The opportunities and challenges presented by AI have been hitting the headlines lately. From the corridors of academia to the boardrooms of Silicon Valley, the promises and pitfalls of this new technology are being analysed and agonised over. In compliance teams, AI has been talked about as the silver bullet for a multitude of challenges. But is it?
Napier’s founder and chairman Julian Dixon thinks there are still some misunderstandings about how, where and why to use it so that it’s cost effective and efficient.
“Often, deploying AI can be like investing in top-of-the-range cycling gear but forgetting to buy the bike,” he says. “There are some basics that should be in place well before a bespoke AI solution is considered. AI can be difficult to get right – it should never be a knee jerk solution to delivering an effective compliance solution. The industry is on a journey towards integrating it with human intelligence so that we can deploy both most efficiently. But the future for AI is exciting. It’s amazing technology.”
As a former economic buyer for a bank, Dixon is first and foremost a pragmatist, with an eye firmly on the ROI of any investment in technology.
“AI can be an expensive upfront cost, but the running costs can also be high. The models can get stale, become corrupted or need retuning, all of which add to the Total Cost of Ownership (TCO),” he says. “You have to factor into the equation the issue that it needs to be explainable to meet regulatory requirements. Ours always is, but that’s not the case with every provider. You could end up needing to employ additional expensive specialist resources to be able to tick that box – some industry players have not made life easy for non-data scientists. It’s important at the outset to consider user and regulator costs as well as the ongoing investment in a state-of-the-art platform.”
So what’s the clever first step towards reducing false positives?
“A decade ago, financial institutions would fix high incidences of false positives by employing teams of expensive consultants. Now, the solution is seen to be automating by using AI. But first, I would counsel any organisation to revisit and improve their rules to sieve the data more effectively. You have to get the rules tailored to the needs of your organisation. That’s why we at Napier introduced sandbox testing. We were the first in the world in AML technology to have inbuilt sandbox testing using copies of real data. Our approach makes the system both bespoke and user-friendly, as well as ensuring a problem-free go live. You don’t need AI to set those tolerances. Start with a smart base to which you can add automation. We integrate AI automation into our state-of-the-art Continuum platform and our point solutions, thus seamlessly introducing the efficiencies that automation can deliver. Continuum gives users control over the rules and the algorithms, which solves the vast majority of compliance challenges. The platform effectively gives you the best of both worlds, which is why it’s won so many awards. But if there still remains a job too large to be done efficiently by human intervention, you can layer additional AI over your existing base system to further automate out false positives.”
Is there any situation in which bespoke AI is the immediate go-to solution?
“It may be the way to go when you have a specific problem to solve to meet regulatory requirements,” says Dixon. “Fines can hurt organisations’ reputations as well as their bottom lines, and bringing in AI to fix things quickly can make sense in these instances. Even at this relatively early stage in its usage, there are times when the value AI can deliver is greater than the cost.”
What needs to happen before AI becomes the ubiquitous solution of the future?
“Ai is extraordinary technology, developed by some of the finest maths brains on the planet. Raw state data is incredibly complex, and I think the sector is still wrestling with some of the pitfalls that come with such a high level of complexity. Here at Napier, our data scientists are constantly working on a plethora of AI driven solutions, with the end goal of making AI better, cheaper and easier to deploy.” says Dixon. “We want to make out-of-the-box AI solutions as widely available as possible, making the undeniable benefits of AI easily accessible in a cost effective and understandable manner.”
Improve your compliance processes with an award-winning solution