CR084: Hyper Automation with Johan Sporre, Ericsson

Agentic intelligence will enable unprecedented levels of automation, it is a next step on from what has been possible for a number of years.  Is it now possible to automate 80% of IT tasks?

This week, Dave, Esmee, and Rob talk to Johan Sporre Lennberg, Vice President Cloud Service at Ericsson about how Hyper Automation is possible, to what extent and the best way to go after it in IT.

TLDR:
01:35 Confusion about AI Ethos
06:28 Cloud Conversation with Johan Sporre
46:00 Showcasing value
54:40 Riding in the season's first snowfall!

Guest:
Johan Sporre Lennberg: https://www.linkedin.com/in/johan-sporre/

Hosts
Dave Chapman: https://www.linkedin.com/in/chapmandr/
Esmee van de Giessen: https://www.linkedin.com/in/esmeevandegiessen/
Rob Kernahan: https://www.linkedin.com/in/rob-kernahan/

Production
Marcel van der Burg: https://www.linkedin.com/in/marcel-vd-burg/
Dave Chapman: https://www.linkedin.com/in/chapmandr/

Sound
Ben Corbett: https://www.linkedin.com/in/ben-corbett-3b6a11135/
Louis Corbett:  https://www.linkedin.com/in/louis-corbett-087250264/

'Cloud Realities' is an original podcast from Capgemini

Creators and Guests

Host

Dave Chapman

Chief Cloud Evangelist with nearly 30 years of global experience in strategic development, transformation, program delivery, and operations, I bring a wealth of expertise to the world of cloud innovation. In addition to my professional expertise, I’m the creator and main host of the Cloud Realities podcast, where we explore the transformative power of cloud technology.

Host

Esmee van de Giessen

Principal Consultant Enterprise Transformation and Cloud Realities podcast host, bridges gaps to drive impactful change. With expertise in agile, value delivery, culture, and user adoption, she empowers teams and leaders to ensure technology enhances agility, resilience, and sustainable growth across ecosystems.

Host

Rob Kernahan

VP Chief Architect for Cloud and Cloud Realities podcast host, drives digital transformation by combining deep technical expertise with exceptional client engagement. Passionate about high-performance cultures, he leverages cloud and modern operating models to create low-friction, high-velocity environments that fuel business growth and empower people to thrive.

Producer

Marcel van der Burg

VP Global Marketing and producer of the Cloud Realities podcast, is a strategic marketing leader with 33+ years of experience. He drives global cloud marketing strategies, leveraging creativity, multi-channel expertise, and problem-solving to deliver impactful business growth in complex environments.

What is Cloud Realities?

Exploring the practical and exciting alternate realities that can be unleashed through cloud driven transformation and cloud native living and working.

Each episode, our hosts Dave, Esmee & Rob talk to Cloud leaders and practitioners to understand how previously untapped business value can be released, how to deal with the challenges and risks that come with bold ventures and how does human experience factor into all of this?

They cover Intelligent Industry, Customer Experience, Sustainability, AI, Data and Insight, Cyber, Cost, Leadership, Talent and, of course, Tech.

Together, Dave, Esmee & Rob have over 80 years of cloud and transformation experience and act as our guides though a new reality each week.

Web - https://www.capgemini.com/insights/research-library/cloud-realities-podcast/
Email - [e-mail]

