AI in banking: Finally getting past the demo stage

The artificial intelligence conversation in banking reached an inflection point last week at the Temenos Regional Forum Americas 2025. Walking through the convention center at Turnberry Golf and Resort in Aventura, Florida, you could sense something had shifted. 

Gone were the breathless predictions and theoretical use cases that dominated the media and conferences just a year ago. Instead, executives shared real deployment stories, complete with surprising successes, unexpected challenges, and the kind of practical wisdom that comes from actually trying to make this technology work at scale.

But perhaps most telling was the underlying current of urgency that ran through every conversation. In an industry traditionally known for deliberate, cautious decision-making, AI implementation has become a race where speed of execution matters as much as technical sophistication.

The adoption reality: Moving faster than expected

The numbers emerging from the latest banking research tell a compelling story about where the industry actually stands. “Today, only 11% of banks have implemented generative AI, and 43% are in the process, which means that more than half have already moved forward,” said Temenos CMO Isabelle Guis.

This isn’t the slow, methodical adoption curve that banking typically follows for new technologies. It’s something closer to the urgent scramble that happens when an industry realizes competitive advantage is at stake. “81% of the leaders also agree that banks that do not adopt artificial intelligence will fall behind the competition,” Guis noted.

The urgency is real, but so is the caution that comes with operating in one of the world’s most regulated industries. “86% of the leaders interviewed are concerned with data protection, and more than half are worried about legal risk and AI accuracy.” 

The tension between speed and safety has created a new dynamic in banking technology adoption. Institutions that can move quickly while maintaining compliance standards are gaining significant advantages over competitors still stuck in analysis paralysis.

Where AI is actually working

The most revealing conversations at the forum centered on where AI is delivering measurable results today — and they weren’t the customer-facing applications that typically grab headlines. Instead, the success stories focused on back-office operations where AI could tackle the industry’s most compliance-heavy, labor-intensive processes.

Temenos Chief Product Officer Sai Rangachari, shared real deployment results: “We just launched our sanction screening agent, and one of the top tier one banks, they send about 20% of their transactions through this agent, and we’re seeing 2% false positive relative to the 6% to 8% that we see without the agent.”

That improvement represents more than incremental efficiency gains. “That’s a huge impact, especially at their scale. That could be 50 to 100 people,” Rangachari noted. In an environment where compliance teams are under pressure to keep pace with increasing volumes , those kinds of operational improvements free up skilled staff to focus on the highest risk, most complex cases, and to ultimately safeguard the bank and its customers.

But achieving these results required a different approach than the “spray and pray” AI strategies that dominated early adoption efforts. “Instead of releasing 10 or 15 new AI products, let’s take the ones that mean the most to our customers and build those out,” explained Barb Morgan, Temenos’ Chief Product and Technology Officer, reflecting a broader shift toward focused implementation.

The execution challenge: Why speed matters

Throughout the forum, a clear pattern emerged: the institutions succeeding with AI weren’t necessarily the ones with the biggest budgets or the flashiest demos. They were the ones that had figured out how to execute quickly while managing the complex organizational dynamics that AI implementation requires.

Morgan, who has assembled a new management team specifically chosen for their ability to move fast — many with direct experience working for banks and fintechs on product and technology — captured this challenge perfectly: “I think there’s a unique breed of people who like to drive transformation, and I happen to be one of those. And so I pulled on a lot of my leaders that have done transformation with me in the past.”

It’s about bringing in people who understand the other side of banking technology from having lived it. When your team has sat on the other side of vendor presentations and managed technology implementations at actual banks, they bring a different perspective to product development and customer engagement.

This focus on execution velocity and customer empathy reflects a broader recognition that AI adoption is an organizational capability challenge. The institutions that can move from pilot to production quickly are the ones positioning themselves for sustained competitive advantage.

The cultural reality no one talks about

Perhaps the most honest moment of the forum came when Morgan discussed something most vendors prefer to gloss over: the human element of AI adoption. “I was talking with one of our US banks last week, and he said I underestimated the amount of cultural change that’s necessary, because so many people are afraid of AI. They think it’s going to take my job away, versus thinking of it as augmenting their job.”

This cultural resistance comes from the need to rejigger how financial institutions operate. It isn’t an optimization or pure cost cutting exercise. The research data supports this nuanced view: “90% of banks are not expecting Gen AI to replace the processes and their people. They expect Gen AI to augment them.”

The banks succeeding with AI implementation have figured out this cultural piece through careful, measured rollouts. As one Temenos exec told about a client’s agentic AI deployment: “They started with 5% and then they said, okay, we’re gonna flow 10% of our traffic through, then we’re gonna flow 20% of our traffic through. And it wasn’t because they didn’t trust the technology. They were getting the rest of the organization comfortable.”

The governance gap slowing everyone down

One of the forum’s most impactful findings centered on organizational structure, and it explains why many AI initiatives stall despite executive enthusiasm. “Only 42% of banks have a dedicated AI team in charge of implementation and governance,” Guis reported. Successful rollouts require people who can navigate the complex intersection of technology capability, regulatory requirements, and business needs. The institutions moving faster have figured out this organizational piece.

“The board of directors is participating in 75% of AI projects, and 99% of them are actively involved,” the research showed. This isn’t just box-checking — it’s recognition that AI transformation requires the same level of strategic attention as any other major infrastructure investment.

But board involvement alone isn’t enough. The most successful implementations combine C-suite commitment with teams that understand both the technology possibilities and the practical constraints of banking operations.

The explainability imperative: Why black boxes don’t work so well in banking

In an industry where every decision can face regulatory scrutiny, the “black box” problem of AI is a compliance requirement that’s shaping how the technology gets deployed.

“We operate in a very highly regulated environment, so having an explainable AI with strong governance… that agent that I talked about that passed regulatory muster and accepted by the regulator — that was important for us,” Rangachari emphasized.

This explainability requirement is changing how AI products get developed for banking. “Everything we do, the output is there. It’s readable in English. The code is delivered. So you have a full audit trail,” he noted.

The challenge is building AI systems that can evolve with changing compliance standards. Banks that get this right early are positioning themselves for sustained advantage as regulatory frameworks mature.

The implementation prediction that changes everything

Looking ahead, Rangachari made a prediction about the reshaping of the banking technology landscape: “If we can half the time and half the cost of [digital transformation] in a tier one or tier two bank… then I think the whole game is going to change for the large banks. They will actually be able to progressively modernize finally.”

The most successful AI deployments at the forum shared common characteristics: focused use cases, explainable systems, gradual rollouts, and teams that understood both technology possibilities and banking realities.

The broader implications

Walking out of the Turnberry convention center, the message was clear: we’re past the point of debating whether AI belongs in banking. The question now is which institutions will master the art of implementation, balancing speed with safety, innovation with regulation, and technological capability with organizational readiness.

The 11% of banks that have implemented AI are still early adopters, and they’re also the testing ground for what becomes the industry standard. And based on what we heard in Aventura, that standard is coming faster than many institutions realize.