Since its launch in November 2022, OpenAI's text-generating artificial intelligence (AI) chatbot, ChatGPT, has been making headlines and stoking fears. There are a number of risks associated with ChatGPT, and as companies try to navigate these risks and capitalize on the rewards to stay competitive, its essential business leaders understand the possibilities of this emerging tech.
In the world of work, generative AI changes the game by encroaching — in theory at least — on a wider range of roles. But is this all just hype? And where do these latest advancements leave the financial institutions (FIs) that, until now, have been happy to harness the power of AI?
Pre-ChatGPT, AI tools played a significant part in the increased automation of critical processes. Machine-learning (ML) tools in particular have taken digital transformation to the next level for financial institutions by making far better and faster use of data than other technologies or human beings.
The models' ability to rapidly separate approvals from rejections accelerates time for approved applicants and wasted effort on applications that will ultimately be declined. It also allows credit officers to focus on more complex cases that could potentially drive growth and increase revenue — without adding risk to the portfolio.
Using AI score breakdowns and recommendations, credit officers can identify risk drivers for business borrowers and simulate how changes to terms of credit will influence the risk profile of each customer.
Within today's financial institutions, AI-powered ML models can work as both "supervised" and "unsupervised" tools.
In the context of, say, commercial lending, think of supervised AI as the new college graduate you train up to approve loan applications. With sample loans defining what makes "good" and "bad" loans, the graduate will steadily learn from the characteristics of previously approved or rejected applications.
With unsupervised AI, you simply give the metaphorical graduate a stack of loan files and ask them to identify interesting common characteristics between them.
Supervised AI is already revolutionizing how bankers approve credit by providing guidance on how well deals match an institution's risk appetite. It's also benefiting customers by helping bankers tailor products, prices and repayment structures to changing customer situations and life events.
Initially, banks needed to supervise the actions and decisions of AI carefully. However, once ML models can confidently select outcomes that align correctly to the strategy, vision and ethics of the organization, financial institutions are promoting them to an even greater position of unsupervised responsibility.
In addition to turbocharging banking processes and saving human employees time, AI tools can help financial institutions build a fuller, more in-depth and more dynamic picture of their customers and prospects.
CEO Ryan McInerney touted the growth of Visa Direct and unveiled a fraud-scoring service called RTP Prevent, which is powered by artificial intelligence, while discussing the card network's earnings for its fiscal third quarter.
For commercial lenders, for example, that could mean seeing where firms fit into the macroeconomic environment and their industry sector. In the financial cycle, they typically need access to funding or insight on how lending affects environmental, social and governance (ESG) scores.
By drawing on more data and digging into the details, AI could also help financial institutions identify lending opportunities that lesser technology might reject out of hand. And as credit risk rises once more, the faster, deeper insights it can provide are ever more critical to informing decisions, outthinking the market and overtaking the competition.
Above all, AI gives financial institutions the ability to draw on a wider, richer set of data: not only historic financials, but also covenant, transaction and market data, news feeds and social media. It's these more current data sources that can show how things are for business customers — and predict where they could be heading.
So far, then, so good: AI tools are already proving their ability to not only make day-to-day working lives easier but also forecast future challenges. But in just a few months, generative AI has rapidly gone several steps further. Rather than simply replicating manual processes and human decisions, tools like ChatGPT dig even deeper into available data to create their own textual or visual content.
Suddenly it's possible to automatically produce an entire, persuasively written dissertation or a disturbingly lifelike deepfake image. This relates back to how generative AI's capabilities exist on a spectrum — it's both astonishingly clever and a potentially dangerous way to cheat the system.
First, it introduces major, more sophisticated opportunities for fraud. What's to stop generative AI from faking documents to show a credit applicant is more profitable and creditworthy than they are?
Within financial institutions, the emergence of generative AI tools is sparking other fears. For example, a number of large banks on
Finally, of course, generative AI only heightens the anxiety that robotic tools will replace more and more human jobs. According to this argument, it doesn't just automate repetitive, mindless tasks — it puts higher-powered roles at risk, too.
But once the financial services industry has addressed its valid concerns about generative AI, there are many positive use cases for financial institutions to explore. These could include regulatory requirements, customer analysis, fraud and risk mitigation and product generation.
While there's much that AI can add to a business, what all AI tools lack is the human ability to read between the lines. So, side-by-side with AI tools, emotional intelligence, human experience and human judgment still have a major part to play in banking. Merging the technology with human judgment will be the key component to unlocking and harnessing the power of AI.
