AI driven solutions are becoming a competitive differentiator for banks and other financial services (FIs)—delivering a highly-personalised customer experience, improving decision-making and boosting operational efficiency. Yet, many of them remain in an experimental phase and will need to accelerate actual AI deployment. Otherwise, they risk being left behind by digitally native players.
AI is rapidly transforming every aspect of the financial world. This transformation has accelerated recently, thanks to evolutionary AI—a new breed of technologies that allow AI to automatically design itself with little need for explicit programming by humans. As it gradually becomes mainstream, evolutionary AI’s capability to innovatively create complex AI models and to optimise decisions considering multiple scenarios is set to reimagine the financial sector. It will enable every player in this field to spot novel strategies that would never have been identified by human data scientists and in turn allow companies to take full advantage of today’s massive data sets.
Fundamentals of evolutionary AI
Emerging technologies that enable AI algorithms to design themselves are allowing organisations to transcend human limitations. Evolutionary AI operates iteratively. Firstly, it randomly generates a set of potential solutions to form an initial population and assigns a score to each solution based on how well it performs relative to other solutions. In the second round, it retains the solutions that performed best, perhaps only 5 percent of the total and recombines their components, sometimes mutating them to create a new population. This new population is then tested and the process begins again. Over multiple generations, the appropriate components of the more successful solutions become increasingly prevalent in the population and eventually a solution is discovered that yields the best outcomes.
Identifying deployment benefits to maximise returns
Compared to human design, evolutionary AI can be deployed far more quickly, avoids biases and preconceptions, and typically performs better. Furthermore, the chosen model will evolve and improve over time based on new data. The technology can be applied in a wide variety of areas at FIs. Some examples include designing quantitative trading strategies to maximise returns while minimising risk and loan underwriting. Rather than relying on human analysis, evolutionary AI solutions can quickly analyse all the combinations of relevant variables to create models that more accurately assess the risk of default by a potential borrower.
In order to reap the benefits of the technology, FIs should focus on the following:
Create and maintain responsible AI applications: Behave in ways that make customers and employees comfortable. This means not making decisions that are unethical or exhibit bias. Companies need to monitor them to ensure they continue to act appropriately as they learn and evolve.
Craft business-driven AI strategies: AI should be viewed through a business lens, rather than as a technology issue. Having AI projects managed by cross-functional teams with business executives in the lead is a good place to start. Companies also need to look across their organisations to identify opportunities to generate concrete business value from AI—not only in reduced costs but also in boosting revenues by delivering enhanced customer experiences and through improved decision-making.
Enhance data management: AI applications depend on access to timely and accurate data, which is a challenge for many FIs that have fragmented data architectures with multiple legacy systems. Companies need to identify which types of data are required for each AI project and ensure they can be captured in an appropriate format.
Adopt an experimental mindset: AI projects need to be rolled out quickly, while at the same time be rigorously measured, so failures are terminated promptly while successes are moved into production.
As AI applications increasingly design and test themselves, the pace of innovation and the accuracy of predictions will vastly improve. It is inevitable that FIs will soon consider it irresponsible to make important business decisions without first consulting with an AI system. Robots will handle routine tasks while flagging exceptional cases for review and resolution by employees. Employees will spend their time on more complex decisions and sensitive interactions with customers, such as resolving complaints or providing sophisticated financial advice. In short, humans and AI robots will be working side by side, delivering more value in combination than either could on its own.