Host Intro (Gene Kim)

So first up today, this afternoon, will be Barbara Arnst. Last year she presented as transformation leader and organizational designer at Telenet, which is one of Belgium's leading telecommunications providers. And she co-presented with her VP of billing counterpart, Johan Morel, on their multi-year journey to rewire their organization to get better outcomes.

I thought it was one of the best presentations of the conference. I loved how their story created value for business leaders who actually had profit and loss responsibilities, and it was being actively led by their CEO. It was such a fabulous experience report on their two attempts to reorganize, to rewire: one that didn't improve things, and she explains why despite many consultants attempting to help, and the other that did. And she explains why it worked.

So I'm so delighted that Barbara and Johan are back to present the continuation of the journey, which also includes their attempts to use AI to provide an effortless digital-first billing experience. Here's Barbara and Johan.

Barbara Arnst

Thank you. So thank you, Gene, for those nice words. For those of you who were there in our 2023 talk, welcome back to the sequel. And for everybody else, great to have you here. Johan and me are thrilled to be back.

But let's dive straight into it. I'm going to start with an uncomfortable truth. What if the biggest bottleneck for your AI strategy isn't technology, but rather it's the org chart? Well, I think it is. And it is because adding AI and adding AI agents isn't like installing new software. It's adding an entirely new actor with its own set of rules to what is often already a very complex system. And to do that successfully, you need to rethink the way you organize work.

Why do I call this an elephant in the room? It is because for many companies, especially large legacy ones like ours, who have spent years rewiring to be more agile and product-centric, the leadership, if we're honest, is really reluctant to go through another rewiring effort. And what I see is that a lot of the mind space of leadership is going through prioritizing AI use cases, prioritizing investments, but is not really thinking about what the organization needs to change to capture the value. And that's a challenge that we're tackling at Telenet head on.

If we think about an operating model, designing a modern one can be very tough. Luckily for us, there are many best practices and principles that exist, and many of them will be shared and discussed today. We have borrowed from these extensively, but we've borrowed with a very deliberate thought and always adapted them to our unique context and strategy. The resulting operating model translates how we think about strategy into concrete value.

Over the last years, we've rewired and changed our operating model numerous times. We've done this either because our strategy changed or because we hit the limits of a previous operating model. Back in 2020, we moved from a siloed hierarchy into what we call the Agile 1.0 model. This one was based on Spotify. When we hit the limits of that in 2023, we moved to the model 2.0, where we eliminated the business and IT divide and started getting serious about outcomes. Today we're at the cusp of another iteration where we add AI to the mix.

I think what's interesting about our journey is not the different models per se, but rather that we have the CEO and political support to make these changes in a very deliberate way, whether that's small adaptations or some of the bigger jumps that are illustrated in the slide behind me.

First, a few words about why we moved from 1.0 to 2.0. The tension that we experienced in the 1.0 is exactly what we illustrate here with the matrix: org design is a fractal and not a layered structure, as what works at the team level must be consistent with what we have at the enterprise level. What we saw in the 1.0 is we launched all of these as Spotify autonomous agile teams, but still had the enterprise structure of Telenet like it was in the nineties, with the very traditional IT department, for example. So not surprisingly, our teams soon got stuck in a lot of dependencies. So we refocused our efforts in a 2.0 in aligning the structure at the enterprise level with the agile way of working we already had at a team level.

So much for the theory, but what did we actually do? In essence, if I were to describe it in one sentence, we stopped organizing around functions and started organizing around the value our teams create. We defined different patterns at the enterprise level, at the N-2 structure we call tribes at Telenet. Tribes are led by a VP, they report to a C-suite member, and typically have 10 to 15 different teams reporting into them. We make a distinction between customer tribes, platform tribes, and enterprise tribes based on the type of outcome that these teams pursue.

Crucially, all of these tribes are empowered with all of the means to accomplish their mission. The means that they have are the business, development, ops, and all the people responsibility included within their remit. And so in a sense, each of these tribes is almost like a mini enterprise.

