Finance One, a non-bank lender, is implementing the first AI strategy outline and framework to improve operational efficiency, improve service and accelerate new product launches.
Finance One is part of the Investor Central Group, which also carries two debt collection businesses acquired through acquisitions.
talk to iTnews PodcastCIO Chris Doyle said harmonization, modernization and automation are key themes in the company’s broader technology strategy and approach.
Where possible, technology efforts are being made to establish common systems that can be used by different entities within the group.
Legacy modernization is nearly complete, and the technology team has been able to “unbundle code, revert direct integrations, make it more asynchronous to achieve more robust processes, and improve performance by moving them to the cloud,” Doyle said.
“[Thirdly]we support companies’ ambitions to build new products and move into new areas, and that’s where the discussion of ubiquitous automation and AI comes in. [here]”
Doyle has a broad perspective on AI and automation.
The financial industry is no stranger to process automation and machine learning technologies, and Doyle sees them as having a continuing role to play in the AI conversation.
“When we think about automation, [tools]there is not one sledgehammer to crack a nut,” he said.
“The key is what kinds of tools are available. There are many vendors with varying degrees of maturity and different types of solutions that can solve your problem.
“This is something I have been directly involved in, working with CEOs and executives to clearly communicate what we are actually working on. [automation and AI] How do you choose a tool? How do you implement them? How can you make them successful? ”
Mr. Doyle is particularly interested in implementing automation and AI tools to help Finance One develop new products and improve the customer experience.
He said the group’s “outline AI strategy and AI framework have recently been completed” and discussions about internal implementation are “progressing well”.
“They’re positive. We’re running all sorts of experiments with LLM, putting in place a platform to manage LLM access, giving it good central visibility, making these tools available to all staff, starting to enable business areas where we’ve identified the use of agents, and supporting how we build effective agents,” he said.
Broadly speaking, the approach you choose, and the technology-based options you employ to build your automation, will depend on the “criticality” of the process to be improved and the amount of work you want to handle.
The company has made its Vibe coding tool Lovable available to business users to create their own automations for non-critical, low-volume processes.
Doyle said Vibe coding provides a way for business units to “start influencing processes that don’t normally reach the top of the priority list” to get the attention of technology teams.
“If you have a low-volume, low-risk process, let’s build something, and that’s fine.
“If the managers and teams that own the process aren’t going to build something with technology because it’s so expensive internally; [for us] to build something and then use it [vibe coding instead]”
More advanced processes, especially those related to core or regulated aspects of the business, will require more central technology team involvement when considering automation and AI use cases.
call quality
One of the areas where Finance One is implementing AI is in the contact center, where Icana.AI’s CallCoach is set up as an internal platform to review inquiries for specific metrics such as quality assurance, sentiment, and difficulty.
“We have over 100 contact center agents working in the collection space,” Doyle said.
“We have other customer service and other roles, but the majority of people who call are in those contact centers, so we want to make sure everyone has access to that capability.
Overall, the company receives approximately 2,000 calls per day and 10,000 calls per week.
Get all “conversations”. This is defined as a call that runs for at least one minute and is added to an S3 bucket from which CallCoach retrieves the conversation, analyzes it, and returns the results to dashboards and reports.
Doyle said AI has proven particularly useful in showing signs of distress that can trigger regulatory action by financial companies.
“Hardship is a regulated process where if someone meets the criteria to say they may be in hardship, they don’t have to use the word hardship. It’s about how we interpret what words they use to say whether they are in financial hardship,” he said.
“These are sensitive calls, and regulators are working very hard on this particularly difficult process because they want to make sure that predatory lenders and predatory organizations don’t cause any distress.”
Running AI means that 100% of the “conversations” can be analyzed for difficulty and other metrics, as opposed to a small sample of calls that only human teams have to review and analyze.
With CallCoach, Finance One’s team leaders now only receive “5 or 10” calls each day that require some action or follow-up.
“So instead of looking at a sample of something that may or may not be useful, they’re going to spend all their time looking at the call.” [flagged by CallCoach] Because they know it helps,” Doyle said.
Reactions vary. It might be asking customers for financial assistance, or it might be recognizing a well-handled conversation and using it as a training opportunity for other staff.
Doyle said that since the company started using CallCoach, “we’ve seen really tangible double-digit improvements in compliance and engagement.”
Quality has improved, which is reflected in the bottom line and the company’s bottom line, he added.
