How Conversational AI Boosts Collections as Delinquencies Rise
As defaults increase collections workload, the benefits of conversational AI can prioritize agents' work and lead to faster resolution

One consequence of rising inflation and interest rates is an increase in delinquencies across various types of loans, including credit cards and auto loans. Recently, The Financial Times reported that credit card loan defaults have reached their highest level in the United States since 2010. For lenders, this jump in delinquency cases means more time, cost, and effort to collect. This in turn begs the question – can AI be utilized to make collections processes more efficient, more effective, and less costly? It can. Conversational AI is coming to the forefront as a technology that streamlines processes and encourages customers to self-serve when appropriate.
What’s the Trouble with Collections?
A rise in collections volume will exacerbate a few problems with many lenders’ collections processes. For example, many lenders utilize people and manual processes to handle each delinquency case-by-case. This approach ties up agents, does not take advantage of much if any automation, and becomes more expensive and less efficient as collections volumes increase.
Past approaches to automation have also struggled to facilitate better customer experiences as intended. In complex processes like collections, customers are typically presented with a limited number of pre-defined options. The format demands that the list of options is limited in number and complexity – there is no ability to capture nuance, and selecting from lists only serves the most straightforward of cases.
Virtual agent technologies have also reached their limits in many cases. For example, the initial wave of virtual agents communicated one-way, primarily functioning as information broadcasters to deliver messages, announcements or alerts to users. They lack the capability to receive or act on customer requests and responses to prompts.
Two-way virtual agents, which can manage more interactions with customers, also tend to be limited to pre-scripted responses that may not fit – or be able to ascertain if they fit – a customer’s specific needs. These agents tend to shine in presenting FAQ responses and automating simple steps, like running checks that determine if a transaction could be fraudulent.
Unlike traditional chat bots, conversational AI takes automation and virtual agent capabilities to a new level.
Conversational AI and Automation
According to Nvidia’s glossary of AI terms, Conversational Artificial Intelligence (Conversational AI) is “a complex form of artificial intelligence” which allows for “human-like interactions between computers and people,” by being able to “recognize speech and text” – including local phrases and slang – and to “understand intent…and respond in appropriate natural language.”
As a result, a virtual agent or chatbot equipped with conversational AI can interact fluidly with customers via their preferred channels, like text messaging or messenger apps like WhatsApp. It is not limited, like previous generations of virtual agents, to pre-set inputs, static responses, or specific channels such as the lender’s web portal.
Instead, virtual agents equipped with conversational AI can:
- Understand, capture, and process a customer’s words. In a chat a customer may share a story filled with pertinent information. With conversational AI, the chatbot can understand and analyze this content, and potentially feed into the automated process.
- Parse key information captured. Customers’ utterances – the things they say which contain information pertinent to the process - are parsed into intents and slots. Intents map to actions a customer can take, like making a payment or opening an account. Slots are data that supports those actions, like names, account numbers, and date ranges like “tomorrow” or “next week.”
- Automate more complex scenarios for better customer experiences. Conversational AI allows for more complex scenarios to be addressed with better customer experiences through automation. It can reduce the time needed to resolve complex cases like fraud, loan originations, and collections by minimizing data collection steps and advancing processes to fulfillment or far enough along to minimize an agent’s hands-on effort.
Conversational AI vs GenAI
Conversational AI differs from popular GenAI tools that many people have encountered. While both types of AI can understand natural language utterances from users and respond to them automatically, conversational AI responds with pre-defined automations.
Conversational AI’s automations, or intents, represent specific, pre-defined processes the lender defines ahead of time and governs over time. Responses and automation are only initiated when a customer’s words – or utterances – can be mapped to specific, pre-defined intents that trigger actions like pay now, pay later, or pay on a specific date. Conversational AI provides the ability to leverage sophisticated AI capabilities in a way that eliminates the risk of hallucinations or improvisations that, for now, may still occur with GenAI.
How Can Conversational AI Improve Collections?
Because conversational AI has these advantages, it can improve business-critical and complex processes like collections in terms of time, cost, efficiency and customer experience. Collections provides a good first example of how conversational AI makes improvements because it can be broken down into a series of intents which can address most of what a delinquent customer might need.
Here are six examples of intents that can automate and facilitate collections with conversational AI:
Negotiation
Rather than live agents navigating the emotional aspects of negotiating terms with customers, conversational AI can do it in a less personal manner. Factors like payment frequency, amount due and payment date may be negotiated by the chatbot. The conversational AI can offer several options to customers, each acceptable to the lender, and then map the customer’s choice back to automated processes that fulfill it.
Pay now or later
If a delinquent customer wants to pay now, the virtual agent ensures the customer is presented with the correct automated payment process and makes their payment. A lender can also configure more payment options for customers, for example: paying a week from now, paying in part now and in part later, or agreeing with the customers on the number of days acceptable to both parties.
Hardship
Every customer has a backstory, and it may relate to hardship when a good customer suddenly struggles to pay. A customer’s utterances may be filled with apologies and explanations about losing a job or a sudden health expense preventing payment. Conversational AI can identify this information and trigger compassionate scripts and automations that give the customer flexible options that are acceptable to the lender.
Speak with an agent
Though many aspects of collections can be automated with no agent intervention, some customers will still want to speak with someone. The trick with conversational AI is keeping the conversation going to capture as many utterances as possible and feed this as data into the process before transferring to an agent. As a result, any escalation to an agent is done with the resolution process already in motion and without asking customers to repeat themselves.
Already paid
The collections process also involves customers who have already paid, or claim to, which means a payment may need to be tracked down and credited to the account. Conversational AI can walk through a dialogue with a customer to track down when and how payment was made to resolve any disconnects while also detecting fraud at the same time.
Fallback
Of course, people and machines are not flawless. Sometimes customer utterances won’t make sense to the AI. In such cases, however, Conversational AI does not improvise. It simply starts over to try and ascertain customer needs by analyzing utterances for slots and intents. And if it cannot, a live human agent can be notified to step in.
Because delinquencies have increased dramatically, financial institutions are engaged in more collections-related efforts that translate into higher costs and often uncertain or negative customer experiences. By automating collections processes with Conversational AI, lenders can elevate their ability to identify and support customers who are experiencing financial hardship. They can extend better customer experiences that give customers a sense of control and relief. And they can make it easy to automate payments and enable customers to negotiate what they owe and when.
How Conversational AI Works Within FICO’s Omni-Channel Engagement Capabilities
- Find out more about FICO’s omni-channel engagement in FICO Platform
- Discover FICO’s solutions for collections
- What should collections operations do next to improve – view the Expert Spotlight
- Read the blog: How compassionate digital collections can improve your bottom line
Popular Posts

Business and IT Alignment is Critical to Your AI Success
These are the five pillars that can unite business and IT goals and convert artificial intelligence into measurable value — fast
Read more
Average U.S. FICO Score at 717 as More Consumers Face Financial Headwinds
Outlier or Start of a New Credit Score Trend?
Read more
FICO® Score 10 T Decisively Beats VantageScore 4.0 on Predictability
An analysis by FICO data scientists has found that FICO Score 10 T significantly outperforms VantageScore 4.0 in mortgage origination predictive power.
Read moreTake the next step
Connect with FICO for answers to all your product and solution questions. Interested in becoming a business partner? Contact us to learn more. We look forward to hearing from you.