Let's now look at the real practice. We translated this into practice into an org chart that we rolled out across the company. Two callouts on that. First, as you see here, each C-suite member, except for the CFO and finance, has a yellow tribe within their scope. That means that they have a customer outcome that they're accountable for. So this is customer centricity at scale and by design. Similarly, we don't have a single CTO anymore, but it is federated across the full company. Most of these yellow and blue tribes own software development. So again, by design, we eliminated the silos between business and IT.

As you can imagine, this was a challenging journey, and a lot of effort has gone into rewiring decades of muscle memory of a functional organization. We're slowly getting there and making the steps.

But a really interesting question now is, are we set for AI with this model? I think the good news is that at the enterprise level, the structure holds: the customer tribes, who own an outcome end-to-end; the platform tribes, who provide a common service or platform for the entire company to tap into; and the enterprise tribes, who have a mission to tie all of that together in an enterprise-level way. I think is really great.

But dig a little deeper. When you look at the day-to-day working of the teams within those tribes, that's where we get stuck. The practices, the governance structure, even some of the ownership patterns, are actually feeling like yesterwork in face of what AI will bring. To bring that to life, I'm going to hand the floor to Johan, who will share his experience.

Johan Morel

Thank you. My name is Johan Morel. I'm a tribe lead of the Billing Experience Tribe. Like Barbara explained, we are one of the journey-owning tribes within Telenet. And as Barbara explained, at that moment in time, we tried to define the tribes as end-to-end as possible, which is quite a challenge in a complex organization like a telco company. But for the billing area, we managed to do quite well, in fact.

As you can see on the slide, we have really centered three big teams in our tribe. We have a team responsible for the customer experience and the journey design. We have one team responsible for the day-to-day bill-to-cash operations. And we have one team responsible for all the underlying IT applications. The nice thing is that they are all centered around that same goal and that same mission. We want to give our customers an effortless, digital-first experience. That common goal, that's where the real difference is made.

We started this exercise two years ago, and back then we also had a talk here in one of the breakout rooms. Then Barbara and I could express the belief that we had in that model. But now, two years down the line, I can really say that it really, really, really works. The big difference for me is really the fact that you bring business and IT together in one team, and then it becomes a world where it's not about dropping code to the business, it's really about co-owning the business. So that's the big step change that we have made so far.

Then now AI comes into play, and AI offers us a lot of opportunities. They're all directly linked to realizing our strategy to give the customers that effortless digital-first billing experience. When you look at a typical telco invoice, I can state that it is rarely incorrect. I do have to admit that sometimes it can be highly complex, and that's where AI then can help us tremendously. You see here on the slide different use cases both for the agents as well as for the customers that can help reducing that complexity.

That's all great news and we're only at the start of it. But already we can show some promising results. We have a customer chatbot live, really in its early stages, really very, very basic, but already we see that we have a containment rate of about 20% of people that start engaging with that customer chatbot. All in all, a great era to be in.

But when you enter an era of opportunities, you also enter an era of challenges, of course, and that's what we face as well. You can summarize those challenges in three distinct challenges.

The first challenge is the fact that we focus on individual touchpoints, and we manage the technical drops as one-off programs. What do I mean with that? When you look at a typical customer journey a customer could go through: you get your invoice, you go to the self-service portal. At Telenet, it's called My Telenet, where you get some basic explanations about your invoice. You have detailed questions, you ask the customer-assist chatbot for some detailed insights on that invoiced amount. And if you then still would call, you get an agent on the line that is also supported by an agent-assist bot. Of course, that consistency from left to right through that whole journey is crucial, and that's something as journey owner that we need to safeguard.

But then enter the blue tribe, the capability-owning tribes. Today we have different teams responsible for the customer-assist chatbot and for the agent-assist chatbots. So by design, they are really capability-focused and not journey-oriented. That's already a first challenge.

A second challenge that's linked to that is the fact that they really look at those things as technical drops. Today, we from the Billing Experience Tribe are very happy that there is focus on billing use cases, but tomorrow those teams will shift their focus, and rightfully so, to more technically oriented journeys. Then my question is, if we want to continuously improve our chatbots, how will that work? If for every single improvement I have a dependency towards a team that has at that moment in time a completely different focus, how will that work? That's the first challenge.

