Let’s face it – early AI rollouts have been anything but smooth. Standout examples include McDonald’s attempt at using AI to take drive-through orders, which caused customers to plead with the AI to quit adding Chicken McNuggets to their orders.
Or, Amazon’s AI-powered HR recruitment tool, which favored male job candidates and eliminated female candidates from consideration – a bias Amazon blamed on data and model training challenges.
Now, let’s pretend for a moment that these failures occurred in banking. While consumers might be overjoyed at the prospect of an AI-powered ATM spitting out extra $20 bills, the loss to banks of such a mistake multiplied across hundreds or even thousands of ATMs would result in massive losses.
Imagine that instead of eliminating female candidates from job consideration, AI instead eliminated female applicants from loan or credit consideration. Mistakes in banking go far beyond funny foibles. Errors on this scale would likely lead to legal action, reputational harm, and loss of consumer trust.
According to Evident’s AI Index, AI references in banking annual reports, press releases and LinkedIn posts are up 59% year-over-year, highlighting both the impact of AI and banking’s desire to reap its potentially astounding rewards. In the scramble to lead or not fall behind, the activity is frenetic and the pace is blistering.
Even as banks race toward windfall returns, they must also take a careful and measured approach to implementation and mature their AI strategies. Failure to do so will inevitably result in a loss of ROI and could lead to even greater losses that won't be measured in dollars alone.
Adoption frenzy and expected returns
In 2024, banks rushed to adopt and implement AI, fueled mainly by the hype around GenAI.While harnessing GenAI’s potential is still a top priority going into 2025, leading banks are homing in on AI implementations that will help revolutionize the business, drive productivity gains, create efficiency, and deliver business transformation.
According to Evident, the top 10 banks are moving at a pace double the rate of the average index bank, and they’re focusing heavily on the revenue gains they hope AI will deliver. JPMorgan Chase alone expects more than $1 billion in earnings from its use of AI.
Are these expectations realistic given the current state of AI maturity? For all but the largest banks, the answer is either “no” or “not for some time.” While we may see the largest banks leading the pack, the realization of those billions in expected ROI is still further afield.
While the focus may be on the endgame and reaping rewards, many banks must revisit the basics if they hope to deliver AI transformation in alignment with strategic priorities.
Not just hammers, not just nails
If all you have is a hammer, everything looks like a nail. And the hammer-and-nail approach simply won’t work to effectively implement AI. Unfortunately, 2024 was littered with these whack-a-mole attempts at AI implementation, and many failed. The foundation of a successful AI strategy lies in data and governance.
Many banks still struggle to turn their AI vision into reality because their data is trapped in silos and scattered across fragmented platforms. The problem is further compounded as data continues to grow exponentially from an overwhelming number of sources.
Banks that have yet to defragment and modernize their tech stacks, clean up and manage their data, and integrate data to drive decisions will likely face challenges in their AI implementations.
Effective governance is also key to driving AI strategy. Without an effective governance framework, banks find it difficult to identify the best technologies and use cases to pursue. The result is often a chaotic adoption that doesn’t always pan out. Instead, banks must develop a clear strategic mandate that's aligned with board priorities and desired outcomes.
The governance framework then becomes a tool that clearly defines criteria for technology and use case selection to support those priorities and outcomes.
This approach builds a strong foundation upon which banks can grow, innovate, and deliver high-impact initiatives.
A tightrope walk through global regulation
Global AI regulation is complex and constantly changing, with political shifts potentially complicating progress. This regulatory maze presents challenges now and could slow the path to greater AI success in the future.
Most regions and countries around the world acknowledge that AI requires regulation and oversight to ensure appropriate application, explainability, and fairness in its use. When applied thoughtfully and collaboratively, regulation can be a tool to drive a more unified approach – streamlining global business, promoting inclusion and access, and supporting innovation within clear boundaries.
Regions worldwide are adopting different approaches. Regardless of where a bank is based, doing business globally will require banks to navigate a complex maze of AI regulations. This will add to banks’ already massive regulatory compliance load.
The European Union has made significant strides with the EU AI Act, which bans AI applications deemed too risky or unethical, provides legal requirements for high-risk AI applications, and leaves applications that don’t fall into the prohibited or high-risk categories largely unregulated.
In April 2024, UK regulators published their approach to AI regulation in financial services, noting the broad adoption of AI already in existence. While rules may still be issued, many believe that effective regulation of AI already exists within the current regulatory framework. For now, it appears the UK plans to continue with a sector-based approach to regulating AI – as has typically been done with technology – with an eye on adapting as technological innovation accelerates in speed, complexity, and impact.
Asia-Pacific – comprised of eight different jurisdictions – is taking a different approach to AI regulation. Across these jurisdictions, several areas of alignment provide a basis for regulation in specific countries. These include data privacy standards, consensus on standardization and guidance rather than binding regulation and penalties, the need for a compliance framework, and the desire for a somewhat uniform approach aligned with EU regulations to support cross-border business operations.
In the U.S., AI regulation remains in flux, with ongoing debates around the right balance between fostering innovation and ensuring accountability. While some policymakers advocate for voluntary self-governance and reliance on existing frameworks, others push for more comprehensive oversight to address transparency, fairness, and inclusivity in AI application.
This uncertainty poses challenges for banks operating globally, as they must balance the demands of a less prescriptive domestic regulatory environment with more structured international standards.
If this inferred approach holds, banks must establish clear governance to self-police, maintain vigilance and transparency in their use of AI, and ensure protective safeguards are in place. Beyond regulatory oversight, banks must also answer to the court of public opinion, with consumers acting as de facto regulators who can – and will – take their business elsewhere.
Opportunities abound
Though the road to AI success is filled with challenges, there are also many exciting opportunities to explore and promising developments on the horizon. One positive note about the successes and failures of AI in banking is that all of the experimentation will ultimately drive innovation. Any fan of the “fail fast and fail forward” approach surely feels excitement about the lessons learned and the potential AI holds for banks and their customers.
Banks that find the magic formula where innovation, customer experience, and profits align will truly reap the rewards of their hard work. While some may worry that AI will take over jobs, creativity, and human ingenuity are still essential to driving the art of the possible.