transforming the financial industry as we know it.
