20 April 2023
Artificial Intelligence
Beyond ChatGPT: The role of AI in banking
Dinner table chatter about artificial intelligence (AI) has surged since non-profit research laboratory OpenAI's launch of ChatGPT in late November last year. ChatGPT galloped to more than 100 million active users by January 2023 and then, in March, an even more sophisticated version was released: GPT-4.
Despite the hype, ChatGPT is not the only AI in town. There's Microsoft's AI-empowered Bing tool, Google's Bard and even paid-for options like Jasper, Otter.ai and YouChat, all of which fall under the banner of Large Language Models (LLMs). LLMs are user-friendly because they communicate using human-like language and have easy-to-navigate interfaces, so engaging with them feels easy and natural. On top of that, LLMs have the potential to take the grind out of mundane tasks such as setting up appointments, replying to emails and even coding. So, we can expect to see a proliferation of more AI tools in the months and years to come, as they become increasingly intertwined in our day-to-day lives.
The mainstreaming of AI has been met with a range of views, from Elon Musk (ironically one of the original co-founders of OpenAI) and his view that 'we are not far from dangerously strong AI' and ChatGPT represents 'one of the biggest risks to the future of civilisation', to Microsoft founder Bill Gates' assertion to Forbes magazine that AI is 'every bit as important as the PC, as the internet'. Locally, Stellenbosch University professor Johan Steyn advocates continuous learning and an adaptable mindset in response to the rapid pace of technological change and disruption which 'is posing significant challenges to individuals and organisations across industries and sectors'.
Much has already been written about the potential for AI and LLMs like ChatGPT to disrupt financial services by replacing advisors, streamlining the client experience and even assessing bond applications and insurance claims.
One man who embraces the potential of AI tools, but takes a more pragmatic view of their long-term disruption, is Dr Mark Nasila, Chief Data and Analytics Officer in RMB Private Bank's Risk division. Nasila, who holds a doctorate in mathematical statistics and probability from Nelson Mandela University, is a frequent commentator in the media on the subject of AI and its ability to impact multiple facets of our lives, including banking.
AI, LLMs in the world of banking
Nasila points out that in the world of banking and financial services, security and privacy are key. Therefore, any discussion around these technologies must consider that open platforms like ChatGPT - even if they are customised to a specific company's needs - will not offer the robust protections required by regulators and lawmakers.
Nasila notes that 'we've even seen some US banks blocking the use of ChatGPT among employees', referring to the likes of Citigroup, Deutsche Bank, Goldman Sachs, Wells Fargo and Bank of America which have responded quickly due to the legal implications associated with the use of the tool in what should be a highly confidential and data ring-fenced internal banking system.
Given that accountability in the use of AI is so critical, this rapid shutting down of indiscriminate use of external AI tools - particularly in such delicate banking divisions as forensics - is understandable. Indeed, right now many organisations, from universities to retailers and marketing teams, are grappling with how best to incorporate LLMs into their existing systems. RMB Private Bank is no different.
Nasila explains: 'In terms of RMB Private Bank as a Financial Services Provider, we have a clear idea of what we want to do with AI and what we want from LLMs. Obviously digitisation is a big one and it represents an opportunity for us to enhance our processes around, for example, how we understand what clients are saying as well as responding to their needs timeously.'
He notes that it is not only text generation applications around clients' experience and marketing where these tools can be deployed. In the banking world, there are important implications around legalities, intellectual property, and client privacy to consider. 'For example, as a big organisation what are the implications if we use ChatGPT to generate contracts and what if this information and input is not aligned to the value proposition of shareholders? Who takes responsibility?'
A hands-on approach
Given these considerations, how does a bank ensure that it can benefit from tech tools like AI and LLMs without compromising essential data, personal information, and trade secrets?
In the case of RMB Private Bank, Nasila explains that the use of AI-enabled technologies hinges on taking a risk-driven perspective. While exploring and understanding the strengths and limitations of generative AI-enabled apps or bots like ChatGPT (which are capable of generating text, media or images when prompted) is important, for use cases that involve any form of IP or confidential data RMB Private Bank's response was to develop its own, in-house AI tools which make it possible to flag risks and mitigate them, while still benefitting from the efficiencies and time-saving potential of AI applications. RMB Private Bank's forensic systems have been using the bank's own AI tool since 2019.
Since rolling out the technology, Nasila notes that the time taken to produce reports has been cut by 70% while generating a draft forensic synopsis for scrutiny by a human analyst has been reduced from hours to mere seconds.
At the moment, RMB Private Bank's Risk team is adding to its AI capability by piloting the use of AI for the digital onboarding of clients, as well as for verifying documents and confirming legitimacy. This represents just one area of potential development and opportunity for financial services institutions, without compromising client information and confidentiality.
What are the potential rewards?
