Banks Must Sell Public on AI Benefits or Risk Losing Them

Artificial intelligence and automation are starting to help financial institutions’ bottom lines. However, pushing AI usage beyond early adopters to a greater share of the industry will require surmounting a vital obstacle: consumers’ broad lack of trust in AI systems.

Part of the challenge is terminology. AI is often written about informally, and words such as “robot” conjure images of job-stealing automation and social media programs designed to hijack human sentiment. But questions are raised in more thoughtful venues as well. In one commentary on the future of AI in financial services, the Brookings Institution think tank spoke of the need to be open about the limitations of AI. The article said that this was “crucial, as miscalculating the true potential of AI algorithms can create wicked consumer protection problems, especially when financial products and services are involved.”

The stakes are significant for financial players, but also for consumers who value what’s already been offered.

How AI has Assisted Financial Institutions Thus Far

Several major banks have reported generating significant savings and positive returns on their investments in these technologies. JPMorgan Chase’s cost to serve customers of all types has decreased 15% over the last five years, while Bank of America’s spending has fallen by as much as $1 billion per quarter, helping it invest in upgrades to branch and ATM networks.

“The example of Erica and other financial institution chatbots highlights the importance of driving customer adoption of AI and automation capabilities.”

Thanks for much of this impact goes to consumers’ own growing use of AI and automation tools to complete banking tasks and interact with financial institutions. Bank of America has cited consumers’ increased use of its Erica chatbot as a key cost saver. Increasingly Erica has automated interactions that previously took place through costly call centers. Bank of America recently reported that more than 7 million of its customers have used the chatbot to conduct more than 50 million interactions.

The example of Erica and other financial institution chatbots highlights the importance of driving customer adoption of AI and automation capabilities. Chatbots, which typically sit within bank mobile apps or websites, still have considerable room to grow. Bank of America previously reported more than 25 million customers are using its mobile app altogether — so most haven’t interacted with Erica yet.

Another benefit for consumers is seen in credit: AI allows lenders such as Marcus by Goldman Sachs and Lending Club to deliver credit decisions within seconds.

Consumers worry about financial fraud, and the use of AI to reduce fraud through pattern detection is becoming widespread. Rather than isolating each transaction as a single event, financial institutions use AI to examine the flow or continuum of transactions. Wells Fargo, JPMorgan Chase and Ally Financial are all employing automated monitoring systems that apply this cohesive view of transactions.

Banks such as BNY Mellon are employing robotic process automation (RPA) to reduce costs and improve operational efficiency. Forrester predicts the RPA market will reach $2.9 billion by 2021. BNY Mellon reported an 88% improvement in transaction time in its trade-settlement system, just one backend operation increasing its use of automation. BB&T and SunTrust — soon to combine as Truist — are both RPA users, which SunTrust has deployed in its payments operations.

Read More:

Building Trust Proves Essential for AI Usage to Grow

Given the significant returns financial institutions have started seeing on their AI investments, and benefits already seen for early adopters among consumers, it would benefit them greatly to accelerate AI adoption among consumers as quickly as possible.

“Multiple studies indicate that most consumers still distrust AI for completing banking tasks.”

Yet multiple studies indicate that most consumers still distrust AI for completing banking tasks. Only 20% of respondents in one Pegasystems survey said they’d trust AI to recommend banking products or to make investment suggestions. Another survey from the company found 68% of respondents would trust a human more than AI to make loan decisions.

Financial institutions would gain much potential savings and streamlining, judging by advances already made, if they would overcome this resistance. The savings could be reinvested in new endeavors to drive greater business growth.

Where can efforts to increase trust begin? With where trust least exists. Pegasystems’ survey data shows some of the underlying reasons consumers don’t trust AI:

53% suspect AI system decision making can be biased.
Another 53% don’t think AI can learn and improve over time.
37% don’t think AI could ever understand their preferences as well as another person could.

Financial institutions will need to prove that AI systems can in fact make unbiased decisions and effectively learn consumers’ preferences.

However, the most crucial factor in changing consumers’ perceptions of AI will be in significantly improving their lives. More than two-thirds — 68% — of consumers say they’d be more willing to trust AI if these systems saved them time or money, or otherwise assisted them.

In this respect, Bank of America taken steps. It has been executing an effective strategy by integrating “Proactive Insights” that warn customers if they’re overspending or have upcoming bills due. This turns the Erica chatbot into an everyday assistant, helping customers navigate financial obstacles and avoid pitfalls. The feature has doubled customer engagement with Erica, according to Bank of America.

Read More:

Acceptance Depends on Cooperation with Vendors and Regulators

At the same time, financial institutions will need to work proactively with technology vendors and regulators to cultivate trust and transparency. Financial institutions aren’t alone in this. An increasing number of companies across industries seek to implement explainable, transparent AI models that can describe how they came to their decisions. This will be especially crucial in highly regulated industries like financial services.

The Federal Reserve has already published guidance stating that banks should be able to validate and assess the decision-making of their analytics tools. Financial institutions should be at the forefront of this trend, collaborating with their vendors to ensure AI-driven decisions are clearly spelled out for all stakeholders.

Regulators also have to be part of the conversation. Financial institutions and AI providers should work with regulatory authorities to establish necessary benchmarks for transparency — as well as modifying existing regulations to account for the growing importance of AI in decision making.

If financial institutions and their vendors do not appear to the public to be cooperating with regulators on how AI solutions are assessed and deployed, it could further damage public opinion about AI’s accuracy and trustworthiness, throwing more hurdles in the way of further AI adoption.