With the rapid emergence of generative AI (gen AI), credit customer assistance and collection functions are taking advantage of the technology’s potential. They can use it to enhance operational capabilities, improve efficiency, increase effectiveness, and—most importantly—create better outcomes for customers.
In recent years, technological disruption has been an inseparable component of credit customer assistance and collections. The shift has been driven by increasingly tech-savvy customers and transparency demands from regulators, both fueled by the COVID-19 pandemic and other credit crises. So far, these technological advancements, such as machine learning (ML) modeling, digitization, and automation, have enabled credit customer assistance and collections to become more streamlined, data driven, and customer oriented. New technology has allowed the offering of more services, more relevant arrangements with customers, new renegotiation pathways, and improved settlement conditions. These can strengthen the customer relationship with institutions, improving customers’ financial health and long-term value to institutions.
Gen AI is the latest and potentially most transformative of these advancements, and it can have an unprecedented positive impact on customer assistance. It can improve and personalize customer contact, boost the capability of agents serving clients, and automate routine processes, such as note taking, interaction summarization, and even some customer interactions. In turn, these benefits can aid the regulatory process through the technology’s ability to organize and synthesize information.
As a result, the adoption of gen AI in the customer assistance and collections space is by no means limited to use in reducing delinquencies. It has the potential to significantly improve customer interactions and treatment and drastically reduce collection-related costs by freeing up resources in operations while effectively addressing credit losses. This enhanced credit efficiency might enable businesses to retain collections in-house as a core capability and capture additional benefits, such as customer loyalty serving as a new source of competitiveness in managing the cost of extending credit to customers. Some early use cases are already yielding measurable results.
In our experience, organizations that deploy advanced gen AI capabilities in customer assistance and collections can achieve up to a 40 percent reduction in operational expenses and improve recoveries by about 10 percent. Additionally, collections could see up to a 30 percent increase in customer satisfaction scores, driven by the technology’s ability to better identify and address customers’ needs on time, helping them become debt-free more quickly.
In this article, we identify the needs of customer assistance and collections functions and discuss where gen AI can add value both to organizations and to customers. We also explain when and where gen AI can be implemented and discuss three gen AI use cases that, in our view, will dramatically change the operations for collections and customer assistance.
Challenges of customer assistance and potential of gen AI
The goal of customer assistance and collections is to support customers in overcoming financial distress while minimizing losses and keeping operational costs low—efforts that enable institutions to foster strong relationships and loyalty with their customer base. These functions must balance efficiency and effectiveness without compromising the overall portfolio risk profile and customer experience.
Collections functions are typically tasked with four main priorities:
- Creating a positive experience in the customer journey. This has become the core obligation of the function. That means giving relevant and meaningful financial advice, offering payment holidays when appropriate, and proactively engaging at an early stage of delinquency.
- Managing value at risk by strategically lowering financial risk. This priority includes identifying which intervention is needed—and when—for each customer, based on their circumstances and ability to pay.
- Minimizing cost without compromising efficacy and experience. This includes knowing when and how to reach out to a customer, automating time-consuming tasks such as data collection and note taking, and providing incentives for using self-serve channels.
- Adhering to regulatory guidelines and customer duty. Strong customer care requires sensitivity for the intensity and tone of messages, analytics-based guardrails to avoid bias and availability, and the identification and implementation of the right products to improve customers’ financial outlooks.
Gen AI can be used as a powerful tool to support the overall digitization of customer assistance. It’s ideal for the many customers who prefer to negotiate with a machine over having to share their difficulties with a human. Gen AI can also provide a more personalized touch in messages sent to a customer base.
We see four fundamental areas, all of which can lead to better outcomes for the customer, emerging for applying gen AI in customer assistance and collections:
- Reducing demand for manual intervention. Gen AI can be used at scale in analyzing call transcripts and chat interactions to identify the core issues a customer is facing, such as when customers didn’t receive statements and forgot payments. By addressing these root causes proactively, institutions can reduce demand for agent intervention, improving customer experience by making interactions faster, less stressful, and personalized.
- Gathering insights and improving operations. Gen AI applications can be fine-tuned on specific call models and employ quality control metrics to semiautomate the continuous improvement of operations. For example, the technology can interpret screen captures of common system reports to generate insights for a call center’s control desk and ultimately automate parts of this function for greater efficiency. Combined, these additions can also enable agent coaching, enhanced performance management, and early intervention in quality issues. All of this can be done at scale using the information from all client communications rather than samples, both improving customer experience and helping to reduce financial risk.