CR084: Hyper Automation with Johan Sporre, Ericsson
[00:00:00] And it's like, sky gray, but clouds coming, Rob's here, everyone's running.
Welcome to Cloud Realities, an original podcast from Capgemini, and this week, a conversation show about the simplification and the hyper automation of IT. Is it possible to automate all aspects of the IT operation these days? I'm Dave Chapman, I'm Esmee van de Giessen, and I'm Rob Kernaghan. So, Johan, good to see you.
We've got Johan Sporre from Ericsson with us. Do you just want to say hello and say a little bit about what you do, Johan? So, I'm heading up an organization called Tech Foundation in Ericsson IT, reporting into the CIO. We've just gone Lie with a new organization from 1st of October and focusing on [00:01:00] hyperautomation is our big theme.
So we'll dig into that in a minute. Looking forward to getting into that. In a year where agentic intelligence has risen and now Rob is almost getting his head around the way that that's going to fit in. I think hyperautomation becomes a very, very valid topic. So Rob, you still think it's a good idea? Uh, modern RPA, or have you got your head around it anymore?
Oh David, oh David, what it is to know you. I'll tell you what I am confused about this week, and actually it's very interesting, and it goes into this conversation we're about to have, which is AI is hitting a very interesting point on the ethos of what's going to take us forward to true intelligence, and you say AGIs, etc.
But, two schools of thought. The first one is, if you keep pumping data into an LLM, and we know it's just using statistics and tokenization, however, the bigger the models get, perceptibly, they become intelligent. And this has been the rise of agentic intelligence, and we see the agents are becoming ever more [00:02:00] sophisticated.
However, There is some evidence to suggest that might be topping out. So the sort of chat GPT, Sam Altman view is, we'll just keep pumping the data in and eventually we'll get to a threshold where you can't tell. However, another school of AI that was actually back in the 90s says, you need to use models to get to the next level of intelligence.
So when you think about it, when you go into the world, you have a load of models, like you understand gravity, you understand how cars work, you understand if you go around a corner, it'll lean you, things like this. You apply that model, and then if you throw a ball, You're able to predict what's going to happen, and you can apply intelligence to that.
You know, the ball's going to come, I'm going to be able to catch it, whatever. And so in the next couple of months, they're suggesting that we're going to hit this threshold of, will we see LLMs top out, and they are what they are, very useful, but they've hit their limit and we need to augment with models and think about how we take that next step.
Or, Will the view that LLMs will perceptibly become intelligent actually happen? And I'm confused about, is it going to go left, or is it going to go right, [00:03:00] and what's going to happen? But we should find out relatively soon, because the amount of computers going into training the new versions of the LLMs is so great now that they think they're going to find out.
And I believe you, there's going to be some winners and losers in this one, depending on which way it goes. Well, what do you think about that, Es? God, I have no idea what he was talking about. I understand LLMs, but, you know, this is the point where I just think, I'm so happy that you're here and you understand these topics.
He's confused about it. Well, I don't, because I'm confused. I don't know which way it's going to go. But I had a really interesting conversation about it when we were discussing AI the other day in a workshop, and it was like, yeah, actually, there are some statistics that say it could go one way or the other, but there are people abetting the farm, um, on particular directions and, you know.
So what's a model, right? Let's attempt to dig into this in a, in an extremely hobbyist way, I would suggest. So what's a model, right? It is where you have recognized a repeatable pattern. [00:04:00] Understanding of how that pattern gets applied as well, like gravity. Let's tell you, right, we all understand if I hold a ball and drop it, it's going to fall to the earth.
Yes. And the model in your head says that's going to happen. And if one day. you release the ball and it just stays where it was, that's broken your model. But we are assured that gravity is a thing that when we wake up in the morning, it's still there and it works. So we, that's a good way to think about a model.
We understand in our head what gravity is going to do when you undertake an action. So we got to that point as humans by examining, observing, experiencing, and sometimes experimenting, improving, and writing down. Now, some bloke got hit on the head by an apple and it was all clipped. Right, right. He's like, things fall down.
Always the bloke. Hello! Excellent. Now, question. Do you get to a point, therefore, where the data set in an LLM is big enough to spot repeatable patterns? Oh, that's it, isn't it? So, so, if you put enough in, and the tokenization statistics work, you can ask it that question, what's going to happen when I let go of the [00:05:00] ball?
So my repeatable pattern would say, an apple will fall to the ground. Exactly right. So, It's perceptibly intelligent, because you don't know any difference. And that's what they're saying with the LLM angle, is that if we put enough data in, we're going to be able to use that corpus to be able to get data.
What is essentially what a model will do. And there's a big, there's a big school of thought that says it can't do it, and you'll need to develop models and apply them to technology to get the intelligence. Is that school of thought, Robert, coming from the architecture community? Yeah, we're going, it's never going to work.
Hands like, you know, hands on the hip. time. That's never going to work. I don't believe what we need. Everybody is. We need some blueprints and models. And it's the also thing it's happening in autonomous driving. And the way that's working is how far can you take autonomous driving with the way it's being codified before we need a model to understand how gravity will affect the car.
Say when it goes around the corner, I'm on the chaos side as where are you? I'm in the like, yeah, put enough data in and let's sort, let's, let's see if patterns emerge. Absolutely. Rather than trying to enforce [00:06:00] patterns. Well, you see, yeah, okay. I'll remember this conversation, and it will be recorded and it will go out, so, uh, we will come back to this at some point, but I'm, I'm, I'm unconvinced.
