The credit and collections industry has traditionally been slow to embrace new technology, however, conversational AI is a perfect fit for contact centres that focus on customer engagement for debt collection. 

Debt collection teams that have taken AI chatbots onboard tell of the difference AI and automation make. They quote the improved metrics such as response and collection rates, as well as the positive customer experiences like no long call waiting times and the 24/7 access. 

However, there are myths about conversational AI that may be holding businesses back from adopting it.  This article takes a look at these and shows why fears around using AI for credit and collections are often unfounded. 
 

Myth #1: Conversational AI chatbots can’t be relational and empathetic 

Customers don’t care that they're talking with an AI chatbot as long as they get their query resolved quickly. 

However, it really does matter how a chatbot uses language. Even though customers know they are communicating with a bot, how it phrases requests and answers makes a world of difference to how the customer responds. Companies need to let their customers know that they are heard and that they are trying to address their concerns. 

And yes, AI chatbots can be designed to speak respectfully and empathetically and give an impression of a ‘human touch’. The days of the early command-based chatbots are over. Instead, AI chatbots have become conversational, that is, they conduct two-way conversations with customers using natural-sounding language.   

By using Natural Language Processing, intent recognition and vulnerability scoring technology, an AI chatbot can pick up on the state of the person they’re speaking to and guide the conversation down the best route.  

For more details, see: How Does Conversational AI Work? 

By integrating with backend systems like CRMs, the AI chatbots have access to customer data that informs the conversation and makes it personalised and useful. 
 

Myth #2: People don’t like talking to AI chatbots about sensitive matters like debt 

In fact, the opposite is often true. Considering that 81% of Millennials have phone anxiety (bankmycell.com survey), it’s no surprise that speaking with a chatbot via a digital messaging channel is preferred. Customers choose texting over phone calls as they can think about what to say, can respond in their own time and it’s more private.  

But what is interesting for debt-related matters is that when interacting with a chatbot they feel less judged. Debt is an emotional matter and speaking about it with a human can be embarrassing, especially if there are others around who can hear you talking. 

Myth #3: AI cannot understand the nuances of debt collection 

AI that is designed for specific industries like debt collection can be trained on focused data to be high-performing tools for customer engagement. They learn the language of that industry and individual company, as well as the compliance and regulations that govern them.  

This laser-sharp focused training keeps the AI chatbots fully under the control of the company and prevents the hallucinations (fabricated facts) that occur with general chatbots like ChatGPT. 

For more details, see: Give AI Chatbots the Edge with Custom Trained Language Models 
 

Myth #4: AI chatbots are not good at understanding humans 

 
Firstly, what do we mean by ‘understand’? AI 'understands' by receiving inputs, analysing them using Natural Language Processing (for more details: What is Natural Language Processing) and then it produces an appropriate response. AI chatbots are trained on large data sets (What is a Large language Model), and for conversational AI they are built to engage in two-way conversations with customers that make sense, are helpful, are accurate and sound natural.  

For example, the conversational AI and automation technology of Webio is built on three layers:  

Intent Recognition – what a person is saying or wants to do 

Entity Gathering – picking out key piece of useful information from a larger unstructured text, e.g. dates 

Propensity – Guiding the conversation down the best route based on what a person reveals in their texts 
 
With these elements of AI working together, the chatbot understands and responds in a conversational manner. AI chatbot technology has come a long way and contact centres are seeing automation levels of 75% where the bot takes care of the common queries and tasks, such as ID&V, account balance checks and payments.  

For more details, see: The Three Layers of AI Automation   


 

Of course, no chatbot is perfect, and there will always be times when a chatbot misunderstands a user. This is seen as a learning opportunity and the data is fed back into the system where Machine Learning techniques are used to growth the bot. 

Myth #5: Conversational AI chatbots can replace human agents entirely 

Although conversational AI chatbots for collections are becoming ever more sophisticated and proficient at customer engagement, there will always be a need to have a human agent at hand. 

When designing a chatbot for customer service you need to have ‘off-ramps’ available at every stage where customers can leave the chatbot conversation and talk to a human. Some customers just like speaking to an agent, or their issue is at a higher level and needs a human to step in and take care of it. 

This blended approach works well: AI takes care of the routine, common and straightforward tasks while the human agents deal with the more involved queries.   

Myth #6: Conversational AI lacks compliance and security 

Generalist AI chatbots like ChatGPT do not have specific regulations built into them. On the other hand, AI collections chatbots are carefully designed with guardrails in place to keep charge over how the chatbot performs and the chatbot conversation designers control what the bot can or cannot say. 

These guardrails come from having the industry regulations coded in, such as GDPR and Consumer Duty, to keep AI chatbots compliant.  

Also, centralised record-keeping and audit trails are part of the AI communications platform which provides checks-and-balances and transparency all along the customer journey. 

See: Debt Collection Regulations & Compliance: What you Need to Know 

Myth #7: AI is biased, inaccurate and unreliable 

It is true that some chatbots have shown biased tendencies and have produced inaccurate responses. We think of Microsoft’s Tay chatbot in 2016 and it’s disastrous and brief existence. Soon after launch, Tay started sprouting prejudiced remarks. Or the early recruitment AI apps that showed stereotypical gender bias. And the likes of ChatGPT that hallucinate (fabricate facts) 15%-20% of the time. 

However, if an AI chatbot is built for a specific task, such as debt collection customer engagement, the bot it is trained on focused and accurate industry-specific data - Custom Language Models. It is then continuously refined using Machine Learning technology. 

In the end, the better the data, the better the AI chatbots performs. And the AI needs to be explainable and transparent - no ‘black boxes’. 

Conclusion 

One of the common myths about conversational AI is that chatbots are some kind of silver bullet that can solve all customer engagement problems. This is not an accurate picture. Conversational AI is indeed a powerful tool, and like any tool, it can be used effectively or poorly. A well-designed and implemented conversational AI chatbot is a valuable asset for debt collectors and it can turn around a company’s debt collection processes and keep customers happy.