Having outlined the possible risks associated with a laissez-faire rollout of AI in financial services, Nasila circles back to the opportunities that digitisation presents in the banking sector.
'Technology enables or helps us to imagine processes, but we need to have the right use cases to create efficiencies,' he explains. This requires time and effort to, for instance, determine exactly how best to use tech tools to enhance employee effectiveness by analysing the entire system from information gathering to how time is allocated for specific tasks, and even revisiting what might be regarded as important performance indicators in the future. A measured approach will, to some extent, also cushion the anxiety with which some might greet the mass arrival of AI tools in the workplace, thereby supporting the adoption.
As Nasila explained in a TechCentral article in 2022: 'For individuals, meaningful inferences can be made by looking at data on their smartphone (with their permission) which can reveal financial behaviour, from call activity to app usage, and even which applications a consumer uses most.... This opens the door to new avenues of lending: for example, smartphone-based microlending, bereft of the usually prohibitive and punitive interest rates such lending models tend to use to insulate themselves against risk.'
Nasila also notes that in sectors such as agriculture, AI can access satellite imagery to estimate past and future income from farming to help the bank make better decisions about loan applications.
But watch out for...
A consideration that Nasila does not feel has garnered sufficient attention in the current AI debate is how blind adoption across all levels of an organisation might have the unintended consequence of negatively impacting critical thinking skills in the future.
'The process of critical thinking comes down to trial and error. If you take away the trial and error process then you are likely to have a generation of people who never thought and never actually executed thinking,' he explains. 'This development would affect the ability of an organisation to come up with ideas, explore ideas, and arrogate their own processes, rather than relying on another system to tell them what to do. At the end of the day, LLMs like ChatGPT can only work with what they've been trained on and programmed to do.' This, notes Nasila, underlines the importance of having teams of highly skilled experts working behind the scenes to ensure that AI is fine-tuned to address the specific needs of the industry and business it serves.
At the moment, the hype lies around the technology as millions engage with AIs like ChatGPT for entertainment and novelty value. Much of the conversation is swirling around threats and possible disruptions to livelihoods and human value. Nasila's advice is: 'Don't buy into the hype!'
Right now, he says, the power of AI lies in its ability to automate and optimise routine tasks and augment the innovation process, but there is still room for the human being in this equation. In fact, critical thinking and human creativity have never been more important.
The role of the private advisor in the RMB Private Bank context is a case in point. In preparation for the AI technology onslaught, our private advisors have undergone extensive training and upskilling to enhance their very human and personal offering, and their ability to offer a holistic approach when giving financial advice. The power of our data-driven AI tools takes care of the routine tasks, giving our people time to form partnerships and make meaningful connections with clients.
Despite the hype, ChatGPT is not the only AI in town. There's Microsoft's AI-empowered Bing tool, Google's Bard and even paid-for options like Jasper, Otter.ai and YouChat, all of which fall under the banner of Large Language Models (LLMs). LLMs are user-friendly because they communicate using human-like language and have easy-to-navigate interfaces, so engaging with them feels easy and natural. On top of that, LLMs have the potential to take the grind out of mundane tasks such as setting up appointments, replying to emails and even coding. So, we can expect to see a proliferation of more AI tools in the months and years to come, as they become increasingly intertwined in our day-to-day lives.
The mainstreaming of AI has been met with a range of views, from Elon Musk (ironically one of the original co-founders of OpenAI) and his view that 'we are not far from dangerously strong AI' and ChatGPT represents 'one of the biggest risks to the future of civilisation', to Microsoft founder Bill Gates' assertion to Forbes magazine that AI is 'every bit as important as the PC, as the internet'. Locally, Stellenbosch University professor Johan Steyn advocates continuous learning and an adaptable mindset in response to the rapid pace of technological change and disruption which 'is posing significant challenges to individuals and organisations across industries and sectors'.
Much has already been written about the potential for AI and LLMs like ChatGPT to disrupt financial services by replacing advisors, streamlining the client experience and even assessing bond applications and insurance claims.
One man who embraces the potential of AI tools, but takes a more pragmatic view of their long-term disruption, is Dr Mark Nasila, Chief Data and Analytics Officer in RMB Private Bank's Risk division. Nasila, who holds a doctorate in mathematical statistics and probability from Nelson Mandela University, is a frequent commentator in the media on the subject of AI and its ability to impact multiple facets of our lives, including banking.
Nasila points out that in the world of banking and financial services, security and privacy are key. Therefore, any discussion around these technologies must consider that open platforms like ChatGPT - even if they are customised to a specific company's needs - will not offer the robust protections required by regulators and lawmakers.
Nasila notes that 'we've even seen some US banks blocking the use of ChatGPT among employees', referring to the likes of Citigroup, Deutsche Bank, Goldman Sachs, Wells Fargo and Bank of America which have responded quickly due to the legal implications associated with the use of the tool in what should be a highly confidential and data ring-fenced internal banking system.
