Why exploiting AI opportunities depends on good data

Capturing accurate, reliable data is the essential building block for the technological revolution about to engulf asset finance

The intoxicating possibilities for the asset finance industry of artificial intelligence and machine learning all depend on the challenging but mundane task of capturing, storing and accessing clean data, according to a panel of experts at the AFC Summer Conference.

Instant risk assessments and credit decisions, enhanced customer services, improved in-life asset management and valuations, and more efficient accounting processes all depend on the foundations of robust, reliable data.

“The future is obviously data-led,” said Ravneet Shah, Chief Technology Officer, Allica Bank.

“When we are ‘solutionising’ problems and building digital tools, we take data as a first consideration: how we are going to store the data, how we are going to generate analytics from it? We need good use of data to automate the decision-making as a finance body.”

This is easier said than done, depending on the task that the data underpins, cautioned Catriona Powell, Head of Data Insight & Automation, Novuna Business Finance. Identity resolution in the assessment of credit risk, for example, can be challenging.

“Can we collate the right data to get sufficiently comfortable that we’ve identified the right entity to make our decision?” she asked.

“There are particular challenges around unlimited [unlimited companies and sole traders], some of them SMEs or even individuals, depending on what data you’ve got, but having confidence that you’ve got the right person is absolutely key to your AML and KYC checks. What we’ve found particularly powerful is to bring in data from a number of sources and then apply our own algorithms in order to make some of those determinations over whether we’ve got the right entity.”

This still leaves questions surrounding the number of data sources required to reach a decision, and even whether data-driven processes can deliver the answers required, despite the tsunami of data flooding into asset management firms.

As with any IT investment, the solution depends on having a crystal-clear vision of the business objective, said Christian Roelofs, CEO, Finativ.

“I see all the time people making technology decisions based on an output, but the output is not the value – it’s the action that the output drives that’s the value,” he said. “So, the question is how do we make sure that we get the value out of technology decisions? Business expectations need to be set around the end result.”

He suggested that if a company cannot codify its underwriting appetite into a systematic order, then AI will not be able to help the decision-making process. It is easier, he added, to think of AI as offering an assisted decision, rather than an automated decision, with the technology eliminating a lot of manual work and identifying underwriting decisions that are simple to approve or decline.

“If you think about the volume of decisions that an underwriter would have to take, these models would reduce the number of those decisions, which brings it back to what we’re probably trying to solve – the speed of underwriting,” said Roelofs. “And, therefore, you might want to look at making automated decisions, or just simplifying the journey for the underwriter to reach a faster decision.”

If it is not possible to codify underwriting decisions, lenders can follow one of two approaches, explained Andy Trimmer, Head of Technology, Simply Finance.

“You can either try and dig even further for more data, or you can assess your risk appetite and say, actually, maybe I’m looking at five things and I should be looking at three. And then it becomes that little bit easier.”

Key to this are data sources that can ‘talk’ to each other, rather than sit in silos where employees have to flip through different windows searching for the information they need. This is easier for start-ups and companies devising their IT infrastructure from scratch, than it is for firms wrestling with legacy IT systems.

“Integration is key,” said Trimmer. “Move data where you need it, where your teams are working, and ideally have one system per team.”