Well let's see how things go. For today though, let's close the book on that one. I think it's chaos and it is, so let's move on. Trust you to pick chaos. Yeah. I want order, you need chaos. I just, I don't know. Embrace yin and yang. It's the yin and and yang at the heart of the show, Robert. Anyway, let's get onto our main topic of the day in the telco miniseries that we're doing on the show at the moment.
In our first episode, in that miniseries, we had Rainer Deutschman on and he was talking about the modernization that telco industry is going through. The fact that it's a legacy heavy business by its nature, and therefore the simplification is a, is a required thread. And that simplification is challenging.
Before we get on to other aspects of, you know, simplification like things like hyperautomation, let's [00:07:00] just take a step back and give us your take on the simplification of the industry that's going on at the moment. Yeah, let's I'll show you my reflection by no means trying to represent Ericsson, but, um, I think if we look into the telco industry, it has a declining revenue for decades, and we can all experiences a lot of the revenue that the business is receiving is from fixed subscriptions.
So as telcos invest in new technology, they have a challenge actually increasing their margin and, and, and. capitalizing on a return on investment on those investments. And when we look into simplification and cost efficiencies, they go hand in hand. So it's very important to get costs down for new investments and really make the networks more simplified and more efficient.
And in terms of the industry itself, the diversification of business models, I think must be quite central to the modernization challenge. [00:08:00] It is. If you look at, as you said, it's quite a legacy technology or legacy, uh, uh, heavy technology industry. And when we look ahead, because the decrease of revenue as a market as a whole, There has been a lot of investments and ideas on how to bring this forward and where Ericsson is doubling down and a lot of the industry is doubling down is to go from what I would call dumb networks to intelligent networks where you can actually capitalize on new features and functions.
And that is intended then to increase the whole telco market value rather than having it decreasing year on year. Right, right. So what does that mean then for the underlying app estates that are supporting, or maybe it's actually broader of course because it's infrastructural estates too, that are supporting these organizations.
So with an IT lens now or a tech lens, the modernization challenge must be intense. I mean, it's, it's difficult enough in, [00:09:00] you know, say retail organizations that have supply chains that maybe have some legacy ERP in them and, and And when you want to swap that out, that's, it's a high risk challenge. It's a high cost challenge and it's difficult and, you know, organizations need to push through those transformation exercises.
But in an organization where the, the business itself is so deeply embedded in the legacy infrastructure and potentially the legacy app estate, the challenge must be kind of 10 X that. Is that how you would hold it? Or how do you manage the modernization challenge from a tech standpoint? Lens perspective.
If I look at this from an IT perspective and not the telco and we can come back to that. But, you know, when I look at this from an IT perspective, I just want network to work, right? So as an IT consumer, I don't get so much excited. But as soon as I put my telco hat on, I do get excited. And I think this is the big shift that being successful in intelligent network will open up functions and [00:10:00] features, which will be interesting for the IT departments and for product and development units to actually capitalize on new security features, new type of consumption patterns for consuming, uh, connectivity.
And, and we're expecting that more services are going to be invented and that will increase your wallet for everybody to, uh, to do right. And, and with that, the thought processes with intelligent networks, that, The complexity will of course be bigger. You need to understand these networks. You need to understand how you capitalize on it.
So it's a big transformation of just having it as like a bit pipe in rather and then now being able to Utilize connectivity in your offerings and capitalize on that. So that I think is a big transformation If you look internally you would need them to be able to do that. You would need to be moving from a more static, uh, network to a more programmable network, automated network.
So to be able to harness that programmable interface for the intelligent [00:11:00] network, you need to be able to replicate that through your infrastructure. And that is, of course, a transformation from an IT perspective. Yeah, yeah, you're right. And what are the telcos at the moment? Are they still predominantly traditionally structured?
Or have they started to move towards like product and platform type structures in the way that kind of, you know, IT as an industry is, is, is broadly transforming no matter what, uh, vertical or any other moment. I see a lot of telcos moving into, especially in the western side, is moving into public cloud and as such they typically follow with a product oriented model.
So I see a lot of shifts on, on the market and Ericsson is now, you know, In the IT department doing a big shift into a product oriented set up, we've been on that journey for quite some years, but now we're doubling down and doing a big shift across the whole Ericsson IT system. So I think the trend is, is quite vivid in the telco space as well in terms of how you organize and how you drive [00:12:00] value in.
Does that also mean that your IT is going to put on that telco hat more often? Do they already do that or is it, does it feel quite separate from each other? No, I think, uh, in Ericsson, we, we contribute a lot to the, the, the business. And I think we get more and more involved to the business areas, especially when we talk about modern technology.
So one of the big shifts, why we do the product oriented model, why we're doubling down on it is that we firmly believe that as an industry. And as, as people we've been working More like a project based organization. You know, the organization has come to us with requirements. We've, we've had that interaction and trying to understand the requirements and then we built to order, right?
But as technologists, we actually know much more about technology and how technology could be used in the business than sometimes the business do. So in order to promote that. We're looking a lot into feature products and [00:13:00] fusion teams, how we can work closer to the business, closer with the business, understanding their needs.
