Those working in debt collection and contact centres are discovering how AI is a powerful partner in customer engagement and vulnerability detection.

Mark Opperman, Chief Revenue & Marketing Officer at Webio, unpacks how this works as he reveals the inner-workings of conversational AI platforms for digital customer communication. 

[Duration - 6:22]

Mark Oppermann gave this presentation at the Collections & Vulnerability summit in Manchester.

The Transcript

So let's crack on. AI. There's been lots of talk about AI over the last while, and I'm sure everybody has heard about ChatGPT and ChatGPT being a large language model. It's been developed over the last many, many years. There's probably billions already spent on it, so these things don't happen overnight. So it's a general AI system.

Hands up. Who has played on ChatGPT so far? Yeah, lots good. So everybody and it really has put the the fizz back into people talking about what can AI do for you?

Well, we've been working at this for the past about four years, and rather than a large language model, ours is very much customised on collections, so using the millions and millions of collections conversations that we manage for our clients every day, training that model to be applicable for the collections world.

The language that is used today in industries from aviation to farming to collections - there're similar words used, but they mean completely different things. So what does that mean to the business? It's finely tuned to make sure that it's actually applicable and usable in the world our industry today. So, I'm going to take a practical approach about how AI is being used because sometimes when people talk about AI it goes over everybody's heads. 

There're only three things I'm going to talk about today.

Entitities

Entities gather and keep the information that your customers are going to give you during a conversation. In digital conversations today, it has to be very structured. You ask for one piece of information and if it doesn't come in exactly the right format, it's going to fail.

If you're having a conversation with a customer and they say, "Look, I don't get paid till this Friday. I'll pay you next Monday," in any traditional digital conversation that would be a complete fail as the chatbot wouldn't be able to work out what 'this Friday' or 'next Monday' are referring to.

With Entities working, it will work out what date that is comparinf to Today and if that is acceptable. It's bringing those key bits of information that you are getting from people into a format that's really practical and easy to use.

Intents

This is where it starts to get interesting because we're all human, we're all so and so's, aren't we? How people say things will vary, but a lot of the time, it's very similar things we're looking for. So what are customers saying? And how do you understand that using AI?

Propensities

The last thing that I'm going to focus in on is Propensities. So what are Propensities? It's understanding from the conversation that's happening what's the most likely outcome. Very simple, very straightforward. So as the conversation is happening, is that person exhibiting certain traits? For example 'vulnerability' is more likely going to end up with the conversation going to a specific end which requires certain actions.

At the start of a conversation, if you knew where this conversation was going to go to, how would you handle it? It's about being able to design the right type of conversation flow to suit both the customer and the business, putting a huge amount of power into your hands, where everything is auditable, it's explainable AI.

I'm going keep it really practical and just show two examples.

The first one is an entity. So no matter which way somebody is giving you the key bits of information, it will go back and it's going to be able to understand these bits.  Almost 60% of the conversations flow through and they don't fail and they don't hand over to an agent. And again, it just speeds up everything, which is a better experience the customer.

The second one is Propensities. As I said, as the conversation is happening, what it's doing is it's identifying the most likely outcomes that are going to be taking place in the conversation.

I've just taken four examples there. The biggest one by far is 'Vulnerable'. Really, that's where most of our clients are seeing a real benefit here. It's identifying that early, tipping between about 85% to 90% accurate conversation. So it's enabling them to identify vulnerabilities and then get customers the right level of support. It boils down to designing better conversations.

We're on the stand, so please come along and say Hi to us. We also work with Policy in Practise, as again one of the things that we're integrating into the Webio platform is the Better Off Calculator. So again, I'm happy to talk about that. And thank you, have a good day.