Why Conversational AI changes the rules for your business
An interview with Aryza Engage founders Cormac O’Neill and Mark Oppermann
Looking at the credit lifecycle from origination to collections – where do you see the most potential for using Conversational AI?
Cormac O'Neill From my perspective, Conversational AI can be used across the whole journey. From origination, when the loan is granted, right through to collections - anywhere there’s a customer touchpoint, there’s an opportunity to use Conversational AI effectively.
Mark Oppermann Any time you engage with a customer, conversational AI has a role to play. But let’s take a step back and look at where we started.
At the beginning, we didn’t have unlimited budgets. So we concentrated on the area we knew well and had deep expertise in: collections. That’s clearly a later stage in the customer journey. What made collections a smart starting point is how well it’s measured. It’s the lifeblood of a business - cash coming in. When debtors don’t pay on time, you quickly get into bad debt provisioning and extra costs that drag down competitiveness. And you need to have those conversations. That’s why we initially focused on collections, and then expanded into recoveries. As we delivered success in those areas, clients naturally began to ask: where else could we use this?
That’s when things moved back up the journey. Some of the products we developed - like credit web chat - are more aligned with origination or early-stage customer service.
You wouldn’t typically use web chat in collections, but it works well when a customer first engages or has a service query. So even our product development began to reflect that shift.
Could you please go into more detail?
Cormac O'Neill Now that we're part of the Aryza family, we’ve been introduced to the Aryza product wheel. Once you understand it, it’s very clear: from origination to lending, collections, recoveries, and even insolvency, every single stage involves customer interaction. And wherever you engage with customers, Conversational AI can be applied.
Mark Oppermann Web chat is a great example. I was recently on a call with some of our customers in the insolvency space. They were still sending a large volume of physical letters. We showed them what Aryza Engage can do - how you can share attachments and send the exact same letter digitally that you’d normally post. With web chat, you’re engaging in real time. So instead of waiting for a letter to arrive and then following up by phone, the customer gets it instantly. That’s a huge shift in both experience and cost.
Instead of waiting for a letter to arrive and then following up by phone, the customer gets it instantly. That’s a huge shift in both experience and cost.
Cormac O'Neill There was a real lightbulb moment: “Hang on - we can stop sending all these letters and just do it through chat?” When you look at the cost of sending a letter in Ireland or the UK, you’re talking about around one pound or €1.20 per letter. Sending it through chat costs just a cent or two.
So switching from post to digital messaging isn’t just more efficient - it also changes how customers engage?
Mark Oppermann Exactly. When you compare the cost and speed, it’s night and day.
Cormac O'Neill Especially when you think about combining the old and new ways of working. Many companies still send physical letters, but now they’re adding QR codes to them. That brings customers straight into digital channels like WhatsApp. So you send the letter as usual, but instead of asking the customer to wait for a follow-up, you include a QR code that brings them into a real-time chat. If you need to send a follow-up letter, it goes directly to their phone. That’s incredibly fast and powerful.
Mark Oppermann And there's another big benefit: you know the customer has received it. In web chat or WhatsApp, you can see when it’s read. With regular post, unless you pay for tracking or registered mail, you don’t have that certainty.
Beyond letters and mailing costs, what other metrics matter - especially in the contact centre?
Cormac O'Neill When we talk to clients, two questions always come up. First: How will this help me reduce costs? Second: How will it help me increase revenue? If a solution can’t answer both, it doesn’t get far. Aryza Engage does both: it cuts the cost of customer contact and it improves collections performance - which, in our clients’ world, is a direct revenue boost.
Mark Oppermann One of the biggest cost pressures - especially after COVID - is staffing phonebased collections. More and more people simply don’t want to have debt conversations on the phone. In my experience, nine out of ten clients say they struggle to recruit and retain agents for those roles.
Cormac O'Neill Digital channels change that equation. With asynchronous chat and high levels of automation, each agent can handle far more conversations, and younger recruits prefer digital work. Companies save on headcount while actually improving customer engagement.
What are the efficiency gains when switching to digital-first customer contact?
Mark Oppermann One of our clients had real difficulties recruiting contact centre staff for collections. So they decided to run a test: set up a fully digital collections team in parallel. That pilot was oversubscribed by 100%.
If we generalise a little: contact centre staff tend to be younger. And for many of them, phone conversations feel unnatural, whereas digital communication is second nature. So not only was recruitment easier, but they also managed to retain staff for longer.
That’s one of the downstream benefits of digital channels: better employee experience and stronger operational efficiency. Just look at productivity. Across our clients, a phone-based agent typically handles between 50 and 70 conversations a day.
Now take that same agent and move them to Engage, powered by Aryza’s Conversational AI. Suddenly they’re handling between 380 and 420 conversations per day. That’s possible because more than 85% of the work is automated. And because digital chats are asynchronous, agents don’t need to respond immediately - they can manage multiple conversations at once.
The result? A massive gain in efficiency for the business, and a better experience for the customer. Most people don’t want to talk on the phone about debt. They prefer to deal with it privately and on their own terms through digital channels.