And by understanding their needs, we can also be more of an advocate of technology and how technology can be used in the product and service life. So we're definitely moving in that direction. And for us, having a modernized platform in public cloud. helps us because we can then talk a lot about those capabilities and trending technologies that the providers have.
And we have quite an in depth knowledge about that from IT versus the telco part. So what have been the challenges on that IT transformation? Um, the move from what I'm guessing was, well, actually, let me not guess, let me just ask you, was the traditional model that you had in place, like an ITIL structure, With central, with, you know, with teams and capabilities centralized, like centralized infrastructure, centralized security, those sorts of things.
Yes, very, very much so. An IT based structure and also a IT for IT version two, which [00:14:00] is a product or service oriented way. So, By, you know, by, by deciding that those are the framework that we're going to work against, all our tools are configured to actually segment and control those type of implementation of the processes.
So from a working perspective, when you come to product setup, it's a big shift mentally in what you want to achieve and how you achieve it. Then that trickles down to, okay, now we need to change our process landscape. And then. Okay. By doing that, we also need to reconfigure our tools and maybe do new tool selections in order to have those process deeply implemented.
And, and it's also a different mindset, you know, in a product oriented model, the product have a much more, uh, bigger autonomy and much more to say. Yeah, yeah. And that needs to be real balanced to versus a more of a command and control type of culture in the project and project portfolio type of setup.
Which is a big culture shift. So how, how did you go about, I'm [00:15:00] interested before we sort of dive into some of the, the process work that you've been doing, I'm interested in that culture shift aspect and the, and the leadership of that. So where did you start with something like that? Because you know, we've talked about like, do you start, do you start bottom up and start kind of building coalitions around what a new way to do things might be?
Did you start top down and start with like. Getting the leadership headspace in the right space. Like just explain how you guys got into such a large transformation. So I think a couple of years back, Ericsson realized that we had to restructure the way we were doing business with our customers. We were, we had a An important thing to digitalize some of our business processes and also our interface to customers.
So what we did was we invested, uh, from the business on a digital factory, which was set up on the side of, of the original structure. So we took in BA consultants, and we built this digital factory, [00:16:00] which aimed to, to improve. So I would say we sort of ended up with 15 new applications, highly composable, uh, highly.
cloud edge type of, of environments. And in that transformation, it was decided that we would run that in a product oriented structure, which was the first sort of interface we had with these type of products done. But we also quickly realized that digital products, it was actually spending multiple BAs and multiple group functions.
And that created a need of actually maintaining these products when in operations in a different way. And we couldn't sort of handle that in the old structure. So that sort of triggered us to start. adopting a product oriented model for these products. So at the same time, we revamped our compute modernization.
We also revamped how we worked. So quite, uh, quite some parts in the organization then went with a product oriented model, which meant that they were living in dual worlds, adding complexity, [00:17:00] adding challenges and friction, but it also made the organization learn. And we were able for a couple of years to run in a dual mood.
And then at this point, we said, no, you know, the model has been tested. It's matured. Let's roll it out to the full enterprise IT. And I think, One of the driving parts is really this customer centric view on recognizing that we are actually, uh, good in technology and that's something we can infuse to business.
So we have something to give back here. And then if we change the ways work and you raise a very interesting point about the learning cycle, I think there's a purity around product and platform operating models that many can embrace and understand when you take The complexity of an environment like the one you operate that has a long history of technology that's powered it, there's that understanding that you must adapt it quite quickly to the organization at hand, and it's imperfect, and we must compromise, but keeping the [00:18:00] ethos of being centric to the experience or centric to the outcome, I think, is the big thing that then resonates with the business, and they can get more involved as this.
Yeah, we all learned and read the book on But there's nothing like trying to implement one in the sort of complexity you're talking about that you learn fast about how you have to adapt them to make them work properly. Yeah, I agree. The thing that triggered in my mind when you're talking about that is how purist do you go?
Yeah. You know, because like the, the sort of the whiteboard, You know, Spotify model of super pure products and DevOps and things like that is in the world of the, you know, high custom build world. And not, not every organization is like that. In fact, most organizations are not like that. They've got aspects of custom and aspects of quite a lot of platform products, whether it's SaaS or ERP.
And by the way, you're also dealing with bringing legacy into that world. So you're not going to refresh all of your tech as part of that journey. So in my, in my [00:19:00] head, it literally makes me want to bang my head on the wall when people get purist about it. Cause it's like, it's just not real. Yeah, no, it's not real.
You can't do it either. I mean, if you look at the traditional enterprise IT organization and let's just, you know, take away the business aspect for a while, you might have what 500 applications. And out of those, how many are actually market differentiating a handful. So if you go puristic on applications that you just want to tolerate, you know, is it worth the investment?
And also they can be highly outsourced. So it's a, it's a good model for that. So of course you need to. Think about that. You change this whole process, set up and the tooling, etc. So when you go all in, uh, you actually shift the culture and the values and the value proposition. And you know what's what's really valued in the organization.
