Artificial intelligence was one of the hottest topics in banking in 2017.
Nearly every major financial institution has made significant announcements about new AI applications this year. This is not hype: Few industries are more susceptible to AI disruption than banking.
That's because banking is a process- and data-driven industry, and AI tools are perfectly suited to leverage data to optimize processes, which in turn allows banks to personalize customer experiences, automate operations, improve risk management, and much more.
However, overall adoption of AI across the industry is still relatively low, and there are few mature case studies to help bank executives decide how to harness AI’s disruptive potential.
A survey of bank executives published earlier this year by Narrative Science and the National Institute of Business found that fewer than a third of traditional financial institutions have launched AI projects.
Moreover, many banks are not familiar with AI due to the difficulty of acquiring scarce and highly expensive AI talent: According to a Paysa report released in November, financial services companies allocated more than $82 million to AI development and research this year, with almost half of that being invested by the three biggest banks: JPMorgan Chase, Capital One, and Wells Fargo.
It makes sense to start with the backend
To get started on their AI journey, banks need a strategy that addresses the wide variety of AI use cases across industries, the many challenges of AI implementation, and each bank's unique digital strategy.
For most banks, the use of AI is Backend As the technology behind these applications becomes more mature, they become more evolved.
Many people in banking and other industries are fascinated by customer-facing AI applications based on computer vision and natural language processing, such as chatbots.
But it is the rapid advances in different types of machine learning technologies suited to extracting insights from the vast amounts of data banks collect that are driving innovation in customer-facing applications.
Additionally, back-end AI applications tend to pose less risk than front-end applications that directly impact customer experience.
Still, banks have a surprising number of use cases to consider for machine learning tools. Prioritizing these use cases comes down to two key factors: the bank's business priorities and the data available to feed into the algorithms.
Where does it hurt?
Every bank has different pain points and business goals that AI can help solve.
For example, a financial institution with many high-priority commercial clients might consider using AI to help automate accounts receivable, as Bank of America Merrill Lynch is doing in partnership with HighRadius for a program called Intelligent Receivables.
Alternatively, banks looking to reduce compliance costs can use machine learning to automate the collection and analysis of data to detect money laundering, as HSBC has done with startup Aysadi.
The options seem nearly endless: AI can help banks better determine credit risk to help expand their lending business, develop more sophisticated investment models to power their asset management departments, or help their small business clients automate expense reporting and fraud detection.
Ten years from now, many banks will likely be using machine learning in all of these applications, but developing an AI strategy must prioritize AI solutions that help solve banks’ biggest and most pressing problems.
The recipe starts with raw data
However, identifying your business needs is only one part of the puzzle in leveraging AI: without the right data, machine learning and deep learning tools can't provide meaningful insights.
Despite its potential, AI technology is only as useful as the data used to train it. Banks have a lot of data, but the right data isn't always easily accessible.
Many banks still collect and store vast amounts of data manually. Even data that is collected electronically is often siloed in legacy systems, making it difficult to extract and analyze. As part of their omnichannel and customer-centric initiatives, more and more financial institutions are taking on the daunting challenge of exposing vast data sets through cloud migration, system upgrades, APIs, and middleware.
Banks that are further along on this path will be better equipped to take advantage of AI, with these efforts laying the foundation for gaining new insights and operational efficiencies through machine learning.
In this sense, a bank's AI strategy is an extension of its digital transformation strategy. Many AI projects typically start with a long, laborious period of finding, extracting, and cleaning the datasets needed for algorithms to draw meaningful conclusions. If the data you need is already cleaned and easy to extract, these tasks can be greatly simplified, giving staff more time to focus on fine-tuning their analysis instead of poring over large amounts of data files.
Efforts must be unified
Therefore, banks must align their business, digital transformation, and AI strategies to lay the foundation for leveraging AI. Once business needs are properly identified and prioritized, and digital transformation efforts are in place to support those needs by making data more accessible, banks will be ready to use AI technologies to tackle their most pressing business problems.