The second challenge is when I look to how we selected the use cases, then I see that it's really focused on as-is thinking and how we can focus on cost reductions and call avoidance. What we lack at this moment in time is really to look from an end-to-end process perspective: how can AI help us in the improvement of the full end-to-end processes? That's something we lack as well today.

The third big challenge is the fact that we have too many cooks in the kitchen. We have several blue tribes. You have data tribes, architecture tribes, you have tribes owning the customer chatbots or teams owning the agent-assist chatbots, all different teams. You have the yellow tribes, you have Telenet as a group, you have Liberty Global as a mother company. All also want to have a say in this. So everyone thinks they own AI, and that slows down the decision process massively.

Here you see a simplified architectural chart of our customer-assist chatbot. One central component in that architecture is the billing agent. That's really targeted to answer billing-related questions from the customer. Today, it is an in-house built agent, and I came across now with the opportunity to go for a vendor-specific agent, which I really want to test out. Then I want to launch a proof of concept on that. Really today it's unclear who can even decide whether or not we can proceed with that POC.

On top of that, we see that we also lack the flexibility in our decision taking. The world of AI is evolving so fast and so rapidly that new opportunities arise every day. So at least you should have the opportunity to look at those things individually without getting bogged down in endless discussions about budgets and some costs.

All in all, it's a very interesting area we are in, very exciting. But indeed we face three main challenges. We miss the customer lens. We look too much from a capability point of view. We limit ourselves to automating the as-is, and we should look beyond that one. And the third thing is that we lack today clear decision authority.

Barbara Arnst

Thank you, Johan, for laying out the challenges. How are we addressing these at Telenet? That's where the enterprise tribes come in. Agility and Transformation is one of the four enterprise tribes, along with HR, software architecture, and strategy. Together we have as mission to drive the transformation vision and design the operating model of the company, but equally, we're accountable for driving the adoption of the different capabilities and proving the value to the organization.

That means very concretely, if we look at the way we view operating-model transformation with a real outcome, for us as enterprise tribes it is quite unique. A few things I want to call out there. First, we don't treat transformation as a project with an end date, but really as a product with a lifecycle. We own the capabilities across the lifecycle and have real skin in the game for ensuring that they're fit for purpose. Second, we don't copy-paste our templates or other frameworks, and most importantly, we ground all of the changes in the reality of the company. We really try to view it as a learning challenge and not a technical rollout. We're able to do this because we're not incentivized to go for quick fixes or follow consultant hype, but we're really owning the operating model and really take that long-term stance. I think this makes a very different dynamic than if this transformation effort was another outsourced program or one that's led from the top, but had an end date. This gives us a dynamic that I think is a lot healthier.

Again, if we look at how we're approaching this, we lean very heavily on the learnings that we had from our earlier agile and digital transformation efforts. As always, the challenge is twofold. On the one hand, designing the structure, designing the way of working, but equally bringing those structures to life in the company, getting our people across the bridge, so to speak. This is a whole different challenge and one where, again, we lean on best practices. In this case, I'll highlight some from Rewiring the Winning Organization.

The first one of those is slowification, and in a world obsessed with speed, we're advocating for slowing down and really very deliberately pacing the journey for our organization.

Let's look how we did that in the past. The picture we see here on the top is actually from 2021. It's our agile transformation plan, and that's the CEO of our company next to it. This plan sets a direction for the company, but also broke down the journey into very manageable steps. Its illustration here is with base camps on a mountain, which deliberately pauses the organization at different stages in that journey.

With AI, we're using the same picture. We still have the same mountain that we use at Telenet. But as Johan described also, there's a lot of different efforts going on at the moment with AI: a lot of pilots, a lot of experimenting in the company. Over the last year, we actually very deliberately let that happen, even though we knew it wasn't the end game or super sustainable. In parallel, we used the time over the last year to strengthen some of the foundations of our company, start to build out the platforms, but also engage all of the enterprise tribes to see how we're going to support the organization in scaling.

Very concretely, what we've done now is we've defined a plan to roll out playbooks for teams like Johan's that go beyond the single use cases and see how we can set in place more sustainable product teams to have their own data pipelines, and also a mission around AI that is more long-lived. I think our leadership wouldn't have bought into the structured AI approach a year ago, and there was a lot of excitement around the hype of AI, but also because us as the enterprise tribes, we didn't know what to tackle. We didn't know the real pain points that our teams faced. So now it's different and we're able to continue the journey.