But there will be areas where it doesn't make sense to in source, for instance, the service just because you're going to have a [00:20:00] DevOps or a product oriented agile way of working. You know, you would need to completely continue to procure that and run it as a service. And then you need to make a decision on where does it make sense to have this high agility, to be responsive.
So I think there's a lot to it to do this iteratively. So you get the feel for where is really the change going to happen and where is it material to get it right? Yeah, because people forget or maybe lose sight of a Spotify or Netflix for their technical brilliance in what they deliver. They have one thing to deliver.
They shunt content and they have a simple billing process authentication, but they focus on one thing. Whereas most large enterprises have multiple lines of business, multiple categories, they serve all sorts of different eclectic requirements, and that's where the variation needs to kick in. So take the ethos, but absolutely it must be adapted.
And I think sometimes we forget the simplicity of what they do, albeit it. underpinned by incredible architecture. [00:21:00] I think that's also the reason why Spotify, it's not a model. They just started to collaborate in that sense. And then people ask, how are you collaborating? That somebody took it into a PowerPoint slide and then suddenly it's a Spotify model.
Uh, but it just fits in their way, in their culture, in their DNA, in the way they want to bring products to market. Yeah. People got wrapped up in the, what they managed to achieve. And it was great, but, um, you know, it's the classic horses for courses, which is. You do need to adapt, there's lots in the thinking that's great, it's just the, the depth of that needs to be considered.
I think it's also, you need to recognize it's a journey, I think sometimes, you know, people complain about middle management not being, you know, change ready or they're resisting change, but I think there's like two lines of thought. When you do a change one is it's very prescriptive. Somebody comes and tell you this is exactly how you're going to do it.
You will face a certain resistance of that because it may not fit right then people may not have been part of [00:22:00] developing it or you would say here's the framework and we would like now middle management and the team within this framework to come up with the ideas that works for you. And then we will consolidate and sort of harmonize as we go forward.
And Ericsson is very much in the second pocket. You know, we have a framework. We have a direction. We're working it out. And remember quite a few of the teams have been working a product oriented model for like three or four years. Highly advanced. It's just that they have different toolings and done it in different ways.
So it needs a bit of harmonization, but there's quite a lot of experience here to do it. And uh, I think one way of harmonizing that is really to allow the teams to work with each other. Work out the processes within that boundary and framework. It gives more of a commit, I think. Yeah, and I think if you're trying to engender a culture of empowerment, and entrepreneurial ship and self starting and kind of moving forward as a team, then it's important, isn't it, to allow them to find their own way to do that?
I agree with you, [00:23:00] if you come in with a templated solution, one, it's probably not going to be as good as the team who are actually on the ground are going to come up with. And two, it's removing some of the empowerment right from the outset, isn't it? I think as well there are some boundary conditions on this.
We're looking very hard now to streamline the backlog tools so that we can put backlog items into each other. It's going to be transparent. Everybody can see and some backlog process KPIs can be done. So we'll not sort of relinquish all control and transparency. And I think you need to be clear on those boundaries as well as you move forward.
So I think you need to be subceptive to, uh, To what, what is negotiable and what is something that actually is there for reason and be able to articulate why you are rigid on some of these boundary conditions. Yeah, I think what I've heard in the conversation so far is that the mode of transformation you went through was took an area, transformed the area into, into product based teams, ran with that for a bit to make sure that the model was [00:24:00] robust and then started to move everything else over to that.
Way of working. I'm interested in the experiment or the exploration you did in those first areas. Did you have like a Hypothesis and then a measurable set of KPIs that you that you use then to say actually yes, it's functioning It's burning down backlogs faster It's getting our time to market down or were you looking at a series of other parameters to sort of judge its success?
For us it was, um, I think it was different in different pockets of, of enterprise IT. I think we were three units that said, you know, we want to have part of our organization moving into products as a spearheading that. So one was the AI and data section and one was digital workspace and one was my area then, which was at that point called IT cloud services.
But now it's called tech foundation with a little bit bigger scope. But. A lot in my organization, the reason why we went was that we modernized the compute platform and we realized that the hyperscalers were [00:25:00] releasing, you know, 1, 500 to 3, 000 upgrades every year. How would we manage that under the ITIL process?
It doesn't work anymore, does it? It would be easier to invent a time machine, go through the year, do a certain number of releases, then get in your time machine, go back, go through the year again. And then if you do that loop a couple of hundred times. You might get a chance. You might stand a chance of getting to that level of output.
And we had a lot of discussions on this. So we were saying, you know, this is 20 years of, um, you know, culture and process refinement. And how would we, how would we handle that pace of change? And also, how would we be engaging with the business and being, you know, customer focus because we were controlling the key for public cloud for the entire Excel, not just for enterprise IT.
So that meant that we had to be customer focused for real, because they were doing, you know, products and service line development here in R& D development. So we had to find a way. And then when this [00:26:00] opportunity presented itself, we said, yeah, you know, let's do Greenfield here. Let's, let's go Greenfield.
