DRUID Talks Season 2 Ep#3 - Part 1 - Liberty Global's Inspiring Journey to Automate Repetitive Work with Conversational AI
Watch this episode to learn how a global telecom leader - Liberty Global - is using DRUID conversational AI and generative AI technology to automate repetitive work.
Join our host, Kieran Gilmurray, and his guest, David Hodsdon, VP of Operations for Liberty Global, to discover how the company used DRUID Conversational AI to revolutionize customer and employee support and learn from the lessons they gathered along the way. Whether you're considering implementing CAI in your business or just curious about its impact, this episode is packed with practical advice. At the same time, David's insights might provide new ways to consider technology's role in business. Let's get started!
Hello folks, welcome to DRUID Talks today I have Dave Hodsdon, the VP of Operations at Liberty Global. Liberty Global is a world leader in converged broadband, video, and mobile communications. Dave, you and I know each other for quite some time now, but set the scene here. Tell us a few words about yourself, Sir.
Yeah. Kieran, thanks so much for the introduction and absolute pleasure to be here. I'm thrilled that we finally found the time to do the podcast, right. So I'm really, really pleased to be here. A little bit of context and you've teed it up nicely. So, I look after the operations in a shared service centre that serves companies in the Liberty Global group. And what that means is we're serving 12-15 operating businesses, you know, household names that you'd know: Virgin Media O2 in the UK, you know if you're familiar with the Swiss business Sunrise over there and these are real kind of you know, household brands and the shared service role is the engine room. It is the common back office. So we talk about operations, we mean finance, transactional finance, accounting, we mean legal, we mean litigation, contract management. We're talking about HR, so, people, payroll and also construction support. So that relationship you have with planning and construction teams.
Just to kind of give you a few words on how the relationship with DRUID has developed in this support centre. We've been working with the DRUID team now for the last 2-3 years, really looking at use cases for conversational AI, how we try and make those interactions with our customers a lot more fluid, you know a lot more regular, and we try to meet our customer where they're already at. So, you know this is kind of replacing traditional telephony and kind of two-way messaging for us.
Well, it might have been easier there, Dave, if we said what you and Liberty Global don't do, because that's a heck of a list. And a lot you're not telling people as well, a lawyer who understands tech as well. So this is a bit of a Unicorn folks. We got what is today. So we're going to learn a lot. So Dave, I'm going to dive straight into the first question, just a little bit more context.
What were the problems that you were trying to solve as you looked for a conversational AI platform like DRUID’s?
Sure. Yeah, it's a great question. Fundamentally, understanding communication both into and out from our support centre, the kind of the back office that we described in the previous segment. And there are lots of reasons why people need to communicate with the back office, it's an extension of the customer and that's how we see ourselves and that's how you know we need to show up in order to deliver for that customer.
And the way that we have done that traditionally has been less effective. And so that really created this kind of need for something new. Traditionally, I'm sure it won't surprise many of your listeners to know, that the traditional back office you engage with either by telephone, unfortunately you dial the number, and you wait in line and it's a case of how many hands you have manning the phones or you’re communicating over traditional e-mail and then it's a ticketing system in the background. And I kind of refer to this as the deli counter where you go to buy your meat and cheese and you take your ticket and you wait till your number’s called.
And that was very much the experience that we were providing to our customers a couple of years ago. And that just doesn't meet with the expectations I think that we have as consumers of information. You know, if you're interacting with an online marketplace you're interacting with financial institutions, your banks, you're able to access those mediums 24 hours a day. They're there when you need them. And if we're going to be a really effective partner for our customers, we've got to offer that, we've got to be there when they need us. No wait times, immediate responses, friendly, professional interface. And that's the really use case for conversational AI for us.
Yeah, I think that's the day and age right now, isn't it, that customers actually expect digital service when they want it as they need it. You alluded a moment ago, Dave, that you had that semi deli counter experience beforehand.