Given that accountability in the use of AI is so critical, this rapid shutting down of indiscriminate use of external AI tools - particularly in such delicate banking divisions as forensics - is understandable. Indeed, right now many organisations, from universities to retailers and marketing teams, are grappling with how best to incorporate LLMs into their existing systems. RMB Private Bank is no different.
Nasila explains: 'In terms of RMB Private Bank as a Financial Services Provider, we have a clear idea of what we want to do with AI and what we want from LLMs. Obviously digitisation is a big one and it represents an opportunity for us to enhance our processes around, for example, how we understand what clients are saying as well as responding to their needs timeously.'
He notes that it is not only text generation applications around clients' experience and marketing where these tools can be deployed. In the banking world, there are important implications around legalities, intellectual property, and client privacy to consider. 'For example, as a big organisation what are the implications if we use ChatGPT to generate contracts and what if this information and input is not aligned to the value proposition of shareholders? Who takes responsibility?'
Given these considerations, how does a bank ensure that it can benefit from tech tools like AI and LLMs without compromising essential data, personal information, and trade secrets?
In the case of RMB Private Bank, Nasila explains that the use of AI-enabled technologies hinges on taking a risk-driven perspective. While exploring and understanding the strengths and limitations of generative AI-enabled apps or bots like ChatGPT (which are capable of generating text, media or images when prompted) is important, for use cases that involve any form of IP or confidential data RMB Private Bank's response was to develop its own, in-house AI tools which make it possible to flag risks and mitigate them, while still benefitting from the efficiencies and time-saving potential of AI applications. RMB Private Bank's forensic systems have been using the bank's own AI tool since 2019.
Since rolling out the technology, Nasila notes that the time taken to produce reports has been cut by 70% while generating a draft forensic synopsis for scrutiny by a human analyst has been reduced from hours to mere seconds.
At the moment, RMB Private Bank's Risk team is adding to its AI capability by piloting the use of AI for the digital onboarding of clients, as well as for verifying documents and confirming legitimacy. This represents just one area of potential development and opportunity for financial services institutions, without compromising client information and confidentiality.
Having outlined the possible risks associated with a laissez-faire rollout of AI in financial services, Nasila circles back to the opportunities that digitisation presents in the banking sector.
'Technology enables or helps us to imagine processes, but we need to have the right use cases to create efficiencies,' he explains. This requires time and effort to, for instance, determine exactly how best to use tech tools to enhance employee effectiveness by analysing the entire system from information gathering to how time is allocated for specific tasks, and even revisiting what might be regarded as important performance indicators in the future. A measured approach will, to some extent, also cushion the anxiety with which some might greet the mass arrival of AI tools in the workplace, thereby supporting the adoption.
As Nasila explained in a TechCentral article in 2022: 'For individuals, meaningful inferences can be made by looking at data on their smartphone (with their permission) which can reveal financial behaviour, from call activity to app usage, and even which applications a consumer uses most.... This opens the door to new avenues of lending: for example, smartphone-based microlending, bereft of the usually prohibitive and punitive interest rates such lending models tend to use to insulate themselves against risk.'
Nasila also notes that in sectors such as agriculture, AI can access satellite imagery to estimate past and future income from farming to help the bank make better decisions about loan applications.
A consideration that Nasila does not feel has garnered sufficient attention in the current AI debate is how blind adoption across all levels of an organisation might have the unintended consequence of negatively impacting critical thinking skills in the future.
'The process of critical thinking comes down to trial and error. If you take away the trial and error process then you are likely to have a generation of people who never thought and never actually executed thinking,' he explains. 'This development would affect the ability of an organisation to come up with ideas, explore ideas, and arrogate their own processes, rather than relying on another system to tell them what to do. At the end of the day, LLMs like ChatGPT can only work with what they've been trained on and programmed to do.' This, notes Nasila, underlines the importance of having teams of highly skilled experts working behind the scenes to ensure that AI is fine-tuned to address the specific needs of the industry and business it serves.
At the moment, the hype lies around the technology as millions engage with AIs like ChatGPT for entertainment and novelty value. Much of the conversation is swirling around threats and possible disruptions to livelihoods and human value. Nasila's advice is: 'Don't buy into the hype!'
Right now, he says, the power of AI lies in its ability to automate and optimise routine tasks and augment the innovation process, but there is still room for the human being in this equation. In fact, critical thinking and human creativity have never been more important.
The role of the private advisor in the RMB Private Bank context is a case in point. In preparation for the AI technology onslaught, our private advisors have undergone extensive training and upskilling to enhance their very human and personal offering, and their ability to offer a holistic approach when giving financial advice. The power of our data-driven AI tools takes care of the routine tasks, giving our people time to form partnerships and make meaningful connections with clients.