And then let's do backward compatibilities towards ITIL where it. made sense. So we handshook those things that we had to report into the back, the old model. Uh, so that's how we did it in, in, in my unit. And I think the other units did it for similar reasons. They wanted to move faster or they wanted to challenge a bit.
And then as we did that for a couple of years, we learned, but the last, um, we actually took in our Ericsson strategy team, which did a great job in, in, uh, looking at what was the latest theories and learnings in the market, then challenging how we've been working and then, you know, taking it to, taking it to another level.
And so this is now the platform or the thinking and here's the values that we're going to pursue. And then everybody starts to align to that in the second phase. So let's talk about the situation where you are now and how you're going to go about tidying the model up. So you've now transitioned into products [00:27:00] and platforms style structure and I believe you're looking at.
How do you automate huge areas of this model now? So first of all, give us your take on that tidier process and how you're getting your heads around what's going to be automatable. Yeah. So I think, um, what we did was we always had a focus on, you know, everything as a code and, and how we could make things more efficient in, in public cloud.
And then we had two tracks, right? One was really to automate What we had, which is in our span of control, we can automate how you provision infrastructure, how you decommission, how you control, etc. And then you had the track of modernizing applications. As everybody working in infra technology foundation know, it can be somehow difficult to get control over the change windows and the willingness for applications to focus on modernizing their applications unless it give real You know, real benefits to the consumer of the application, of course, right?
[00:28:00] So we said, let's focus on hyperautomation, and that sort of coincidentally was at the same time, more or less, when Microsoft announced ChatGPT. And we were then thinking, okay, so can we get automation on steroids and call it hyperautomation by infusing AI into this? And that's where it started. It's like two years back, we started sort of strategizing about how to do this.
And I think we are. well on our head to do a lot of good stuff here with automation. So let's start with how much of the operation that you look at at the moment do you think is automatable? So let's say a hundred percent of the activities that IT are doing, some are dev, some are infrastructure, some are service management, some are workplace management, some help desk, you know, the full stack.
How, how much of that hundred percent in a, in a hyper automated world do you think the outside edge of automatable is today? Today, I think quite a lot [00:29:00] doing in. If I first comment on service desk and intro, we can talk about development later on. Um, but service desk and infrastructure. I think what we've seen is there is examples on like Broadcom.
They have a serial service desk agents. Um, now I think they've been operational for at least a year and we see these examples where actually the human in the loop has been being reduced to just Okay. continuously automating by introducing automation. So our target is to automate 80 percent of our operational activities.
And how do you measure that? Well, you can measure that in in certain areas. You know how many of the tickets was actually sold by an automation script or an agent or an AI assistant. So how much of your infrastructure is now provisioned by automated script or blueprints versus manually configured. So.
What we're doing is we're really targeting an estate where the infrastructure is fully provisioned by CICD [00:30:00] pipelines and validated and tested also decommissioned by CICD pipelines and for this year, our target is to start. Implementing modernization on clusters of applications and the reason it's cluster of application is we would like to revoke production access and as we revoke production access, we would fully automate like the operational environment of the application and the infrastructure.
So at that point, I think we will see a lot of incidents going down. You know, typically we say 70 to 80 percent of the incidents is coming off their change window. So how much of that can we get away if we get away the manual? Challenges and we have a big test database that we can run scripting versus before we deploy and I think that's a huge potential and that is before we start debating and discussing how we can infuse AI into the hyper automation track, right?
So that is where we're now testing [00:31:00] and trying out. So we think over time. The operational part of IT is going to be commoditized to a very high degree, very high degree. And I think it's going to go rapidly, especially now with agents. Yeah, I was going to come on to how agentic intelligence fits into this, because I sort of see that as being an important, almost final part of the puzzle when it comes to the automation of underlying tech in the way you are describing.
See, exactly as you're saying, you know, cloud tooling, dev tooling, cloud tooling. I've got huge amounts of configurability these days, and automation within, you know, the native tool sets. So you almost don't have to do anything other than spend the time to, you know, to automate and configure properly. Then it looks after itself going forward.
The problematic part has always been the, kind of, the nature of something like a service desk or something like that, and I always saw that you could, you could, There was a [00:32:00] very automatable route for things like incident, particularly in the, in the infrastructural back office space where you've got like, you might have a server out or a, you know, an instance down somewhere.
Well, that's very fixable in an automated way now. And you could even automate that through to the help desk, for example, that could send out alerts, as well as the, you know, the service itself, pretty much self repairing. So you can sort of see how that track, Automate, again, a lot of discipline in getting that working, but you can see how that gets there.
And then it's always been that layer of, well, where's the human in the loop? And it's generally when it's talking to other humans or managing a service desk, but with the Gentic coming along, that feels like it's that final jigsaw piece. Is that broadly how you're holding the stack? And just to note, Rob's still trying to get his head around agentic intelligence, so when we get to that bit, Johan, you might have to talk a bit slowly.
