2023 was the year of AI in every respect. Following the launch of ‘ChatGPT’ by OpenAI at the back end of the previous year, it seemed like everyone in the fintech industry was thinking about its potential. But now over a year has passed, have firms worked out how to best, and safely, implement AI into one or several elements of their business?
Both Gary Higham, CTO, and Kieran Molloy, machine learning engineer, at car finance marketplace, Zuto recognise the opportunity of AI; but also the importance of customer service. Here, the two discuss what they have learnt about utilising AI so far, and the barriers they see slowing down adoption.
The awe inspired by the launch of ChatGPT in November 2022 was accompanied by fear – surrounding its power and its potential to eradicate the need for humans. AI’s rapid development has since wowed even its early adopters. And, while it is becoming more widely adopted for its operational and analytics capabilities, even the most tech-savvy firms – like fintechs – are still figuring out how best to use it.
Companies are currently trialling different AI models to get the right balance of speed, functionality and cost, and to ensure that processes such as communicating with customers and gathering data are secure and adhere to regulations. Here’s what we have learned about where it can best be used and what the barriers to adoption are.
AI as co-pilot
One of AI’s greatest strengths is its ability to enhance customer interactions, not to replace the human element. In a survey of more than 2,000 Zuto customers, 82 per cent said that good customer service was most important to them when taking out car finance. That’s why we see AI as a customer service co-pilot.
Checking documentation and inputting data is an arduous but necessary part of a customer service agent’s role.
When securing car finance, customers need to send various documents, for example, their driver’s license. We use AI to check that all documentation has been retrieved and input into the database and that it’s correct and valid.
Generative AI (GenAI) is able to cover every communication channel across the full customer interaction space, whether that’s SMS, voice or email. Automating this not only avoids the opportunity for error, but also frees up time for our people to have more meaningful interactions with customers.
AI as a listening tool
Another useful application in fintech is AI’s listening capabilities. GenAI can analyse needs and intentions more quickly and accurately than a human, so it can send prompts and information to support conversations, leading to more relevant and satisfying outcomes for both the customer and the agent.
This also makes it an indispensable tool for enabling compliance with Consumer Duty regulations: listening to and transcribing interactions will ensure all required information is shared with the customer. Not only will it record evidence that customer interactions have happened as they should, but it will also prompt the customer service agent in real-time, to ensure they cover everything they need to.
Taking this one step further, GenAI can pick up on any potential vulnerabilities the customer has and alert the human worker so they can follow the due processes and deploy the right approach, depending on the circumstances.
AI as a training assistant
At Zuto we have a rigorous training programme in place for sales and customer services new starters. This ensures our people can give customers the best options and can cover all that is required from a regulatory perspective.
Having AI working alongside humans means that, rather than having to remember all compliance regulations and finance information, the relevant details can pop up in front of the sales or customer service agent in real-time – while they’re communicating with the customer. In this case, AI enhances automated processes by listening to the conversation and knowing when to provide what information.
Barriers to AI adoption
While businesses know they need to be looking at AI, many have not progressed very far with its implementation. One of the greatest barriers to adoption is likely to be cost. That, and the huge amount of energy that GenAI will consume, are the elephants in the room.
Several advanced AI models are available allowing for some choice in price, but the fact is that using AI extensively can be expensive and will use a lot of energy. There are cheaper models than OpenAI, for example, that can provide results, but don’t necessarily have the breadth and depth of the larger offerings. At Zuto we’re trialling different models and are factoring in speed, cost and capabilities to work out the most cost-effective and functional solutions for us.
When it comes to energy consumption, GenAI is seriously power-hungry. With the largest tech firms like Google and Microsoft embedding AI into their search technology and software, the demand for electricity is significant. EM360 cites PhD research that calculates that by 2027, Google could consume roughly the same electricity as a small country annually (between 85-134 terawatt-hours).
With all of this in mind, businesses will need to prioritise the areas where they believe AI will have the biggest impact – running it across everything is simply not a viable or affordable option. The glimmer of hope is that AI will help the tech giants to optimise their energy consumption as the tech develops.
AI working alongside us
As GenAI is relatively new it will continue to evolve and grow, increasing its capacity to do more on its own. Its ability to accurately check data for errors, inconsistencies and anomalies and take on more tasks that human workers find repetitive makes it a superior candidate to do these jobs.
However, human interaction remains crucial, so AI is not going to do us all out of a job any time soon. It could be the perfect colleague though.
