From Transactions to Trust:
How generative AI helps banks build deeper customer relationships
11.20.2024 | By: Nico Stainfeld, Tireni Ajilore
This is the second installment in our series examining how generative AI will reshape retail and SMB banking, which is based on our research and conversations with 20 banking executives across business, compliance, product and AI research. In our previous piece, we explored how AI is transforming customer acquisition, onboarding, and underwriting.
Today, we explore the next two stages of the banking lifecycle: customer engagement and delinquency management. Here, generative AI is helping banks solve two of their most pressing challenges: shallow digital engagement and costly collections processes. For incumbents and startups alike, the opportunity lies not just in automating existing processes, but in creating new ways to build trust and deliver value at scale.
Stage 3: Customer engagement
The digital engagement paradox
Customers today expect their digital banking experiences to be seamless, convenient, and tailored to their needs. To meet these high expectations, banks must continually invest in new technology. However, delivering a consistent experience across channels—whether it’s mobile, online, or in-person—can be difficult.
Digital channels, despite their convenience, often generate surprisingly shallow engagement. Most customers use banking apps merely for quick transactions—checking balances or making payments—without exploring the full range of valuable services available to them.
Consider this data from Central Bank: their mobile app records over 7 million monthly logins, with customers checking their accounts 20-26 times per month. However, as EVP Daniel Westhues observes, “The average time on app is still less than a minute. When they’re coming to our app, they’re very transactional.”
This superficial engagement carries real consequences for banks’ bottom lines. Sarah Welch, Managing Director at Curinos, quantifies the cost of this digital disconnect: customers who open accounts online bring just one-fifteenth of the deposits compared to their branch-acquired counterparts. They’re also two to three times more likely to leave within their first year and buy fewer products over time. This digital profitability gap represents one of banks’ most pressing challenges—and potentially one of their greatest opportunities for innovation.
The banking industry’s early attempts to bridge this gap brought us chatbots that could answer basic questions and recommend products based on what others had bought. While these systems reduced call center volume and introduced basic automation, they couldn’t replicate the deep understanding and empathy that a human interaction can provide.
Looking forward, generative AI can help banks deepen these digital relationships. Banks will leverage AI to proactively anticipate customer needs—whether it’s setting up a rainy day fund, offering tailored advice on how to improve your credit score, or even helping you manage subscriptions you might have forgotten about. These interactions will be seamless and proactive, with AI working quietly behind the scenes to enhance customers’ financial well-being.
Opportunities for incumbents
AI-powered virtual assistants and customer churn analysis
Banks can greatly benefit from using generative AI to create virtual assistants that offer personalized advice to their customers. These AI-powered assistants can do more than just answer basic questions—they can provide tailored financial guidance, helping customers make smarter decisions about their money.
By understanding each customer’s unique needs and goals, these AI advisors can suggest the right products, offer budgeting tips, or even help with investment planning. This kind of personalized support can make digital banking feel more like a one-on-one conversation with a trusted advisor, helping customers feel cared for and confident in their financial choices.
Leading institutions are already implementing these solutions. Bank of America has developed “Erica,” an AI-powered virtual assistant that helps customers manage their finances more effectively. Wells Fargo has introduced a similar AI-driven tool called “LifeSync,” which helps customers stay on top of their financial goals and make informed decisions.
Opportunities for startups
Advanced voice bots
A primary technical challenge in customer engagement involves creating voice bots capable of natural, human-like conversation. Previous attempts often produced rigid, scripted interactions that frustrated users rather than helping them. As one AI Lead at a Mortgage Lender notes, “You can have human-like speech, you can have accuracy and information, and you can have lack of latency—but you can’t have all three right now.”
Startups can leverage generative AI to develop voice bots that are far more advanced. By using LLMs trained on diverse datasets, these systems can understand a wide range of customer inputs, adapt to various dialects and languages, and provide nuanced responses that feel more human. They can manage multi-turn conversations, maintaining context throughout extended dialogues while delivering consistent, relevant answers. This makes interactions smoother, more intuitive, and ultimately more satisfying for customers.
Companies like Posh and Glia are developing voice bots specifically for banking, addressing the unique needs and complexities of financial services. These solutions comply with strict regulatory requirements and prioritize security, offering peace of mind to both banks and customers.