What were you doing to solve challenges before conversational AI, and where was it leading to those frustrations that you started to look for that technology to solve? You know, the problems that you're just describing.
Sure, sure. So, the challenge that we had, and I think this is across transactional finance, across HR, payroll, where you have the relationship with your customers and they need information from you. Some of that information is accessible through self-service, but perhaps that kind of route to access that information might be cumbersome. It might be that you don't want to interact with that particular query type often. And so you're looking for just somebody who can offer you expert support guidance in the moment and that kind of traditional medium of that ticketing system, the deli counter example that we talked about. You're left holding an awful lot of data, qualitative data, as to why people are interacting with you, but you end up solving the query that's presented to you rather than the underlying issue. o what I'm really getting at there is, you know, you can allocate a lot of human resource, intelligent human resource, to dealing with a surfacing of information for your customer.
Let me give you a very articulate example there. If you're looking at transactional finance, it's quite common for a customer to come in and say “Hey, I'd like some information on the status of an invoice. I'd like to know when it will be paid. I'd like to update some information on the supplier that I'm working with. How do I do that?” And so it's kind of regular repetitive, but here's how you do it. You end up putting a lot of resource on just solving that problem. And it's quite easy to think at the end “Well, I've done a great job” because, you know, I answered 50 queries today or 100 queries today. But actually what you haven't done is provided that next level of customer support while you're really getting to the “Well, why did that happen?”
Why did you need to come to me to ask me to surface that information for you? And actually the underlying cause there might be the supplier I'm working with hasn't been paid on time. Or, you know, I'd send a request for some information and I've not seen it updated. Or, you know, and actually, using our human resource to solve the underlying issue with where I'd like to spend the time. And that's the thing that really wasn't being addressed before the conversation AI interactions.
So what did you do or what convinced you to pick DRUID’s conversational AI platform over another platform? What were the features or the benefits or the functionalities you really needed.
Yeah, that's a great question and ultimately, we were looking for a partner. The word partnership is so important to me when it comes to selecting technology vendors to work with. Our business is going through a journey, a transition. You know, we're we're digitising the services that we used to offer in an analogue way. We're looking at the value creation of having humans undertake administrative tasks. In moving our humans to knowledge-based tasks, we're trying to get closer to our customers. We're trying to move up the value chain from simple transactional processing to intelligent processing. So, producing monthly accounts, completing journals, undertaking balance sheet reconciliations, undertaking litigation.
And the way that we do that and still preserve value for the customer is to enter into a very strong partnership with a vendor that really understands how their technology works in a real corporate setting.
So, one of the catalysts for us was the great partnership with DRUID - and that is the people, that is the passion and the excitement. You know, you've really got to enjoy the work you do and be passionate about the products that you have and more importantly than that, be passionate about solving problems. And we really recognize that right from the outset with DRUID, that they were here to try and solve some of our corporate problems, and in the process, we'd form a very close relationship, and it'd be a commercial relationship that would work for both sides.
In terms of the kind of features, multi-language is really important. When you teed up this conversation, we're serving customers in the Netherlands, in Belgium, in Switzerland, and there are 8-9 languages spoken in those territories and whilst as a group of companies the common language there is English, it makes sense to meet our customers where they are. And so where you have people raising HR queries, for example, and they want to know about leave balances and they want to know about family leave policies, those questions are coming in Swiss, German, they're coming in in French, they're coming in in Dutch. And so DRUID's ability to be able to integrate complex language modules and really solve that language barrier.
The ability to be able to offer things like way messaging service where I can put UK based agents using the workbench of the DRUID tool, waiting to receive queries that come in in Swiss German that are automatically converted. So my UK user can converse in English and my customers or employees can converse in a natural language to them. And that's a really slick thing, I really like that.