Oh, okay. Love you, David. Love you. You're welcome, Robin. I think, uh, so the way we look at it is the four [00:33:00] streams, right? I will come back to your question soon, but I think just laying out the basics. So the way we look at it is we really need to get Everything has code, which is really the infrastructure in the application.
We then really need to revoke the access into production. So we get touchless operations. That is quite straightforward. In parallel to that, you need to sort out your data foundation, because automation and AR will run on your quality of your data. So that, you know, triggers you to sort of start running your observability.
So with everything as code and a data foundation and observability, you will get far. Then the magic comes, right? How quick will the assistants and agents actually change the way we work and revolutionize? And I think here you asked me for what is the use case? How do you see this? And I, I tend to see that The use case that makes a lot of bang for the bucks is in the sales process.
So I think the way when I talk to my [00:34:00] peers and when I talk to our business is how, how, how should you look at your process, right? And it sort of boils down, it's my interpretation to try and look at your incoming request, either if sales or if it's operations as an API call. Let's, let's try and configure the input part to be an API call.
That's, of course, takes a lot of effort sometimes. But imagine that's true. Then an API calls comes in. It goes to the first team that actually takes this on. If you go there with a AI agent and you, or an AI assistant and you interview the team. So you now start populating a database with And you ask them, what are you doing when this comes in and they explain it to you and the intent for you is, can I codify this?
Can I remove this stage? What are the blockers that actually calls for manual intervention? And then you backtrack and you step this down. And you [00:35:00] constantly record it into a database where you then can train your AI assistant. I think you sort of start building up a knowledge about your process, about how everything works.
And then you have to go back and iterate and see, what about this can I do as a code? And one about this might be an RPA or some other tooling for it. And as your AI intelligence gets better, it will start proposing for you as well what to do. And it can potentially also connect some of the dots that you cannot do between processes.
So I think we need to start thinking about our problems in a different way. I think, I think you've, you've, for me, you've hit the nail on the head with the, once you can get to the point where you can describe your need to the thing, The agent, whatever we call it, and then it can go and create the answer for you or do what you need it to do.
You've got to a very different world. There's this thing going on in the development community, which is, as the developer cuts the code, they describe the need for what this code needs to [00:36:00] execute in. And then the rest of the process is abstracted away, so it then goes into an execution engine and it runs.
And I think that ability to describe what I need and then the system going, um, Work out the rest is absolutely where you get this incredible level of automation that can occur with like. you know, very complex things as you describe it. I think you can. I typically talk about two use cases that I like.
One is a really basic one, which is a U. S. company. I will not mention their logo. I don't know if I'm allowed to. So what they do is they sell Facility stuff, right? So they do carpeting, they do lots of stuff. So they have lots of field visits and they have lots of service in relation to that. You know, they have paper, uh, paper and they have first aid kits.
They check some of the fire systems and alerts and alarms and stuff like that. Validating stuff, certifying stuff, right? So they do a lot of visits, but They've had a quite decentralized model and spent a lot of time on centralizing. And [00:37:00] the reason they centralize is, of course, to be able to drive upsell.
So what they've done is they have a huge fleet of people visiting sites, right? Thousands of people. And they now have real time feedback through devices on what is actually being replenished. And what they then can do is they can run, uh, a lot of their. Customers versus their product and service offerings and you have an AI engine looking at the bills coming in.
You're looking at the offerings you're looking at when is something was lately lost, supplied, etc. So it's a super active agent pushing data to the sales team and to the field teams to check, you know. We haven't changed paper for a long time, you know? Check it. The carpet has been there for quite some time.
Is the garment okay? Do they need to buy something new? When they buy something new, have we checked if they changed something of the logos? You know, is there something that they're doing as an advertising thing? So, the AI engine gets very potent, and it has skyrocketed their sales, you know. They, they [00:38:00] cross sell, they increase their margin, they, you know.
So, I think this type of, you know, Quite simple AI cases is where you can start, you can get a lot of money in the sales form. I think you can see as well from models like that and also the, the use of AgTech in the IT operation. This, to me is beyond say, the copilots of the world, like productivity enhancement.
This is gonna be the scaling engine, I think for it across large scale organizations and the real sort of transformation driver. Well, the, the, the thing about you describe is often we know. as a human, we should go and check things, especially like the sales process, query things, investigate, understand data.
But we don't have a lot of time. It's time consuming. And sometimes as humans, we just go, Oh, I'm tired. I'm going to go. Whereas the agent you described never gives up and just saying, look at that, look at this, do that. I prepared this for you. Ring them, talk about that, this, that, and the other. And it's that bit.
It's, it's, it's just filling in the gaps for the humans, which they probably could do, [00:39:00] but it's so much effort. And they've got so much on, it's a massive productivity boost. Isn't it? It's that thing where you can suddenly get huge results out of something that's constantly working for you in the background, taking away what can potentially be mundane, or maybe not things that you don't think about.
And that's where I get, I think we'll get the, the, the human. And it's this access to massive amount of data, right? You know, you could probably do this type of cross selling and say you should, you know, look at this sales part and this sales part, but because it's real time, and it can actually look at the bills from that customer, it actually knows.
