For all the talk about artificial intelligence (AI) in financial circles at present – it seems everything is “AI-powered” – it turns out there’s a lot less genuine AI in place than we might have imagined.
The March 2020 Unlocking AI Playbook: FI Edition, a PYMNTS and Brighterion collaboration, explains that while the use of AI solutions by banks and financial institutions (FIs) skyrocketed 70 percent in a single year (2018-2019), less than 10 percent of all banks say they use AI today.
“AI’s real-world usage may appear limited compared to the fanfare surrounding it today, but our research aims to accurately depict its adoption, so we precisely define AI,” the report states. “Systems fitting our definition must have current business applications and be able to work with and learn from dynamic data sets in real time, and these capabilities must be able to associate with specific entities within a system.”
Fair enough. But what’s holding things up?
Fear of implementation cost and complexity are major deterrents to adoption, with 82 percent of banks turning to existing technology, such as a business rule management system (BRMS), to mimic the data insights promised by true AI. Use of BRMS is up almost 23 percent since 2018, which shows an eagerness to upgrade technology – even though such systems aren’t true AI.
The Semantics of AI
The very definitions of AI are challenged and clarified in the March Unlocking AI Playbook: FI Edition, so that clever (but not “intelligent”) systems don’t continue to be categorized as AI.
“The term ‘AI’ is applied to an abundance of technology products and services today, but genuine AI is in fact relatively rare in the systems available to and employed by FIs,” the report notes. “PYMNTS aims to provide a clearer picture of genuine AI’s fit into this landscape by defining six major computational systems today’s banks use.”
Those six computational systems include the previously mentioned BRMS, along with data mining, case-based reasoning, fuzzy logic, deep learning/neural networks and genuine adaptive AI.
True AI, as defined by PYMNTS, entails the massive computational power needed to analyze impossibly large datasets, combined with the “smart agent” system structure that allows AI to learn and “know” the legitimate behavior of every card and account in its purview.
Big Banks Get AI
Not surprisingly, larger FIs have the resources to afford true AI and the scale to get performance results from a well-configured system. “PYMNTS’ analysis found that larger banks were more likely to employ multiple systems, including AI,” the report states. “FIs with assets over $100 billion use four systems on average, compared to the 2.1 systems leveraged by those with less than $5 billion. Larger banks represent the lion’s share of AI users, too, with 87.5 percent of FIs with assets over $100 billion and 42.9 percent of those with assets between $25 billion and $100 billion utilizing the technology.
“This contrasts with the 3 percent of banks with between $5 billion and $25 billion in assets that use AI,” the report continued, adding that, “Not a single FI possessing less than $5 billion in assets reports using the technology.”
While cost is absolutely a consideration, what’s clear is that not all computer systems claiming AI powers actually possess them. That’s one of the major takeaways from the March 2020 Unlocking AI Playbook: FI Edition, which contains a wealth of tables illustrating its findings.