The roles and capabilities of artificial intelligence (AI) in the finance sector are constantly growing. How can the industry best use those capabilities?
‘Artificial intelligence’ is one of those terms, like ‘the Cloud’, ‘Blockchain’, or ‘Gamification’, that describes a specific and sometimes useful tool, but which can be bandied around as a buzz word so often it can be hard to determine the actual use for it.
However, once you get past the hype it’s impossible to deny the technology is already having a big impact on the finance sector. In 2018 Fannie Mae’s Mortgage Lender Sentiment Survey, ‘How Will Artificial Intelligence Shape Mortgage Lending?’, determined that 63% of mortgage lenders were familiar with AI technology, 27% were already using AI tools for their mortgage business, and 58% expected to adopt AI solutions within two years.
Since then, the case for AI in finance has grown stronger, as a shifting economic environment, accelerated by the Covid-19 crisis, has shifted the emphasis on lending away from credit scores and towards analysing applicants’ entire digital footprint, from social media account use to internet browsing, to geolocation and smartphone data.
Irish finance firm Bluestone has introduced a sales tool that uses AI to offer a real-time, automated approvals process to motor dealers at over 600 dealerships across the country.
Using algorithms to determine a customer’s credit suitability in seconds, the tool replaces manual underwriting and enables dealers to complete the entire sales transaction with a single meeting.
These new tools can streamline the sales and approvals process while ensuring motor finance is approached prudently. Bluestone has reported that the technology allowed it to give an immediate decision on 30% of applications for motor finance from the outset, increasing that to 40% by 2019.
AI hasn’t just seen use at the front end of motor finance transactions. Flexys is a specialist debt collection software supplier providing design, research and development, delivery and support.
The system Flexys provides is a cloud-native, end-to-end debt collection system that has used AI and machine-learning tools to dramatically improve the speed, performance and cost-effectiveness of debt recovery systems.
“We use clustering and classification tools that help us to identify customer preferences and behaviours which we then use to improve communications and prioritise agent intervention for those most in need,” explains Jon Hickman, chief executive of Flexys.
However, Hickman argues the potential applications for AI are far wider than that.
“Within the debt management aspect of motor finance, there is a range of applications for AI including customer engagement and channel models, and propensity (to pay or engage) models,” Hickman says.
“So, for example, using AI to suggest the right message, the right channel and the right time for an individual customer engagement. In addition, AI could help define the best payment method and amount. All these elements help to deliver successful results.”
The primary benefits of machine learning are borne out of the fact programs can process far greater volumes of information than their human counterparts, and can see the patterns that emerge from the data.
“Machine learning can uncover relationships and patterns that we wouldn’t otherwise be able to find with simple analytics. Once trained, the software can recognise those patterns when shown new data so it will continue to optimise and stay up to date,” says Hickman.
Of course, recent history has demonstrated the importance of being able to adapt to changing trends in the data.
“A topical example would be the Covid-19 pandemic which brought many new debtors with unseen behaviours and preferences to which AI solutions are better suited for quick iteration and learning,” Hickman says. “Organisations using AI could adapt quickly and scale their operation, avoiding delays and making sure customers can access the help they need.”
Power and responsibility
Of course, it’s natural that as financiers gain access to ever greater technology and capabilities, the responsibilities and regulatory environment will become more complex. Firstly, before the AI can do anything with the data gathered, its quality must be confirmed.
“The main practical challenge involves handling data,” Hickman tells us. “Before we can start any learning, we must ensure the data is as good quality, understandable and error-free as possible.”
Once past that hurdle, the challenge becomes making use of that data while preserving the privacy of customers.
“Most security concerns around machine learning centre on privacy and autonomous decision making. Although these concerns are legitimate, Flexys implements systems and controls that mean AI is used in a manner that both protects privacy and protects against the potentially negative effects of unmoderated AI decision making,” Hickman points out. “For example, we strictly anonymise learning data so that our AI systems never see any information that identifies any individual.”
One point that Hickman is very keen to make is that machine learning and AI solutions can augment, but are no substitute for, human engagement.
“AI should only ever be used to support or augment human decision making. At no stage should AI systems make definitive or irreversible decisions that might adversely affect a person’s future,” Hickman insists. “AI should empower its users to make more informed and effective decisions but not attempt to replace existing knowledge and experience.”
To meet the challenges and responsibilities AI and machine learning tools present financiers with, it’s essential to make data a business priority and ensure that the necessary expertise and domain knowledge are in play.
“It is good practice to start small and build on successes safely and incrementally,” Hickman advises. “A competent supplier will test new ideas, new algorithms and new features in a low-risk, controlled and measurable way.”
At the same time, these tools are not magical cure-alls.
As Hickman says, “It’s also important to be realistic, to accept setbacks as part of the process and be prepared to move on if a particular AI method isn’t producing desired results; sometimes all that is required is a reframing of the problem the AI is trying to solve.”
But while the challenges and limitations facing AI in the finance sector are real, there is no doubt that as confidence and expertise grow, AI will take a more prominent role in optimising a range of issues relating to motor finance.
“As well as specific results, the use of AI can have a knock-on effect on wider efficiency,” Hickman tells us. “For example, in debt management, when AI is used to personalise digital journeys, this makes digital self-service even more accessible and increases take-up and completion rates. This, in turn, will free up contact centre agents to manage the edge cases or exceptions.”
Ultimately, machine learning and AI tools always come back to enabling the user journey.