- Supporting agents and freeing up time. Gen AI can bolster the capabilities of case handlers in real time to improve experience and help reduce financial risk. This can range from adding a knowledge assistance tool to clarify a policy or offer eligibility to interpreting conversations and suggesting an interaction approach, tone, or product to the agent. Ultimately, this could occur through automation. In turn, such a boost can reduce or fully eliminate the need for agents to spend time manually writing post-call notes into a system, freeing up their time for cases that require a high-touch approach.
- Automating interactions. Gen AI can help power the next generation of chatbots, human-like interactive voice response (IVR), and even virtual agents. These tools can potentially offer increased empathy and high-quality solutions for customers while speeding up the process. Additionally, they can power hyperpersonalized messages both in these channels and in mass communications (such as emails and text messages), further improving their effectiveness and the user experience.
Gen AI implementation across credit customer assistance
Getting gen AI up and working in customer assistance isn’t as simple as plugging in a computer. Customer care leaders need to be sure capabilities put in place during early development enable the efficient growth of the gen AI ecosystem (see sidebar “Principles for implementing a generative AI customer journey”). The potential benefit of an all-in approach may be tempting, but simple, small, and manageable steps better serve functions initially.
When considering the implementation road map, leaders will have to balance value creation against disruption to the business and the potential for bugs. One smart approach that players are adopting is prioritizing high-value, internal use cases. These use cases can be built in a modular way, allowing for later deployment for customers when data, regulatory, and risk constraints are lifted.
Innovative customer assistance functions are choosing gen AI use cases that can be built and implemented rapidly without the need for complicated technical investments. These use cases typically involve using ready-to-use large language models (LLMs) that require limited development efforts and have minimal risk, as they rely on public or internal data and aren’t client facing. Additionally, they tackle a function’s area or process that is clearly defined, not scattered, and can capture impacts such as customer call insights and quality control effectively.
Early on, these use cases shouldn’t require sophisticated fine-tuning or content interpretation. Instead, they should have a limited yet clearly defined set of guardrails. For example, a gen AI use case could be for analyzing call data to identify factors contributing to successful outcomes. In this scenario, the use case is simple, manageable, and easy to measure: the low-cost ability to analyze call volume has a short implementation timeline, minimal integration expenses, and limited change management or retraining requirements.
On the midterm horizon, players are considering gen AI use cases that involve real-time output. These use cases often require more controls and security measures than less-advanced ones do, as they may involve the use of confidential customer data. However, the output of the model doesn’t directly interact with customers, as it requires human intervention instead.
Advanced applications of gen AI typically require a larger set of unstructured data from various sources to be fine-tuned. As a result, they require more advanced testing and validation processes and are more likely to be built and deployed across different areas or functions within an organization.
The most advanced applications of gen AI will require significant development effort and investment, which often leads to implementation timelines of roughly two to three years. These use cases are typically client facing. They will require both sophisticated environments to reduce latency to acceptable levels and robust guardrails to safeguard both the data exchange and the output to customers. They might be costly using today’s technology.
In the long term, to truly capture the benefits of gen AI, leaders should consider how its deployment affects the end-to-end journeys of both the customer and the customer care team. Combining different use cases has much more impact than developing individual use cases does. When coordinated, one use case can leverage another to amplify the individual impact while building on the same modular architecture.
Moving to a mature gen AI system is transformational. Each area enhanced by this innovative technology will need a revised operating model to fully capture the value generated. Adjustments will be needed for existing processes, policies, human intervention, staffing, and more.
Three concrete gen AI use cases for customer assistance
Our research shows that end-to-end transformation of a business domain such as collections with gen AI use cases involving augmentation, automation, and demand reduction can yield up to 30 percent productivity gains. Customer assistance functions across institutions around the world are already implementing gen AI. Here are three examples of how gen AI has enhanced the process. These examples come with the caveat that capturing the full potential of gen AI requires the deployment of a whole portfolio of use cases that integrate with one another.
Gen AI as a low-cost, high-value performance booster
Gen AI can be used to quickly analyze unstructured data to generate actionable insights. The most intuitive application of this in the customer assistance space is to analyze call recordings for comparison of interaction quality against a proprietary knowledge base of a call model. The comparison should include objection management and empathetic approaches, among other measurements.
With minimal development or integration effort, this capability allows institutions to improve strategy and performance management by applying insights from specific calls. It can be used to improve coaching conversations by automating part of the process through self-guided dashboards, suggestions, and training programs. Gen AI algorithms can also identify patterns and use them to help leaders rethink their institution’s existing strategy and call-model approach.
