5 takeaways from Evident’s Responsible AI Report
/As AI adoption grows, banks are staffing up to manage the risks. AI has a multitude of exciting use cases in banking, but institutions are also preparing for its challenges, including data and privacy hurdles, bias, hallucinations, and compliance concerns.
“Past investments provide a starting foundation for banks to manage emerging risks specific to AI, their ability to address the dynamic, black-box, and/or data-intensive nature of AI systems has been tested. Operating in a constantly evolving regulatory environment only adds complexity (and urgency),” wrote the authors of Evident Insights’ Responsible AI Report that was released earlier this month.
Evident’s Responsible AI Report highlights how major banks are hiring to address AI risks and the steps they are taking to encourage responsible adoption. Here are five takeaways.
1. A growing number of roles at banks involve responsible AI.
Of the 50 large banks assessed by Evident, 41 had roles dedicated to responsible AI in 2024, up from 31 banks in 2023. Overall, the number of employees whose responsibilities include responsible AI is still a relatively small group, with just 249 employees that fall into this category. On a geographic basis, U.S. institutions are leading and have the greatest number of these types of roles. JPMorgan Chase was identified as a leader, with more than 20 roles, more than three times the number found at the average bank.
2. Banking leaders’ remits are now more likely to include responsible AI.
Year over year, the report found a significant increase in executive leadership positions that included responsible AI oversight. Two thirds of the banks assessed — 33 out of 50 — had a senior responsible AI leader in 2024, which Evident defined as a VP-level or above executive who oversees responsible AI activities, drives awareness and builds internal expertise. This number more than doubled from 2023, when only 15 banks had such a role. Responsibilities include risk assessment; education and training; stakeholder engagement; policy development; and thought leadership.
“The great value of having a responsible AI function, of having a data ethics function, is we can look over all of these risks and consider them in a holistic way and put the end impact to the customer, to our clients, to our employees, to our other stakeholders, at the real heart of what we’re doing,” Luke Vilain, AI governance lead at UBS, told Evident.
3. AI training for employees is on the rise.
Of the 50 banks evaluated, 31 included some form of AI training for employees.
“Access to our Generative AI tool is contingent upon team members completing required training and securing the appropriate permissions for its intended use,” Ozge Yeloglu, vice president of advanced analytics and AI at CIBC, told Evident.
4. Banks are mapping responsible AI principles and controls.
Eighteen banks published their set of responsible AI principles, a three-fold increase from the year-earlier period. Publishing responsible AI principles helps banks build trust among
customers, partners, investors and regulators, but banks still face the challenge of “translating those lofty principles (sometimes derided as ‘platitudes’) into pragmatic AI controls that provide structured, actionable policies and safeguards,” the report said.
5. Banks are investing in AI research to drive adoption.
The rollout of responsible AI research reports saw a 136% increase among the banks covered by the study. For example, research on explainability served as a building block for control processes CommBank put in place for its model developers, Dan Jermyn, the bank’s chief decision scientist said.
Dan Latimore, chief research officer at The FR, argues governance approaches to responsible AI are essential, but banks also need to back them up with concrete action.
“Like any governance or compliance function, having a formal approach to responsible AI is a necessary but not sufficient condition for managing the risks associated with implementing AI. If they’re to realize the benefits of AI without undue downside, banks must follow through with robust execution plans so their strategic intent is reflected in operational reality,” he said.