That's first, solidifying. Next is simplifying. How are we doing that? In the past, what we've been very deliberate about is creating different patterns for the organization that address real concerns for the company and being super disciplined and relentless about rolling them out and using them consistently. Here you'll recognize the different colors that we had in place for our tribes and the agile model.

How are we approaching this for AI? I think a first insight is that with all of the buzzwords and the breakneck speed that AI is entering the world and our company, we're seeing this really as a recipe for cognitive overload for our teams. So simplification and creating a mental model for our company is more important than ever. We're starting here also from the existing team patterns that we have in place, but augmenting them now for the AI reality.

What we're seeing is that if AI and all the different flavors of AI-augmented teams come into play, it brings a whole set of different challenges. How do we build trust between teams? How do we revisit collaboration models? And how do we manage algorithmic design?

Given that trust is such a big issue, we're very thoughtful about designing patterns that actually reflect the logic and the foundations behind the level of human and AI collaboration that we want. There's a lot of unknowns. So we're starting in very small steps, and the first step we're taking is answering who in the company owns a given AI bot and what is expected of that team.

This isn't trivial, as AI bots of course have a big blast radius, and we have them all over the company now. So we're giving very practical guidance to the teams on what is expected of them. Equally important, I think, is that there are also areas that we put on the parking lot for now, or we are very clear to the organization that are not part of the pattern library at Telenet at the moment.

In summary, for a 3,000-person company like ours, we want to get the narrative right and we want to get the guidance right. But equally, we want to start experimenting in small steps and not wait until we, or somebody else, has figured everything out.

Simplifying is one. Last is the amplification and getting signals from the field to shape our transformation journey. Again, this matters in a world with AI more than ever, as there is a lot of anxiety about AI and what it will bring to the organization. Rather than ignoring those signals, we're really seeking ways to amplify that and capture that feedback.

In previous efforts, as we see here, we've been very disciplined about gathering feedback around the adoption rates, for example, of the different agile capabilities, and also using that data to inform what works, but also what doesn't work. For AI, we know we need more. To that end, we want to get feedback from our teams already before we start redesigning workflows. Next to the technical feasibility, we're mobilizing teams from throughout the organization to shape and to decide with us how we're going to redesign workflows and where we will use AI to augment them. I think this is super important, as doing that collaboratively with members from the field really builds trust and beats dropping the AI solutions blindly onto the organization. This is an investment of time, for sure, but one I think that is really important and one that we will over-engineer in the rest of the rollout.

That brings us to a next question there. Are we there yet? Is everything crystal clear in the organization with no loose ends, and are all of our teams now wired for +AI now? Or as the slide here shows in Dutch: are we ready for tomorrow?

Johan Morel

I think the honest answer is not yet, but we got this. We have solid foundations in place, and our organizational fine-tuning, we know how to do it. It's really an integral part of our company culture. We have done this already three times, and we will now do it again. Second thing is that we have solid feedback systems in place that really feed that organizational fine-tuning. And the third thing is that next to the journey-owning tribes, we have also transformation tribes who have really skin in the game. So side by side, I'm really convinced that in the very near future, we will craft that AI-ready operating model.

Barbara Arnst

To summarize, I am going to quote Charles Darwin, who said the ones who survive aren't necessarily the strongest or the smartest, but they're the ones most responsive to change. I think in essence this is what you can learn from Telenet. It doesn't really matter what your operating model looks like or what the starting point is, but it's all about being able as an organization to experiment, to learn, and to adapt.

Whether it's a small or a big mechanism that you are changing, it's just making sure that as an organization you have a system set up in the company to enable that organizational learning, and if somebody has real skin in the game to continuously improve that. My final takeaway is: take your operating model as seriously as you would your AI LLM strategy, and version it, test it, and learn from it.

That brings us to the closing slides, where we'd like to thank you for your attention and we'd love to continue the dialogue with you on how you are elevating your AI-ready operating model to be a top leadership concern, and equally how you are engaging your organization to actually want to slow down in this exciting journey. Thank you very much.