Four Key Considerations for Adopting AI for Debt Collection
April 14, 2021 • 7 minute read

Four Key Considerations for Adopting AI for Debt Collection

Automation and artificial intelligence (AI) are making significant changes to the way the financial services industry handles consumer credit. From consumer lending to debt collection, machine learning is increasing the profitability of managing receivables, through streamlined services and data-driven insights that maximize returns on investment, while simultaneously improving customer experience. But navigating the myriad of opportunities to improve the way companies interact with customers is complex. Constantly evolving regulations and best practices can make it challenging for companies to decide on the best solutions to improve their businesses.

In this post, I’d like to share how currently available AI solutions are transforming the industry along these lines. But before doing so, I’d like to explain why my research team and I have been working in the credit & collections space for the past few years. Most customers don’t plan to be in debt, but instead are affected by circumstances outside of their immediate control. Examples include, unexpected expenses such as medical bills, expensive repairs, or the loss of a loved one. Unanticipated life events such as loss of a job, mental health issues, caring for family members can also put someone past due on balances. And sometimes it can be just a lack of experience with financial planning.

Likewise, collection agents have the unique challenge of bearing the emotional burden of collection calls. Given the emotional energy required, it’s not surprising to hear statistics that report 50% of new collection agents resign in under three months. As a researcher, finding unique solutions to debt collection presents the opportunity to not only improve call center efficiency, but also to alleviate the emotional strain of the work, both for the customer and the agent. And that’s why I want to share my insights with you today.

With that in mind, as collection agencies seek innovation, they should always keep the following customer needs in mind:

    • Empathy: Few people want to be in the situation of owing money. The agent on the other side of the communication should understand that the debtor is going through a hard time and they want to figure out how to get out of the mess.
    • Simplicity: Sometimes a virtual agent is the fastest and most accurate solution to resolve the customer’s problems. For complex requests, sometimes the easiest solution is a simple conversation with a human. Every minute on the phone is a reminder that the problem is not resolved.
    • Privacy: Debt is a sensitive matter. If at all possible, debtors want to resolve their debt discreetly so as not to harm their reputation, whether that be with their immediate family or within their social circles.
    • Flexibility: When a customer says they can’t pay their debt right now, they’re generally telling the truth. Sometimes they need options to pay their bills on a schedule that matches their circumstances.

How do these needs intersect with the operational metrics, such as conversion rates, cost reduction and brand reputation? When considering debt collection solutions that use AI, I recommend considering four key considerations: trust, empathy, compliance, and productivity.

1) Trust

Virtually every credit and collections conversation involves the gathering of personally identifiable information, including account ID, social security number, and payment information. Whether a live agent or a virtual agent is responsible for collecting the information, it is essential to protect that information from unauthorized access. 

The Interactions Virtual Collection Agent (VCA) can capture sensitive information without involving call center agents, whether it be through digital or voice channels. Never solicit customers to provide sensitive personal information over SMS or an insecure chat panel. During live agent conversations, some technologies redirect customers to a secure channel to submit sensitive data, while newer AI technologies capture the information in real-time while masking the information from the agent.

Prior to exchanging sensitive information, it’s essential to ensure that the [virtual] agent is speaking with the right party. Virtual agents follow a consistent set of steps, as customized with the agency, to confirm the caller is the correct party and if not, can work with the caller to identify a better way to contact the actual customer. Innovations such as voice biometrics ensure that the speaker’s voice and speaking patterns match the information of the customer on file. 

2) Empathy

As mentioned earlier, any solution, whether it be AI or human powered needs to be designed to maximize empathy. Virtual assistants are never judgmental, by design. As a result, a properly designed conversational flow can service most customer requests in a manner that doesn’t leave customers having to explain themselves. Using intent detection, our VCA understands what the customer is asking for using a combination of voice recognition and natural language understanding. Paired with emotion detection, AI technology can detect emotions such as frustration, anger and can inform a dialog manager to select a response that’s appropriate for the situation. Speech generation and text to speech synthesis can enrich this by choosing an empathetic voice and generating a response that includes the customer’s name. For example, if the customer says, “I can’t pay because my paycheck hasn’t arrived yet,” an advanced system could detect the reason for nonpayment and respond, “I’m sorry to hear that, John. Let’s explore other payment options.” and proceed with a payment arrangement.

For live conversations, new collection agents can be supported by real-time conversation analytics that suggests the next best action, based on the events in the call. Post-call summarization and analytics can describe the actions in the call and assess agent performance to identify opportunities for coaching.

3) Compliance

As AI and automation continue to reshape the accounts receivable management industry, a new wave of regulations have been released by the Consumer Finance Protection Bureau, such as the recent amendments to Regulation F to prevent false or misleading representations or harassment by outbound dialers. A well-designed virtual assistant can help agencies automate call flow and reduce errors made by collection agents. A Virtual Collection Agent is incapable of deviating from business rules and scripts. In live conversations, AI can monitor agent speech to make sure that required disclosures are provided before proceeding with debt collection.

4) Productivity

As the main measure of operational success, virtual assistants eliminate caller wait and hold time by servicing the majority of inbound requests. Outbound collection campaigns can be scaled by enabling a Virtual Collection Agent to handle tasks like identifying the correct party, freeing your live agents to handle any complex debt collection tasks requiring human intervention.

While virtual assistants can handle most payment tasks including, payment arrangements, it’s important to ensure that the virtual assistant knows when to escalate to a live agent. Effectively every AI prediction comes with a confidence score that can be used to decide whether to trust its decision. Likewise, emotion detection can help identify where customers may be getting stuck in the dialogue and may need help from a live agent.

And a fifth… Fairness. 

While most agencies may stop with the four considerations above, fairness is another challenge. Some customers may speak a primary language aside from English and requiring other language options. Most AI call center solutions in the United States have mature applications for Spanish and French. In other cases, English accents and noise can affect the quality of AI systems for speech understanding. It’s important to evaluate the experience of all customers to make sure that your collections strategy does not leave customers stuck in a virtual assistant that doesn’t understand them, or waiting forever in a hold queue while other customers have their needs met quickly.


Whether your debt collection strategy focuses on outbound and/or inbound calls, email, or digital campaigns, keeping these four key considerations top of mind will help you develop the right blend of AI and human touch to create an amazing experience for your customers while maximizing your profitability.

Want to learn more? Let’s talk.