Understanding the application of AI in auto finance

Starting from a working definition that AI is the science of using machines to do something that would be considered intelligent if done by a human, a workshop at the Asset Finance Connect Summer Conference, sponsored by Alfa, brought together vehicle manufacturers, finance houses, technology specialists and start-ups to explore the practical implications of the AI revolution.

In broadbrush terms, AI divides between narrow models trained to perform a single task, such as a credit check, and generative AI, like the ground-breaking ChatGPT, which is based on trillions of data points and capable of answering questions in an uncannily human fashion.

Both have valuable roles to play in the automotive sales and finance chain, roles that will increase in importance as vehicle manufacturers adopt agency distribution models in which they take over sales responsibility from their former franchised dealers. Converting a greater proportion of visitors to their websites into e-signatures on a virtual dotted line is going to be vital.

With the first contact between manufacturer and customer likely to be on an online platform, AI has the power to track every step of the buying journey. It will be able to establish the page-by-page routes and interactions that typically lead to sales. And it will identify the early warning signs of a drop-off where a customer is on the brink of losing interest in the process, and respond with nudges in the form of messages and offers to keep the customer engaged and direct them along more fruitful routes.

AI-based technologies will be able to respond 24/7 in real time and with a human tone to customer questions, and when a sales team is online it will help them to single out the handful of prime customers who have pursued browsing behaviour likely to lead to a sale, from the thousands of people who may be visiting the website. These hot prospects can then be targeted with live, human support to clinch the sale. In many respects, this is simply a 21st century version of experienced department store staff being able to spot definite buyers within the crowd on their shop floor.

Buyers’ routes to vendors’ websites, the pages they visit, how long they spend on different pages, the questions they ask and answer, and the data they are prepared to share all provide valuable insights into their intentions to buy, creating patterns from which AI can learn. For example, a customer who has pre-qualified their car finance is four times more likely to buy a car.

Moreover, AI has the power to identify non-linear relationships in datasets, honing the accuracy of both its predictive models of how a customer will behave and its prescriptive models that recommend the best steps to secure a sale. The more data, the more nuanced the insights.

While manufacturers have previously been able to conduct this analysis, AI’s transformative power is its ability to do it in real time. So, if a customer sets out on a browsing route that suggests high interest in purchasing, AI can reinforce this commitment with messaging that maintains the momentum of the process, each nudge honed from learnt experiences of previous successful transactions.

This includes using generative AI to deliver real-time sales advice and scripts to human agents. They can filter and edit these automatically generated answers (which are constantly enhanced and improved in a cycle of machine learning) to provide customers with a personalised sales approach.

“Your job will not be taken by an algorithm, but by a person with an algorithm,” said one delegate.

And yet to use AI merely to digitise former analogue sales processes would deny vendors massive opportunities to interact with customers in the way that they would prefer, which creates a challenge of knowing when and how a salesperson should contact a prospect, alongside potential GDPR privacy complications. For example, inviting a customer to a test drive too early in the process can be a turn-off – customers are fully aware that they can contact a dealer for a test drive; the reason they are shopping online is precisely because they want to avoid the pressure of a face-to-face sale. So, a phone call from a sales executive may not be an appropriate approach, either from a sales or privacy perspective.

Moreover, from a compliance perspective, alarm bells are starting to ring if vendors allow AI to provide unsupervised answers about finance and insurance, with the spectre of miss-selling lurking in the background. Consumer Duty already demands that finance houses and insurers can explain their decisions, both positive and negative, so it will be imperative that there are no unwitting biases written into AI’s decision-making processes.

Finally, despite the science fiction-type advances of AI, it is important to remember that fewer than 10% of car sales are currently completed online. The vast majority of transactions have a break in the chain between online research and in-person acquisition, which means AI needs to find a way to separate customers in a showroom who have spent considerable time beforehand informing themselves about a specific car, from those who are there to browse.

Similarly, having an identifier that linked a showroom sale back to the customer’s browsing history would massively enrich the dataset of the sales journey from which AI could learn.

The technology also has a growing role to play in supporting finance and fleet management companies with the in-life operational management of vehicles. Connected cars are generating a tsunami of data about how a vehicle is being driven, its mileage, and its mechanical performance. Being able to warn a customer that they are likely to exceed the mileage parameters of their lease agreement, for example, could nip potential end-of-contract problems in the bud, while sensors that detect engine temperature and vibrations are helping fleets to adopt predictive maintenance schedules, replacing parts before they fail and minimising unscheduled downtime, a key requirement for last mile and mission critical businesses.

And looking ahead, AI is starting to analyse the correlation between the charging profile of electric vehicles (how often they charge, the number of charges/discharges, and the speed of charging) and their battery health, which is set to become as important a determinant of their residual values as a full service history.

The possibilities presented by AI to improve and enhance every element of a car’s life cycle seem limitless, and the finance sector will need to be crystal clear about the challenges it wants AI to solve and fine-tune the questions it asks ever-growing datasets to exploit the full potential of the technology.