Generative AI is making huge leaps and bounds, sending ripples across every industry. AI market size was estimated at $136.6bn in 2022 and is expected to reach $196.6bn in 2023.
Financial services is one of the industries that could be transformed by GenAI. However, AI has been in use in the financial sector for years, so what does the recent innovation of GenAI bring to the table?
GenAI in trading and risk management
Jürgen Eckel, managing director and partner at BCG X, Boston Consulting Group’s tech build and design unit, argues that the key benefit for the financial sector in using GenAI is in structuring data, particularly external data, adding that it will also make compliance departments more efficient.
“If you have tools that give you more freedom and time because they take away mundane tasks from your typically overloaded compliance and regulatory departments, that will ultimately make those departments more effective and efficient and will allow them to make better decisions — ultimately allowing you to have a more profitable business.”
Drishdey Caullychurn, founder of TEXpert AI, an AI-powered hiring platform that helps assess and improve workforce diversity, and a former trader, agrees.
“Data extraction, text summarisation, reporting, especially compliance, all of that takes a lot of time and these can be easily handled by GPT and quite safely — it can make things way quicker.”
One of the things that we do is use our AI models to sift through unstructured data to identify collateral
Two months ago, Bloomberg announced BloombergGPT, its large language model (LLM) being “purpose-built from scratch for finance”. But whether the model will be able to outperform human bankers and traders is yet to be seen — right now, it's unlikely.
“There are distinct human elements, for example, the pandemic or Ukraine war, which are factors that a large language model just cannot predict or analyse. In a world where everything’s changing, human insight is absolutely required in trading,” says Lewis Z Liu, cofounder at Eigen Technologies, a startup launched in 2015 that uses AI to transform documents into structured data.
Liu says that GenAI could also be used in predicting financial crises. “One of the interesting things in 2008, when I was working in finance, was that there was one bank that we were working with that couldn't find $2bn worth of collateral — it was hidden somewhere in all the data.
“Banking systems have terrible data and it caused a financial crisis. So today, one of the things that we do is use our AI models to sift through unstructured data to identify collateral.”
GenAI in marketing and customer experience
According to a survey from The Economist, 77% of bankers believe that the ability to unlock the value of AI will be the difference between the success or failure of banks. The study also showed that banks will continue to adapt tech into their internal structures in order to enhance customer experience, product offerings and new revenue streams.
When you ask ChatGPT to do recommend or ideate something, it's just mansplaining as a service. Even if it's not certain about the answer and might be incorrect, it's very authoritative
Eckel says that there’s huge potential in using GenAI in marketing and customer experience. “With a powerful tool like GenAI they can get highly automated, structured, data-driven imagery and marketing messages to get people to invest, open accounts and more.”
Chatbots and virtual assistants are already in use in financial services, but GenAI will help them respond to increasingly complex customer inquiries. “In an ideal world, you have an almost real-time coach for complex interactions — and for less complex topics, there's still benefits around brand communications when you respond to emails and other communications,” says Eckel.
But it’s not simply a bed of roses. Liu says that he’s sceptical about the use of generative models in offering tailored growth and innovation strategies. “When you ask ChatGPT to do that [recommend or ideate something], it's just mansplaining as a service. Even if it's not certain about the answer and might be incorrect, it's very authoritative.”
Caullychurn says that companies should also be mindful of upcoming regulatory changes while adopting GenAI. “We know that regulation is coming, so the possibility of spending a lot of money on it right now and then discovering in the future that we can't use it is something to think about.”
The risks of GenAI in finance
Liu says that one of the major hurdles in the adoption of GenAI in finance is the risk of inaccuracy or hallucinations (model errors). “How do I know if the results are accurate or if they’re not biassed? How do I know that the predictions it's making today are relevant even when the world changes, the market changes or the documents change?”
He says that financial institutions and companies should also think about the huge computing costs involved in employing GenAI models. It costs GPT4 around $14 to answer one question on one 100k-word loan document, but it costs a human being only $6-7 to do the same using Excel.
The domain-specific approach could be the answer. “LLMs by themselves are not fine-tuned and are extremely expensive to compute. So it makes sense to create smaller LLMs that are fine tuned to do certain domain-specific tasks like BloombergGPT.”
Have we given people comfort that we're not going to take your personal data or your highly proprietary company data and have it be regurgitated somewhere else by a competitor?
Data privacy is also a key area of concern in the adoption of the technology, especially in financial services.
Caullychurn gives the example of GenAI-driven high personalisation in personal finance management: “When talking about a hyper personalised service, where they're able to project the financial requirements of an individual from the cradle to the grave and even generations, you need to understand the level of privacy that you would be giving up, as a client, to achieve that.”
For Eckel, public perception could be one of the major hurdles in the adoption of GenAI, especially in the financial sector as finance directly affects people’s lives. “Everyone's so fascinated by the topic now and it's in a very positive space, but that might change.
“It could move towards ‘Oh my god, I don’t want the robots to talk to me’ or if there’s a headline that reads ‘GenAI chatbot advised elderly woman to put her life savings into penny stocks’.”
Liu says that regulation could be the answer to these issues, but it also comes down to accountability and transparency within the AI industry itself.
“It comes down to the industry and I'm talking about us — the overall AI industry. Have we given people comfort that we're not going to take your personal data or your highly proprietary company data and have it be regurgitated somewhere else by a competitor?”
Nevertheless, Eckel is bullish GenAI in finance is coming.
“There’s a whole spectrum of players who are starting to look at this, and financial services is probably in one of the best positions to start using it soon.”