Artificial Intelligence (AI - do we even really need to show the AI abbreviation at this point?) is likely dominating your morning news feed scroll and infiltrating your inbox with its promise to reshape the way your credit union can engage with members, streamline operations, and unlock new growth opportunities.
The global market for AI was estimated at $46.9 billion in 2020 and is projected to reach $341.4 billion by 2027. With the rapid pace of AI, you might be considering whether to jump in or pump the brakes.
In this blog, we'll clarify what AI and machine learning are (we see a lot of confusion on the differences), dive into AI's current applications for collections and shed light on the benefits and risks for consideration.
What is the Difference Between Machine Learning and AI?
While AI and machine learning are related and often used interchangeably it’s worth breaking down what each mean.
AI, in the simplest explanation, is the science of creating machines that can imitate human intelligence and behavior. These tasks may include problem-solving, learning, reasoning, understanding natural language, recognizing patterns, and adapting to new situations. AI encompasses a wide range of technologies and approaches to simulate human intelligence in machines.
Machine Learning is a subset of AI and refers to the practice of enabling machines to learn from data and improve their performance over time without being explicitly programmed. In other words, machine learning algorithms use data to identify patterns and make predictions or decisions instead of following strict instructions. Machine learning is trained on large datasets, and their performance improves as they receive more data and feedback.
Machine learning can be applied to create actions based on customer behavior patterns within recovery and collections processes. By analyzing historical data, like payment patterns, communication preferences, and past interactions, these models can identify and recommend a suitable collections approach for different customer segments.
AI incorporates various technologies, such as machine learning, natural language processing, chatbots, predictive analytics, and others to enhance its capabilities.
Tools like ChatGPT, which process natural language, use machine learning to build large language models (LLM). By training on massive amounts of human knowledge to become experts, they can answer questions by predicting what words should come next.
Applications of AI in Collections
Chatbots are computer programs that mimic elements of human conversation and are used as one communication channel.
With the goal of delivering a cost-effective alternative to human customer service, it’s one of the main reasons they are quickly gaining popularity across various industries. Approximately 37% of the U.S. population is estimated to have interacted with a bank’s chatbot in 2022, a figure that is projected to grow.
There are two applications for chatbots. The first is rule-based, where the chatbot will serve specific commands and use programming to respond to already established and mapped out alternatives.
The second application is AI-powered chatbots, which emulate human intelligence by predicting the appropriate response based on the data it ingested and its training parameters. They can handle routine customer interactions without human intervention, such as payment reminders and inquiries.
Ex. A member asks how they can make a payment in full on a late bill, and suddenly they pivot and say I can only afford to make a partial payment. With Conversational AI, intent is recognized, and the chatbot can shift gears.
Conversational AI is designed to assist users on any communication channel - spoken or written. It’s the tool used to build the AI-driven chatbots we discussed above.
Conversational AI can have a multi-turn dialogue, ask further questions, recognize the intent, and compensate when a member goes off-topic.
Ex: Conversational voice AI can handle automated phone calls to late accounts, delivering payment reminders, past-due notifications, and personalized messages. These voice interactions simulate human-like conversations.
IVR systems can also be powered by conversational voice AI where a member can interact with the agent using natural language. Members can make payments, inquire about their accounts, and access self-service options through voice commands.
Predictive analytics uses advanced data analysis techniques, statistical algorithms, and machine learning models to analyze historical data and predict future events, trends, or outcomes. It involves identifying patterns and relationships within the data to forecast potential outcomes.
AI can predict which members are more likely to default on payments, enabling early interventions and targeted outreach. Predictive models can help identify high-risk accounts and prioritize debt collection efforts for maximum efficiency.
Understanding the Benefits and Risks of AI
The financial services landscape is competitive, and AI is helping transform the level of service consumers now expect - with ease and convenience at the forefront. Incorporating AI-driven systems in your collections strategy can help credit unions uncover new ways to increase efficiency, reduce costs and enhance decision-making.
AI-driven processes can handle a large volume of member interactions and routine tasks, significantly reducing agents' workload. This reduction in workload can lead to a decrease in full-time equivalent (FTE) headcount and allows agents to focus on more complex cases.