And I also think, and I know we'll get on to it later, we're talking about kind of rapid advancement of technology at the moment. You and I, Kieran, talk about this a lot. You know, the market's moving quickly. I think it's important that the partners that we work with recognize that and are agile, flexible enough to see the changes that are happening to the market and look at how they respond to them. The deployment of new features, the iteration of their product road map, has constantly kept us feeling like we're at the front of the queue, the top of our game, when it comes to conversational AI.
I really adore that. Dave, you and I have talked about this in the past, but I see too many vendors actually reacting to the market as opposed to actually anticipating what's coming and spending time with companies. And I have to say I enjoy the multilingual part. Although a funny story, I was in Barcelona once, or Milan, and I was using Google Translate and somehow I told the taxi driver I loved him. I did technically, but not emotionally. So I'm glad to hear that is now all working many decades later and I love as well the way you mentioned you're focused on the customers and you're focused on your employees. There was no hint there whatsoever of taking people out of the business. It's very much about augmenting them, allowing them to do great things.
But Dave, the result of the pudding is in the tasting. What sort of results have you got from implementing the DRUID platform so far?
It's a great question, and I do listen to the show Kieran, when I listened to season one, some of the talks that really stood out for me were those where the audience can get some real kind of actionable insights. So I thought it was really important to bring some statistics, and stop me if you've heard too much.
The deployments that we've set out so far have been achieving about 65% query resolution ratio. So what that means is yes, 6.5 out of 10 queries that are coming through are being answered automatically by the virtual assistant without the need for a human user intervention. And you know why that's really exciting for me is we do a lot of analysis, and I'm sure a lot of your listeners do as well, on the reasons that people interact with us.
And absolutely somewhere between 50 and 70% of the queries that come into our teams today, the answers are pretty straightforward and I don't want our committed, experienced and loyal workforce picking up the same query 10/15/20 times in a day. It's administrative, it's frankly quite soul-destroying bits out there repeat, rinse, and repeat. A valuable public service, but the automations can take care of this. So let's let them lift the weight of that 65% and lets us focus on driving the resolution of the underlying cause of those queries, which is where we've been able to deploy our resources. So our people now understand what's driving the queries that come into the teams and instead they focus on partnering with our customers to resolve those queries at source.
Now in terms of the development of the tool, I'm really excited because I can see a glide path taking us from 65% to 85%. And some new features with the DRUID products that I know the team have been working on, that I really like. And this is about, let me give you just one specific example here, about something that's going to be really exciting, if that's OK. You know, the conversational AI itself, we're not just talking about the ability to be able to engage with a virtual assistant that can service fixed information. We're also talking about the ability to be able to configure the virtual assistant to go and take a next action based on something that the users inquired about.
And so I like the example back to accounts payable. I like the example of “Hey, tell me the status of this invoice” OK, it's currently unpaid, but why is it unpaid? Well, because the purchase to which the invoice relates hasn't been receipted so the business user hasn't said “I've received the goods or services that I was expecting to pay for” so effectively invoices come out of sequence for a traditional kind of three-way match. And that's useful information.
Today our AP clerk if you like to receive that information No problem, I will take the action to follow that up with the purchase order owner and I will say “hey, have you received these goods and services because I've got an invoice here”. But actually, the DRUID product is now developing and it's taking us on a journey where we can say to the virtual assistant “Hey, it's on hold. Would you like me to go and speak to the purchase order requester to see if they have received the goods and service? Yes, please.” And so then the push notification there on the same medium that we use to surface the virtual assistant, which for us is Microsoft Teams, push notification to say “Hey, if you've received the goods or services, would you like to receipt them? There's an invoice waiting. Well, yeah, I would. Please. OK, I can do that for you” You know, and actually and it becomes a bit of an evolution. Where does it stop?
But this is how I like to see, this is why I described the relationship as a partnership earlier, Keiran. This is about the DRUID team really understanding our use case and how we're embedding this technology and saying rather than just being a two way you tell me what you want and I'll give you an answer, let me do next best action. And through doing that, you can actually get really creative with your use cases.
