By Samah Chowdhury
Customer interaction technologies have evolved as a result of consumer demand for fast and convenient banking, developing from basic interactive voice recognition systems to sophisticated generative AI-powered virtual agents. Over time, each technology built on the strengths of its predecessors while addressing its limitations. These technologies aim to facilitate automated, efficient and increasingly human-like communication between banks and their customers across various channels, including voice, text and digital interfaces.
Timeline of advancements
Click chart for larger view.
IVR systems first were introduced to U.S. bank customers in the 1980s. These systems laid the groundwork for automated telephone interactions using prerecorded or synthesized voices to guide users, answer basic queries and route calls. IVRs significantly reduced call center volume for essential queries, allowing human agents to focus on more complex issues. While not inherently “intelligent,” original IVR systems paved the way for increasingly sophisticated automated customer interactions.
As digital platforms gained prominence in the early 2000s, rules-based chatbots extended the IVR concept to digital channels through text-based interactions. Chatbots provided consistent answers to standard questions using preset rules and decision trees, improving response times and availability. Today, 30 percent of banks with under $3 billion in assets have either implemented or plan to implement chatbot technology within the next one to two years. This adoption highlights the growing importance of chatbot solutions in modern banking, as banks look to improve efficiency, cut costs and provide automated 24/7 customer support to remain competitive.
The integration of machine learning capabilities in the 2010s heralded a new era for chatbots, enhancing their functionality with improved natural language understanding and learning abilities. This technological leap allowed chatbots to handle more nuanced queries and adapt to changing customer needs. Bank of America’s chatbot, called Erica, stands as a testament to the success of machine learning-powered chatbots, having served over 19.5 million customers and processed more than 230 million requests since its launch in 2018, showcasing the technology’s ability to scale and evolve with user interactions. Companies such as Boston-based Posh are helping to make this type of ML-powered chatbot accessible to community banks, as well.
Today, virtual agents combine voice and text interactions with deep learning capabilities, offering comprehensive and intelligent automated systems. For example, Capital One’s Eno delivers personalized insights, fraud detection and account management across voice and text channels. While the boundaries of these technologies are still being tested, the progression of automated customer interaction systems shows a clear trend toward more natural, intelligent and versatile interactions.
Pace of adoption
The development of these technologies in banking has been relatively rapid but adoption has been uneven. Large, tech-savvy banks have quickly implemented advanced solutions, while many smaller institutions have yet to begin researching these technologies. Moreover, significant differences exist in automated customer interaction system capabilities within the same institution across different channels (e.g., mobile app vs. website). The pandemic accelerated adoption across the board as banks sought to maintain customer service with reduced in-person interactions. However, the pace has varied by technology:
- IVR systems are widely adopted, but many banks are upgrading to more sophisticated versions;
- Simple chatbots are still common, while generative AI-powered chatbots are still in the early stages of testing and implementation; and
- Tech-forward banks are implementing virtual assistants, but they aren’t yet ubiquitous.
Adoption mainly has been driven by the need for banks to be available to customers 24/7 with instant responses to common queries. For banks, this has meant that they’ve been able to manage scalability during peak times and collect insights to serve their customers more effectively.
Design complexities
Product development is an ongoing process that extends beyond initial product-market fit. This need is evident in the progression from traditional chatbots, which rely on predeveloped outputs and manual updates based on solicited feedback, to generative AI systems that craft responses in real time. In particular, machine learning has enabled customer interaction products to understand customer questions, predict follow-up inquiries and dynamically improve with nuanced, context-aware responses. However, despite the progress, a design challenge exists, with user flows often requiring refinement to ensure a quick and seamless hand-off to a human agent.
With tools powered by emerging technologies, ensuring the accuracy and reliability of responses requires rigorous testing and training. User sentiment, adoption and patience also play a significant role, as customers need to feel confident in the system’s capabilities. Geoffrey Moore’s book Crossing the Chasm tells us that refining high-tech products with user feedback is an ongoing process that can take considerable time as different user groups engage with the product; automated customer interaction systems are not an exception.
Generally, with higher levels of automation, there is a need for risk mitigants, such as a human in the loop or, in this case, a “human behind a button.” How your user flow is designed and how effectively it is integrated within your systems will translate to customer support and satisfaction levels. A primary focus for banks implementing customer interaction systems is ensuring compliance with consumer protection laws, which remains an ongoing and critical concern. Equally important is the development of fair and nondiscriminatory AI decision-making processes that are transparent and explainable. Adhering to privacy regulations and data handling are essential to a bank’s assessment of technologies and strategic applications. Banks face the ongoing challenge of meeting customer needs amid growing supervisory scrutiny. In the near term, the CFPB’s plan to issue chatbot rules could influence the course of innovation by constraining banks’ ability to leverage this technology..
Looking ahead
The integration of advanced technologies like AI, machine learning and natural language processing is set to revolutionize customer interactions in banking. AI-powered virtual agents can increasingly anticipate customer needs by analyzing data such as transaction history, browsing behavior and economic trends. For instance, based on a customer’s financial patterns, a virtual agent might predict when they’ll need a new product like a mortgage or investment account and proactively offer support. This is especially critical as all generations — Generation Z, millennials, Generation X and baby boomers — share a common frustration with the lack of personalized recommendations in digital banking experiences. In addition, predictive analytics models that include machine learning are helping to improve efficiency by detecting fraud in real-time, assessing risks holistically and streamlining customer inquiries, making interactions faster and more secure.
Multimodal AI is also set to enhance customer engagement by combining voice, text and visual interfaces, creating more intuitive and seamless interactions. For instance, a customer might start a loan application process via a mobile app, continue the conversation with a voice-activated AI assistant and complete the process through a video chat with a human representative — all while the AI system maintains context across these different modes. Multimodal AI significantly can enhance accessibility, making banking services more inclusive for customers with diverse needs and preferences. For banks, this technology can lead to increased engagement and improved customer experience by allowing customers to interact through their preferred channels.
Emotion AI is also rising as a powerful tool that allows companies to analyze customer emotions and stress levels during interactions, enabling real-time adjustments to improve service quality. The foundation of emotion AI lies in affective computing, with companies such as Affectiva, which also is headquartered in Boston, pioneering technologies in emotion recognition. Typically, IVRs are integrated with advanced AI systems capable of recognizing, interpreting, processing and simulating human effects. Recent advancements have focused on analyzing agents’ and customers’ vocal patterns and facial expressions during interactions to analyze micro-shifts that impact sentiment. Ultimately, the benefit of advanced technologies offers real-time ability to detect subtle cues in customer interactions, allowing for immediate improvements in service delivery.
The current pace of emerging technologies in product innovation indicates a future where banking is more responsive and attuned to individual customer experiences. Perhaps, in time, chatbots and IVRs will become the conversationalists we never expected.
Samah Chowdhury is senior director of innovation strategy in ABA’s Office of Innovation.