AI offers the advantage of reducing manual labor and the need for a larger FTE headcount. A smaller team, and more efficient resource allocation and collection efforts, significantly reduce operational expenses.
AI-powered predictive analytics can assess historical data and customer behavior patterns to predict which accounts are more likely to default on payments. This enables early intervention and targeted outreach, helping to prioritize debt collection efforts for maximum effectiveness.
What are the Risks of AI in Collections
Without question, the potential for AI is massive, but so is AI’s potential for misuse. Any credit union looking to implement AI must also be aware of the inherent risks and pitfalls when dealing with high-stake applications (members’ financial well-being and sensitive information).
The most considerable discussion around the risk of AI is that we still don’t know where AI is learning the information from - this is referred to as the “black box problem.”
The term “black box” is a metaphor used to describe a situation where the internal workings of a system are hidden from the outside observer. In the context of AI, it becomes challenging for humans to trace the step-by-step process of how AI arrives at its decisions or predictions - which can be problematic in financial applications.
Nabil Manji, head of crypto and Web3 at Worldpay by FIS, said a key thing to understand about AI products is that the strength of the technology depends a lot on the source material used to train it.
Manji describes the challenges for financial services as a lot of the back-end data systems are fragmented in different languages and formats.
“None of it is consolidated or harmonized,” he added. “That is going to cause AI-driven products to be a lot less effective in financial services than it might be in other verticals or other companies where they have uniformity and more modern systems or access to data.”
Additionally, in a recent CFPB spotlight article called “Chatbots in Consumer Finance”, the CFPB highlighted that when chatbots fail in the markets for consumer financial products and services, they not only break customer trust, but the stakes for being wrong when a person’s financial stability is at risk are high.
Security Risks and Privacy Concerns
AI systems require access to sensitive financial data and personal member information to make decisions. If not adequately secured, these systems can become targets for cyberattacks, leading to data breaches and compromised member data.
Lack of Human Judgment and Empathy
Making members feel valued and understood, fostering strong relationships, and a sense of community makes credit unions stand out. With an AI-driven collections process, members may lack the human touch and emotional intelligence needed in a time of financial trouble. They could find it challenging to connect with AI-driven interactions, leading to potential misunderstandings that can negatively impact customer experience.
Legal and Regulatory Compliance
The financial services industry is heavily regulated. AI in collections must align with existing regulations and guidelines governing debt collection practices to avoid potential legal complications. Internal leaders must ensure AI-driven actions and decisions comply with relevant financial, privacy, and consumer protection laws.
Bias and Discrimination
Because AI models are trained with an unknown dataset, the AI systems are only as good as the data we put into them. AI algorithms can inadvertently inherent biases in the data used for training. If the training data is biased, it can create an ongoing cycle of bias that can lead to discriminatory or unfair outcomes in collection decisions that impact disparate groups.
The bias in language models can also arise from humans who assess and validate the model’s responses. This validation process presents an opportunity for bias to be introduced into the model’s outputs.
Staffing, Recruitment and Skills
Consider whether your credit union has the right in-house skills to understand, use and appropriately supervise the AI solutions adopted. You will likely need additional skilled professionals to help test and manage these AI applications, and with a current talent scarcity this can become a recruiting challenge.
What to Consider Next?
Many CIO’s and IT executives are speaking out to make organizations aware of the potential risks and benefits, explaining the importance of working with experts in the field to develop responsible strategies for its use.
As the market grows, eventually, AI tools will become more heavily regulated by policymakers and industry stakeholders, with stricter rules that promote transparency, accountability, and trust in AI systems.
Until then, ask the right questions, lean on industry experts, and always start with establishing what problem you’re trying to solve and put your customers at the centre of that discussion. Doing so will ensure that the AI tools you evaluate align to improve member experience and satisfaction.
If your goal is to improve efficiency, productivity and recovery rates, while still delivering a personalized experience, consider leaning on automation collections software that still requires a level of human input that will give you control of your member experience. Get in touch with us if you need some guidance.