Oh, wow, I love that, in so many different layers there. Particularly one when none of us come to work to do mundane, boring stuff, you know, And it's amazing how many people today still have to do that. And then we wonder why they disappear, aren't engaged, don't give the wonderful customer service that we've got.
Well, Dave, what, what lessons have you learned that you're happy to share with the audience here as you've installed and implemented and built and used and obviously got massive success from conversational AI?
Yeah, I've learned many lessons and I'm very happy to share them. I think about the 1st and probably the most pronounced lesson is don't wait to deploy until it's perfect. And I know you hear this a lot, and you’re like me Kieran, we read a ton of literature on evolving tech, emerging tech, and start-ups, and it is a common lesson. So, I realise this isn't news, but the importance of versioning, the importance of getting a product out there in front of your customer, that is pretty good. The bits that I would get right before I was ready to launch, the user interface, the user experience, so that when my customer connects, what they're seeing or what they're interacting with looks professional, feels, meets the expectation, if you like, with this virtual assistant.
But actually configuring too many pathways for automated queries and responses, I've learned just to get out, get version one out there and then update every, whatever it is, every six weeks, perhaps for the first six months, and then move to quarterly releases and the product gets better and better. And what that builds is a real relationship between the customer using the product, let's talk about the operational teams as well, that are also users of this product at the other side of the interface, and they are much more engaged in iterating the product. This is what it needs to look like. This is the next feature that's important because, so it's been super, super relevant and a great lesson.
You know, focus on the platform that you're using to surface the virtual assistant. And for us, when we started, we deployed our virtual assistant in the same place that we had our ticketing tool and access to our telephone numbers. And so for us this was a business SharePoint, it made sense. You know, all of the frequently asked questions are there. You're used to using this to kind of access your tickets. That's fine, but actually, that was still too many clicks away from the user. And actually, bringing that virtual assistant really close to the end user was very, very important.
And so with the businesses that we serve, they prefer Microsoft Teams as a common medium for engaging with one another. So, they talk to their colleagues over teams. Teams calls, Teams messages popping up all the time. And so, we surfaced the virtual assistant there. We locked it in as an employee. We gave it a snazzy name. It's important as well for engagement, but actually this has reduced the friction between “I have got a query how do I access the right form for that query”, now it's just the virtual assistant. So that was super relevant and a great lesson for us.
I think the 3rd and equally as important is that level of engagement, and I talk a lot about, and Danny, who leads our transformation function and works with me, talks a lot about this power of engagement. Engaging both our teams that will use the products that our operational teams and our customers that will interact with the product. Super important.
I'm told that 7 out of 10 transformation large scale change programs fail because they haven't secured sufficient engagement at all levels. And, tended to be in the early days, we were talking to the decision makers, feels natural, right? So, the CFO, the CPO, and their leadership teams about the value of deploying virtual assistants, got the buy in at that top level. But actually, the users of the platform, it’s super relevant to bring them along on that journey both from a user feedback perspective, and that goes back to that versioning that we're talking about a moment ago, but also the success of that project.
So, if you think about the operational teams that sit behind our conversational AI and pick up the what I probably class as Tier 2 or more complex or more bespoke queries, if the workbench that they're using to do that work isn't efficient, isn't effective, doesn't look the way that they'd expect it to in order for them to be able to navigate through, they'll find workarounds, and they don't always tell you because you're busy, 100 miles an hour deploying product, they don't want to say to you “Hey, I'm not using the platform as you thought I'd use it. I found another way of using it.” And so, we can drive kind of waste into the process that we hadn't anticipated by just not having that open channel of communication. So, I think that engagement layer is critically important.
Yeah, people forget the change management. But I would say in my own head, Dave, I'm picking a word out there out of all of that. That's friction. And you remove the friction from the user. You remove the friction from the gap between the actual team and the user and the outcomes. You've removed friction from the leadership and the comms and the change management. Really brilliant lessons for everyone.