Does the AI handle most conversations on its own, or do customers often ask to speak with an agent?
Cormac O'Neill You could say that roughly every seventh or eighth customer interaction ends with the person saying, “Okay, now I need to speak to an agent.” So yes, agent escalation still happens - but the key question is: are the majority of customer queries being successfully resolved? And the answer is yes. When you combine human agents with AI, you get the best of both. As Mark said earlier, AI can handle a huge number of common queries. In fact, we find that Conversational AI can resolve around 40% of inbound queries straight away.
In credit and collections, these tend to be things like:
What’s my balance?
How much do I owe?
When is my next payment due?
Those are all solvable with automation.
And when you integrate Engage into your system of record or CRM, you can provide that data instantly - no waiting, no phone queues. The customer gets their answer right away, after a quick ID&V check to confirm their identity. That’s the power of having AI embedded in your process.
So automation handles routine queries - but what about the more complex cases?
Cormac O'Neill That’s exactly where the human team comes in. AI can save customers from having to wait 45 minutes on the phone to get simple answers. And by doing that, it frees up your contact centre team to deal with more complex issues - people who want to reschedule a loan, who’ve had a change in circumstances, or who are facing personal difficulties.
These are the harder conversations - the ones that require empathy, flexibility and experience. And that’s where trained agents make a real difference.
Mark Oppermann It’s all about balance. Some organisations rush to automate everything, but that’s not the right approach. It’s about finding the blend - where automation delivers real value, and where human expertise is essential.
You’ve already invested in your team. You’ve trained them, developed their skills - so it doesn’t make sense to sideline them. We’ve even seen it in the media lately - companies that went all-in on automation are now stepping back, admitting that it’s not working as well as they hoped. You can’t replace everything with AI. You need both.
Cormac O'Neill And here’s something really interesting: digital channels often open up conversations that simply don’t happen over the phone.
One of our clients looked into their data and found that customers who wouldn’t answer a phone call were much more likely to respond over a digital channel.
So digital channels are not just more efficient, they also improve engagement?
Cormac O'Neill One client analysed their data and found something remarkable: a whole group of customers who had never answered the phone - no matter how often they were called - suddenly engaged when approached via digital channels. In fact, 75% of those previously unresponsive customers replied in a digital conversation.
And when you think about it, it makes perfect sense. If someone isn’t comfortable on the phone, calling them once or ten times won’t make a difference. They simply won’t answer. Now imagine being in their shoes - someone calling to discuss money you don’t have. The natural instinct is to avoid that situation.
Mark Oppermann If you give people time to think - maybe they sit down with a cup of tea, read the message, reflect, and then respond - you create space. That’s why we see much higher engagement rates across digital channels compared to the phone.
And there’s another important point: the quality of digital conversations. You might assume a live agent would have the better interaction, but that’s not always the case.
A good example is Promises-to-Pay. Across nearly all our clients, we see a consistent pattern: around 10% more of the promises made via digital channels (within a 14-day period) are actually fulfilled, compared to promises made over the phone. Why? Because on a call, people may just say whatever it takes to end the conversation. But with digital, they have time to consider their answer.
Cormac O'Neill That doesn’t mean there’s no place for agents, of course. In cases where someone shows signs of vulnerability, that conversation should always be handed off to a properly trained human. That’s where blending automation and empathy really matters.
So the key is matching each customer’s intent to the right response?
Cormac O'Neill Exactly - there’s no one-size-fits-all approach. It all comes back to finding the right blend between automation and human support. And a crucial part of that is understanding why a customer is contacting you in the first place. Think about it - if I owe you money and I’m reaching out, it’s not because I’ve got time to spare. You’re probably the last person I want to talk to. So there’s always a reason behind the contact. The key is to understand the intent.
How does Conversational AI improve the way customers navigate support compared to traditional phone menus?
Cormac O'Neill Traditionally, this was done through clunky phone menus: press 1 for accounts, press 2 for something else - you go round in circles. But Conversational AI handles this much more naturally. Instead of a rigid menu, you simply ask: “How can we help you today?” And the customer can respond in their own words. Our system then uses a fully trained intent model - 42 global intents, straight out of the box, fully configurable - to understand what the customer wants and route them correctly.
Some companies might worry about how long it takes to implement AI. How do you address that?
Mark Oppermann For many companies, building something like this internally would take months - if not years. And even then, they’d only be able to cover a subset of use cases. What we offer has already been trained on millions of conversations, so it immediately understands what the customer is trying to do. That allows for a much more personalised journey - rather than forcing people through a rigid A-to-B-to-C path, hoping they eventually land in the right place.
AI and automation in customer engagement have made great strides in credit and collections, and the pace of change is accelerating. More and more businesses are embracing AI automation, and in particular, Conversational AI. But while the term is becoming familiar, many business leaders still have questions: What exactly is it? How does it work? And how can it add real value to digital debt collection? In a 45-minute webinar session, we’ll demystify Conversational AI and show how it helps credit and collections teams do more with less. What We’ll Cover
What Conversational AI really means
Custom Collections Language Models (CCLM)
The Impact on Customer Engagement Rates
Automation and Agent Efficiency
Getting the channel mix right (SMS, WhatsApp, etc.)
