Scaling debt recovery with an AI assistant – a success story

Let’s face it, nobody wants to talk to an enforcement agency. People often have a negative perception of debt recovery agents, as if their aim is only to take money from the poor. 

Sadly, many people are affected by debt. Debt can be overpowering. Some people get into a deep debt spiral, with no idea how to get out, and that affects their mental health. The thought of a conversation with enforcement agents can cause a great deal of anxiety. 

Kieran Norton, a Technical Support Engineer at a leading UK enforcement agency says, “actually what we want to do is try and improve things, get people out of debt and help them stay out of debt as well.” To achieve that, they’re utilising Conversational AI designed, built and managed by VUX World.

Read on to find out about a chatbot that reduced query resolution times to 2 ½ minutes (saving 400 agent hours in under a year) and improved CSAT by 30%. 

If you thought you needed to be a developer to build an AI solution, think again. Learn from Bret Ingerman from Tallahassee State College how it’s benefitting from AI automation and from Cobus Greyling on how you can build your own AI chatbot for your business without the need to code.

On August 1, we're hosting an exclusive webinar to dive into the key benefits of implementing AI chatbots and how you can identify best use cases that work for your business and your customers.

Find out more

Trying times

The debt recovery agency where Kieran works was experiencing an increased workload. The cost-of-living crisis meant that more people were falling into debt, and their live agents were overwhelmed with the volume of calls and chats they received. 

It was in this climate that the agency saw an opportunity to incorporate Conversational AI. Assistants appealed to them because they can scale as the business grows, but also they can handle occasional spikes in the contact centre. If there’s a sudden influx of chats, CAI can handle them and escalate to live agents only when necessary. It would also be available 24/7 for users who need to call before or after work.

Can Conversational AI succeed in debt recovery?

They had concerns though. For them it was vital to make the AI Assistant as secure as their live agents. Only the case-holder should be able to access their case. 

Also, they didn’t want a solution that felt robotic. They wanted their vulnerable customers to know they’re being cared for. It needed to have the company voice and feel like a conversation with one of their live agents.

Add to that the fact that the enforcement industry is strictly controlled by the Ministry of Justice. The results would have to be compliant.

Finally, it had to connect with the various systems they use. It was no good to create a debt recovery chatbot that couldn’t access cases, as that would limit its ability to help.

Knowing where to start

VUX World created the AI Assistant to meet this client’s needs.

We knew we had to start with the customer's needs. While the business saw an opportunity to automate conversations, we knew we shouldn’t start by automating everything. The best place to start is with high volume chats that are easy to resolve.

To get there, VUX World analysed transcripts of past conversations. It became clear that many customers would start a chat to find out how much they should pay, when they should pay, and how they could pay. Customers can make one-off payments and create payment plans, so both needed to be part of the experience. This meant that the assistant’s first use cases would be related to deflecting conversations from live agents and interpreting customer needs.

While those aren’t truly transformational implementations of CAI, they are an excellent place to start. Once an assistant can do those things well, it’s possible to move on to more complex use cases. 

With a clear understanding of what people wanted to discuss, our aim was to find the best ways to automate the journeys that would resolve their needs.

Creating journeys to answer needs

Once we had the use cases we could begin the design process. While designing flows to answer FAQs were simple, helping customers create a payment plan was more complex. For that journey to begin, users first need to be identified and verified. It’s essential that only the case holder accesses the case. They’re asked various questions to identify them (the image below shows a simplified version). 

Once they’ve passed the ID check, they’re told how much they owe, and when they need to pay it by. Users who want a payment plan are given a link to an income and expenditure form, which they need to fill out. 

This is the same process a customer would go through when they ask a live agent to set up a payment plan, but no live agents are needed for these conversations anymore! Plus, the user’s request for a payment plan has been logged on their case details on the line-of business system. A few days later, we automatically check to see if they completed the form. If not, that data is useful to the client who can follow up with the user.

If a user is vulnerable, then that’s flagged after the ID check. Vulnerable customers are escalated to a live agent as their situation requires a more sensitive discussion. When that happens, the agent will know both who they are and what they need, as this is passed across during handover. The conversation can continue from that point. 

This is one reason why query resolution times have been massively reduced. The chatbot is always a customer’s first point of contact. For some use cases, the chatbot can handle everything. If it does need to escalate, the live agent will continue the conversation after the chatbot, so both live agent and customer have already been helped, and everyone’s time is saved.

As anyone who has made an assistant knows, the launch is the beginning. From that point onwards, conversations need to be monitored to prove design assumptions. We continually analyse conversations on a business level, at a customer journey level and at an interaction level, at each turn of every conversation.

Replicate their success

We’re stoked to tell you that this chatbot has worked wonders! 

• The chatbot helped the client to win more business, as it simplifies the logistics and allows them to scale to meet demand more easily
• They have less pressure to hire live agents (the chatbot is equivalent to one full time employee which saves them £27,500 per year)
• The reduced query resolution time of 2 and ½ minutes per conversation has saved 400 hours of agent’s time, which is also time saved for customers
• CSAT has improved by 30% since launch
• The chatbot has moved nearly £1 million to arrangement since launch, which was previously handled by live agents alone
• Finally, it’s won the CIVEA Innovation in Enforcement award!

So how can you replicate that success?

1. Know that this is a long game, and you should start with the best use case(s) - pick those that users frequently need and can be easily automated
2. Get the support of your stakeholders - CAI will transform your business for the better and everyone needs to be onboard
3. Contact VUX World and have us do it for you!

To dive into this case study in more detail, you can rewatch the webinar we recently ran with Kieran, or you can download the case study here. 

About Kane Simms

Kane Simms is the front door to the world of AI-powered customer experience, helping business leaders and teams understand why AI technologies are revolutionising the way businesses operate.

He's a Harvard Business Review-published thought-leader, a LinkedIn 'Top Voice' for both Artificial Intelligence and Customer Experience, who helps executives formulate the future of customer experience ad business automation strategies.

His consultancy, VUX World, helps businesses formulate business improvement strategies, through designing, building and implementing revolutionary products and services built on AI technologies.

1. Subscribe to VUX WORLD newsletter
2. Listen to the VUX World podcast on Apple, Spotify or wherever you get your podcasts
3. Take our free conversational AI maturity assessment