By Slaven Bilac
The use of artificial intelligence is not a novel idea, even in the risk-averse financial services industry. While most of the effective uses of AI have traditionally taken place in the background - spam identification, fraud prevention, and risk assessments - that is not the case for AI systems like ChatGPT.
According to Demand Sage, ChatGPT has already achieved over 100-million users as of January 2023, a mere two months after its initial launch. This statistic makes ChatGPT one of the fastest-growing consumer applications, showing how immensely it has already permeated into multiple spheres of the workforce.
Although ChatGPT achieved significant advancements in natural language processing, natural language understanding, and generative text, chatbots are not a new or unique concept. With the range and complexity of the responses and the ease in conversation flow, there could be opportunities in the banking sector, but there are also significant obstacles that can hinder its deployment in the near future.
The Hurdles of AI Models
The member experience should be streamlined by chatbots; when a member asks a question, the chatbot responds or points them in the direction of a solution, saving everyone time. However, credit unions could find it challenging to use sophisticated chatbots like ChatGPT in environments where members are present, because the goal extends beyond simply speaking with the member and learning what they want. The interaction also involves taking action to address the issue, which creates several obstacles.
The back-end infrastructure that supports AI chatbots determines how valuable they are. As a result, credit unions must use API interfaces to make it possible for their systems to exchange data and promote easy interactions. That frequently necessitates substantial investments in IT architecture and collaborations with outside parties, which is no small task.
Everyone knows that regulatory and compliance requirements are significant issues in the banking industry. AI models like ChatGPT on occasion can make mistakes and generic models could use out-of-date data or erroneously interpret statements for certain circumstances. Nevertheless, the chatbot always responds with a tremendous amount of confidence, which can be especially problematic when it comes to ambiguous subjects like financial services.
The Advantage of Human Interactions
Understanding ownership and responsibility for advice given is an advantage of human interactions, which can often times be underestimated in today’s digital world. There is clear, unambiguous accountability. This can also result in people responding cautiously because they don't want to be held accountable for mistakes, which naturally leads to strict systems with clear hierarchy and standards. With AI, it isn't the case.
In addition, there is no empathy when using chatbots like ChatGPT. While employees can offer answers and solutions, they also sympathize with the member. Money and financial issues are delicate matters and managing these issues frequently necessitates emotional support, which humans alone can supply.
The Positive Impact ChatGPT Can Have
Currently, high volume, repetitive operations with little variation are best suited for AI systems. This includes content generation, particularly for areas that aren't subject to intense examination or scrutiny.
An AI chatbot might create templates for website pages or summaries of goods and services. Using a hybrid model creates efficiencies – a human proofreads, edits and polishes the writing and ChatGPT tackles the tedious heavy lifting. This method could save hours of time and improve efficiency.
Future Effects and Impacts of ChatGPT
ChatGPT will continue to make improvements every six to 12 months and accuracy will improve. We're moving faster toward a world of AI-to-AI interactions and services where individuals may have their own AI serve as their agent. Before the human equivalents take the reins to ratify or endorse findings made by their AI agents, your AI may communicate with a financial institution's AI to move through the earliest steps of an arrangement or negotiation, such as a mortgage loan.
However, while AI will certainly change workflows and roles, we shouldn't anticipate a widespread replacement of human-filled positions and interactions, especially in a service-based sector like financial services.
Advice for CUs Considering ChatGPT
Credit unions need to avoid dangerous KPIs. The main goal of an AI like ChatGPT is to improve the member experience, not to lighten the workload of the staff. Conversely, chatbot systems that place a high value on deflections (i.e., discussions that are not escalated to a person) are risky. Institutions run the risk of unintentionally boxing people in and reducing critical human-to-human interactions by stressing containment metrics.
Furthermore, as members consider negative experiences much more heavily than favorable ones, this may cause member relationships to fall apart.
Recognize that no AI system is complete enough to answer all questions. We've all shouted at chatbots or typed furiously at them, demanding to talk to a representative since our question wasn't being addressed. Credit unions must therefore arrange pleasant transfers to a human agent. Obstacles and delays in achieving this result may cause annoyance and reputational harm.
The Constant Threat
Long-term initiatives face a constant threat of interruption from updates and breakthroughs as AI systems like ChatGPT develop quickly. Smaller-scale operations that enable weekly or even daily modifications would benefit credit unions' overall performance, particularly in terms of obtaining the most for their money.
Slaven Bilac is CEO of Agent IQ, a provider of digital customer engagement solutions specializing in making financial services more personal again. Before launching Agent IQ, Bilac worked at Google for more than a decade in both their Japanese and American offices.