Real stories (Practical examples of how credit and collections teams use Conversational AI today to get results)
Register now: 30. September
Let’s say I’m a bank and want to use Conversational AI for origination - how long would it take to set that up? And what about the risks?
Cormac O'Neill Before you even talk about setup, the first thing you often have to deal with is fear. Some of our customers are very forwardthinking, but others are understandably cautious. They worry about how something like Conversational AI might disrupt what they’re doing today.
Take a contact centre agent, for example. There’s a natural fear: is this AI going to replace my role? And the answer is no - we’re not here to replace anyone. We’re here to augment what people do and make it better. You’ll always need your team. You’ll always need human support.
Mark Oppermann And then there’s the other fear: the AI going rogue. We’ve all seen the horror stories - generative AI giving strange, inappropriate, or simply wrong responses to customers. That’s a big concern, especially in regulated industries like credit and collections. No one wants to be in the papers for that kind of mistake.
Cormac O'Neill That’s the next level of fear - risk to reputation, risk of fines. But once we sit down and talk through how our AI works, how it’s trained, how it’s controlled and monitored, those fears usually ease. The key is understanding that it’s not about replacing people or taking risks - it’s about making things more effective and secure.
How do you ensure that the AI stays safe and compliant?
Cormac O'Neill That’s where our approach is very different from general-purpose AI. The AI we’ve built is 100% focused on credit and collections. We’re not using large language models that try to answer everything from cooking tips to travel advice.
Mark Oppermann We work with what we call custom language models - small or medium-sized models trained to do a specific job. In our case, that’s credit and collections. So our model won’t help you make the perfect carbonara, but it will identify when a customer is vulnerable, wants to make a payment, or needs support.
Cormac O'Neill And the responses it provides are all structured, pre-approved by compliance, and governed by the business. There are currently no customer-facing AI responses going out without human oversight and approval. In regulated environments, that level of control is non-negotiable.
What about data security - where does the customer information go?
Mark Oppermann People wonder if their data is being sent off to some nameless AI company overseas - maybe in the US or China - and what’s being done with it. It’s a valid concern. Everything we do runs within Aryza’s AI infrastructure. Nothing is sent outside. The entire system is built to keep sensitive customer data secure and in-house. That control is essential, especially when you’re working with regulated financial information.
Cormac O'Neill But AI isn’t something we’ve just jumped into because it’s suddenly popular. We’ve been working in this space for nearly five years, and from day one we knew: this has to be an enterprise-grade solution. And that starts with data security.
How does the platform meet regulatory and compliance requirements?
Mark Oppermann Regulations like ISO 27001, DORA, and requirements from bodies like the FCA come up in every due diligence process we go through - and rightly so. You’re dealing with one of the most valuable assets a company has: its customer data.
From day one, we built Engage with enterprise architecture in mind - especially in terms of security and compliance. We’ve been doing this for nine years, and things like data processing agreements, DPIAs, and security reviews are second nature to us.
Mark Oppermann You could have the shiniest product in the world, but if it doesn’t meet basic hygiene standards - security, regulation, compliance, it won’t even make it past the starting line. That’s why we’ve made those fundamentals a core part of our platform from the beginning.
Is the due diligence process a barrier for clients?
Mark Oppermann Not at all. We’ve invested heavily - money, time, and expertise - to ensure that all compliance and security concerns are addressed upfront. That really shouldn’t be a worry for any client.
Of course, there’s always going to be a 40-page questionnaire. Every enterprise has its own version. The questions are usually quite similar, and we’re well used to that. This isn’t a plug-and-play tool you start using the next morning. Large organisations follow a structured process, and we’re built to support exactly that.
Are there particular sectors where Conversational AI delivers the most value?
Mark Oppermann The sectors that benefit most are those where a large part of the business is B2C, there’s a high volume of customer engagement, and they operate sizeable contact centres to support that interaction.
If you think about financial services, retail, utilities, banking, insurance, and telcos - these are all sectors with large customer bases and frequent communication needs. Whether the engagement is regular or occasional, the ability to handle it efficiently and at scale makes a huge difference. And of course, what used to be just “call centres” have evolved into full contact centres that manage far more than just phone calls. That shift makes them an ideal environment for automation and intelligent routing through AI.
Any final thoughts for organisations considering Conversational AI?
Cormac O'Neill If you’ve already taken some steps toward self-service or digital channels, that’s great - but the next conversation should be about scaling that intelligently. Any company with a sizeable contact centre should be exploring how to move to the next level without abandoning what already works.
It’s not about replacing traditional channels entirely. It’s about blending digital and traditional in a way that plays to the strengths of both. That’s something we’ve become very good at - helping businesses evolve while keeping their customer engagement consistent, compliant, and effective.
Book a meeting with with one of our Aryza Engage Experts worldwide:
Mark Oppermann: Book a Meeting in English
Stefan Schetter: Book a Meeting in German
Eric Abcouwer: Book a Meeting in Dutch
