CIOs in financial services embrace gen AI — but with caution

Despite the myriad temptations to rush in, KeyBank, Gen Re, and Genpact business tech leaders set the path for ethical, efficient, and secure generative AI usage.

Highly regulated, customer-centric, and dependent on layers of human involvement and manual processes, financial services are ripe for automation through artificial intelligence (AI). Those same characteristics, however, reveal the risks AI pose to this sector. So business technology leaders in financial services are carefully navigating a path toward AI. But as they reveal, it’s a route they must navigate with caution.

AI dominates discourse and headlines, but financial services technology leaders know there’s a lot of substance amid the noise. “There are many tools that go into a hype cycle, and then you come out of that cycle a little dismayed, but there’s a difference here,” says Dominic Cugini, chief transformation officer at KeyBank. “We’re seeing how fast this technology is maturing, so it’ll have a very different hype cycle.”

AI is not the future of financial services — it’s the present. Genpact, a major business and technology services company that assists banks such as JP Morgan and Goldman Sachs, is already utilizing AI. “It’s really good at summarising, filling in blanks, and connecting dots, so generative AI is fit for purpose,” says Brian Baral, global head of risk at Genpact. “We’ve been able to leapfrog and do in months what had taken three years, but the data is key. Banks have to get ready to take the step forward.”

Conscious of the recent history of disruption to financial services, the sector’s technology leaders are already looking for opportunities in AI. “Generative AI is starting off a new age of exploration in IT,” says Frank Schmidt, CTO at insurance firm Gen Re. Cugini at KeyBank agrees, and adds that the exploration has to include a cross-functional team from all areas of the business, not just IT. “We also pulled in some experts from Microsoft and Google to really understand what AI means to our sector.” Schmidt sees AI as having potential in process automation, particularly underwriting submissions. “AI will play a role in this workflow and classifying information,” he says.

CIO Tiago Azevedo of Boston-based low-code development platform OutSystems agrees. “In order to get meaningful productivity from AI, we need to rethink workflows,” he says. “And I expect AI will become composable so it can play different roles in the organization.” For this to succeed, financial services organizations will need processes that are far more modular.

Just as the adoption of AI needs all parts of the business to be involved, so too does the ethical mandate of financial services organizations and their use of generative AI. “We’ve started an ethical AI committee that consists of the legal, compliance, technology, and cybersecurity teams,” says Cugini.

Setting out the rules

Such is the transformational potential of AI that financial services organizations will need to construct rules on its usage that reflect the regulatory environment, expectations of customers, and geographic and cultural differences. “We’ve issued guardrails for the use of AI,” says Schmidt at Gen Re. “One of these was human-centric, that every employee is responsible for their work, which held true before AI, and still does.” This is a timely reminder that although there’s much hype surrounding AI, the cultural norms of an organization and greater society must always be respected.

“We put policies in place along with legal on the use of AI including use cases,” says Cugini. “We communicated to the whole company that we’re not locking down and ignoring this technology, but creating a counsel for the business to look at how to bring AI into KeyBank in a responsible manner. So over the last several months, we’ve been taking a disciplined and educated understanding of large language models and generative AI. This means now we’re taking an intentional approach, and know we’ll always have a human involved for the foreseeable future.”

As financial services adopt AI, the technology roles that have protected organizations and enabled change will once again come to the fore. Cugini believes that business analysts will become even more important to CIOs as they step into their traditional role to reduce the gap between engineers and the business. “We’ll need that technology rigor, and you want people that talk to the business,” he says.

AI in a box

CIOs are under pressure to deliver productivity improvements and reduce costs in financial services. As a result, many CEOs have high expectations of AI and its ability to transform their businesses. However, developing and deploying an LLM is costly. For this reason, out-of-the-box LLMs such as Bedrock from Amazon Web Services (AWS) and those from Microsoft may give CIOs the speed to market they desire.

One such organization that’s taken this route is Genpact. The New York-headquartered business is using AWS Bedrock as the LLM foundations for RiskCanvas, the fraud prevention and reporting technology service Genpact provides to financial services providers, including Apex Fintech Solutions, a provider of clearing services for e-commerce businesses.

“With AWS’ strict data security, it prevents any data going outside of AWS, that ensures the models are clean and don’t access the whole internet like OpenAI does,” says Brian Baral, Genpact’s global head of risk management. This is being used to automate suspicious activity reports (SAR), which financial service providers have to produce if they identify transactions in breach of sanctions, for example. “In the US alone, there are four million SARs filed with the government a year,” he adds. “They take time to complete, typically two to three hours as it has to be written in a specific format, and it’s a sensitive artefact for the regulator, and if you get it wrong, you’re in trouble.”

Generative AI is now automating SAR generation for the financial services customers of Genpact, and Cugini at KeyBank is also considering generative AI for SAR automation. “AI is able to take all the information and write a case for inspection and review, so it benefits the clients,” he says of how better fraud prevention helps customers. Baral adds: “There’s an efficiency and effectiveness as you want your analysts to be writing these near-perfect every time, but there’s a lot of variability in humans. Generative AI gives a perfect answer, but there’s a human in the loop.”

This AI in financial services debate often returns to the need for human involvement as the sector’s business technology leaders are acutely aware of the need to retain consumer trust in their organizations.

“Technology enablement is about increasing accuracy and quality control,” Cugini says. “Humans get fatigued. If you have large teams that require a lot of governance, AI can reduce that variability and meet our regulatory, client, and stakeholder needs.”

Baral sees the same benefit: “Organizations will save money, and analysts can be analysts again, free from grunt work so they can fight financial crime, and their jobs will be more enjoyable.”

So as a model, none of these three financial services business tech leaders believes generative AI will actually face their customers.

AI threats

Generative AI won’t only increase the productivity of financial services workers, but also give cybercriminals a powerful new tool with which to attack banks and insurance firms. Over a third of business technology leaders indicated in a recent Nash Squared Digital Leadership survey that they’re concerned with protecting data privacy as a result of AI. “With the advent of generative AI, we’re seeing more attack vectors such as synthetic identities being created,” Baral says. For financial service providers, this increases the need for watertight customer identification systems. “There are continually more challenges and more bad actors, and they’ll always find ways to attack.”

Customers naturally expect financial service providers to be careful with their money, and that demand has led to a culture of meticulous analysis and adoption of technology, especially now with AI.