A consumer finance institution deployed gen AI to improve the effectiveness of its frontline customer assistance workforce. It was able to quickly identify the specific call model elements that helped keep arrangements intact, all with limited model fine-tuning. The company also used this information to create a 360-degree, personalized, digital performance management dashboard. The dashboard included call-level feedback for supervisors to use when providing coaching and personalized training, leading to a 10 percent improvement in performance.
Similarly, a major European credit manager company used the gen AI capability of natural language processing with traditional ML techniques to help identify collateral and match it to accounts. They also created a personalized digital performance-management dashboard with call-level feedback for supervisors to provide coaching and personalize training, leading to a 10 percent increase in payments.
Gen AI as a live copilot: Expanding frontline reach with real-time integration
Gen AI can serve as a copilot to boost the performance of agents in real time throughout customer conversations (exhibit). This enables a better overall customer experience through more structured and targeted interactions that focus on what matters to the customer.
In early versions of this deployment, agents can ask a chat interface to provide a summary of previous interactions with a customer, how to respond to a specific question, and if a specific product or discount is available to an account. More advanced deployments can be integrated into telephone calls or other electronic discussions to suggest actions, products, or approaches to the agent during the evolving conversation. They can also include automatically identifying if a conversation is going outside policy, gauging quality control, and triggering the intervention of a supervisor to prevent a negative customer experience before it escalates.
For chat-based interactions, gen AI can prepopulate suggested responses for customer replies, with agents editing as needed, thus increasing the efficiency of the interaction. These conversational responses can be personalized based on customer profile, previous interactions, and current exchange to enhance customer experience and the likelihood of a positive outcome.
An implementation of this use case by a bank resulted in an estimated agent productivity increase of up to 14 percent. Using gen AI as a copilot enabled agents to handle more interactions and spend less time on research and typing. We project that average handling time could be reduced by 10 percent by providing personalized and empathetic responses, resulting in less time spent on customer service. Collection agents using this capability are also likely to have more successful debt or restructuring negotiations, leading to a 6 percent increase in recoveries.
In a simpler copilot implementation, a large bank in the United Kingdom is training existing LLMs with regulatory documentation and internal policies to provide a chatbot interface. Frontline agents will soon use it to quickly navigate product eligibility and compliance guidelines, greatly enhancing customer experience and call quality metrics. It’s a step up from architecture originally developed for anti-money-laundering and know-your-customer rules.
Gen AI as a customer-facing virtual agent: Bringing full power of automation
Gen AI is already being used across industries to improve customer interactions, from restaurant drive-throughs to customer authentication in call centers. In the customer assistance space, players are looking at elements in the journey that could be automated with virtual agents to create 24/7, empathetic support to customers and free up time for real-life agents to focus on the cases that need the most attention.
The technology offers a huge benefit in efficiency. Frontline agents often spend excess time on process-heavy customer interactions, such as authenticating customers and finalizing payments that weren’t completed because of technical issues. Additionally, many customers hesitate or feel uncomfortable when speaking about their financial distress to someone on the phone. Others might need to have discussions outside typical business hours.
Gen AI can alleviate much of the friction by using traditional, script-based chatbots and IVR that provide a human-like interaction experience that is both empathetic and personalized. This technology can also be integrated with existing systems to search for and provide responses to customer questions and suggest specific arrangements in real time. When the technology is stumped, it can automatically escalate to a human agent.
A utility company is currently migrating several use cases of its call center, including authenticating customers and solving specific billing issues, to a gen-AI-powered virtual agent. In this migration, the company aims to handle more than 45 percent of its inbound volumes through the new virtual agent at a fraction of the cost of customer representatives, who could then devote more time to more nuanced cases or other tasks.
Credit customer care can lead an institution’s gen AI journey
The impact and benefits of implementing gen AI in the customer assistance and collection space are already being realized by fast adopters across the world. While short-term benefits can be captured immediately on specific use cases, a structured road map is necessary to capture the most value, minimize risks, and make the most out of cross-organizational investment for long-term success.
By building a scalable gen AI capability in the credit customer assistance space and coordinating with other functional areas of the organization, institutions can combine the power of data, automation, and human capital into collections that keep customers and improve finances.
The adoption of this new technology in customer assistance shouldn’t be seen only as a way to quickly realize value and fund the broader adoption of the new tools. It’s also a way to pressure-test an organization’s capabilities and technical infrastructure needed to scale.
Integrating gen AI can improve the level of support provided to customers in financial distress in a way that can benefit everyone’s bottom line.