Stage 4: Delinquency and collections
The hidden cost center
While non-performing loans represent only a small fraction of most portfolios, they consume a vastly disproportionate share of resources. In mortgage servicing, default-related expenses—including loss mitigation, bankruptcy, and foreclosure—account for nearly one-third of servicing costs despite affecting only a minimal portion of loans. This imbalance stems from several factors: the emotional strain of collections work leads to high agent turnover, complex compliance requirements demand constant vigilance, and legacy systems struggle to support the dynamic, empathetic interactions needed for successful resolution.
In the early days of AI in delinquency management, chatbots took on routine tasks like payment scheduling and account updates. While these initial AI tools were basic, managing only straightforward queries with predetermined responses, they laid the groundwork for more sophisticated applications by reducing workload and making early collections processes more efficient and customer-friendly.
Today, generative AI represents a fundamental shift in collections strategy. Rather than simply automating tasks, these systems provide real-time agent assistance, continuous compliance monitoring, and personalized engagement strategies. This evolution transforms collections from a pure cost center into an opportunity for relationship recovery and enhancement. As Andrew Beddoes of Provenir observes, “Banks are starting to view these channels, call centers, etc., not as places for complaints to go to die, but as places for complaints to go to turn into profit.”
Opportunities for incumbents
AI agent assist and conversational payment rescheduling
The emotional demands of collections work often lead to high staff turnover. To address this challenge, banks should implement generative AI solutions that directly support agents. Real-time AI coaching during calls can guide agents through difficult conversations, suggesting effective responses while helping them maintain empathy and composure. This support not only reduces stress but also builds agent confidence. AI-driven training programs can also simulate realistic scenarios and provide immediate feedback, enabling new agents to become proficient more quickly.
Cresta AI’s implementation at a major U.S. bank demonstrates these benefits: initially, key components of negotiation occurred in only 5% of calls. With Cresta’s real-time guidance, team leads and collectors 4x-ed their playbook adherence within a month, resulting in larger and more timely payments. This shift echoes observations from Mike Murchison, CEO of Ada, who reports customers are seeing a 25% increase in automated resolution rates through generative capabilities. He emphasizes that “the skills producing these results are much more coaching-related than workflow-related.”
Beyond agent support, the technology’s predictive capabilities open new frontiers in early intervention. Advanced AI systems can now identify accounts at risk of default long before traditional indicators would raise alerts, enabling banks to develop personalized outreach strategies and optimize intervention approaches based on real-time effectiveness data.
Opportunities for startups
Compliance management
The compliance challenge in collections presents perhaps the most immediate opportunity for startup innovation. Traditional quality assurance processes typically review less than 5% of customer interactions—a dangerous blind spot in a heavily regulated industry. As Christopher Reese, Director of Customer Value at Cresta, points out, “Random QA on a subset of calls, a very small fraction of calls, because this QA is typically done by humans… we’re QA checking for compliance on probably less than 5% of the calls that actually take place.”
Startups like Sedric are addressing this gap by using AI to monitor all customer interactions for regulatory adherence, offering banks a comprehensive, real-time view of their compliance posture. Ivan Edwards, Vice President of Customer Interactions at Cadence Bank, highlights the potential: “AI can come in and say, ‘Hey, here’s the pop-up, make sure you read this disclosure.'” This capability extends beyond mere monitoring—as one AI Research Lead at a top five bank put it to us: “AI solutions are being built to do real-time redaction—imagine speaking to a customer rep and your sensitive data is automatically redacted in real-time, never stored on the system.”
AI voice agents for collections
The second frontier for startups lies in developing AI voice agents for collections. The business case is compelling: contact center labor costs currently consume up to 95% of operational expenses. Gartner predicts that conversational AI could reduce these costs by $80 billion by 2026.
More than just providing cost savings, the next generation of AI voice agents promise to transform the collections experience itself. Rather than simply handling routine payment processing and account inquiries, these systems will personalize interactions based on sophisticated analysis of customer data and behavioral patterns. They’ll determine optimal contact times, select appropriate communication channels, and adjust their tone contextually. This approach borrows from successful engagement strategies in retail banking, where AI-driven personalization has dramatically improved customer satisfaction and loyalty.
Importantly, these voice agents aren’t meant to replace human collectors entirely—nor should they. Instead, they’ll create a hybrid model where AI handles routine cases while human agents focus on complex situations requiring nuanced judgment and emotional intelligence. This division of labor promises to make collections more efficient while simultaneously improving the experience for both agents and customers.
This post is part of our series on how generative AI is reshaping banking. Start with our overview of AI’s impact across the banking lifecycle, then explore how AI is transforming customer acquisition and underwriting in part one.
Published on November 20, 2024
Written by Nico Stainfeld and Tireni Ajilore