You know, you haven't sold product X, Y, and Z. Why not? And then that gets very transparent and you, you know, with transparency, you can also take decisions and act. So I think it's a very good example of how you can quickly revamp it. And if you look at where that data is typically stored in a large company like Ericsson, it's probably can Oracle or an ARP system like SAP or something like that, right?
And you already have the data readily available. It's just, you know, [00:40:00] Connecting the dots, putting an AI engine into it, it's not a huge investment. So how do you get going on this journey? So there is a tipping point moment in the way that we've been describing, which is once the AI can start to learn and it becomes then kind of a positive flywheel effect on how then that's gonna that's gonna drive forward not only your Automation, but also the productivity and business value in the way you've been describing.
But to get from a, you know, kind of low automation or even zero automation, and let's just focus on IT, IT estate to something that's 50, 60, 80 percent automated, is a big leap for a lot of organizations. So how did you execute on that? Oh, we're not there yet. It's our trajectory. So I mean, we're trajectory towards 80%.
I would say that we're well on the way, but you know, we're taking the first step on the infrastructure, but I do get this question though. So I think it's, uh, uh, you know, the CIO [00:41:00] and, and other. peers to him is asking, What are we doing then in terms of AI, right? So one of the first thing that we're doing is we're using AI in a data ingestion.
You know, you need to have logs, events, metrics, and tickets, and you need to have quality of that. And there's quite a lot of AI that is already readily available to ensure that you have data quality. So you should use that. And that's something that we are looking into and doing. Then we all get flooded by false positive alerts across the estate.
You know, fantastic tools and AI is already available on that and, you know, utilize that we're doing it. Next thing is really this, you know, auto classification, prioritization and triage of tickets or workloads. You know, fantastic support already now. So, by leveraging that, you get the custom to sort of start using the output.
You know, because if you do a good AI installation, nobody's using it. Like, you [00:42:00] install AIOps and nobody's actually executing on the advice or enabling the, the, the human in the loop to reduce, even though you have evidence that you could. Then, you know, it's no use. So, you, you need to get that. Experience and ability to actually execute.
And then I think in the end, what we will see or what we're doubling down on is that we will see agents coming in on different areas. So you would have potentially one for network, one for cloud, one for service desk, or we'll see if they overlapping. All of those would either be agents by themselves or controlled by a super agent, and that we don't know yet, so that we are experimenting on, but we see a lot of initiatives coming here.
Do we call the super agent Skynet? Are we going to go there? We're going to create the super agent and that's it, we've actually created Skynet. I think it's a good question, right? So as they start, you know, if we get LLM models like agents talking to each other, where's the [00:43:00] limit? Where's the limit? You know, it's a good question actually.
Isn't it? Isn't it just, well, let's not go there in today's conversation, but that is an obvious, that is a, that is a thread that we could be on for a couple of hours. Maybe just to bring today to a bit of a close, what I'm interested in is. We talked a lot about maturing product and platform organizations and automating within those structures, but there'll be a lot of organizations I'm guessing will be a number of years behind where you've progressed to and then they're probably looking at their I.
T. at the moment and thinking, Oh, my God, I've got technology modernization to do. I've got cultural change to do. I've got so many moving parts. How do I even start on a journey where I can get to a point where I'm thinking about 80 percent of Automation. So just to bring us to a close, what advice would you give organizations that are really just at the foothills of this journey.
How do you cut through the noise to actually start to drive your change? I think from [00:44:00] when we talk about this, and I think this is from all business aspects, my advice would be revamp your compute platform so that you can harness the power of the hyperscalers and modern technology. If it's not the hyperscalers, you know, where do you harness the new technology?
Where's your. ingestion of new technology strategy. It's very important because you can't invest and build everything yourself. You have to go from a builder to a consumer, uh, and configure type of mentality to keep up with the speed. Um, and then you need to paint your North Star. And the North Star we painted was really, you know, Everything is code and as people start realizing that everything is code is making it more efficient, freeing up time to do things like development and things that a lot of people thrive on.
It's you get a lot of traction with the teams and they will come to you and they will tell you what they want to do in investments that they need to do in order to get everything as a code. [00:45:00] And as you're on that on that trajectory, if you then start doing these use cases where, um, you try to infuse AI and AI can then revamp the processes or revamp everything as code, then you start getting a lot of 10x out of the just everything as code type of automation.
So if you get an automation that is then looking at the data, making proposals, being able to actually. redeploy or recode your existing code to include new functions and features and constantly evolve your evolution of hyper automation, you will continuously commoditize over time your stack of operational activities.
So I think summarizing it is ensure that you have access to latest technologies. You can actually infuse your. operations with it. But in order to harness the power of those technologies, I think everything as code is key to [00:46:00] actually get there.
Ez, what have you been looking at? Yeah, so I want to talk to you about showcasing value. I've been talking about demos before and I've been at a conference. This is what Rob thinks he does on a, on a day to day basis as he does hashtag value added every day. Hashtag showcase. Hashtag action. Hashtag value.
If you talk to yourself in the mirror every morning, you're going to add value today. I'm sorry. You actually do that, right? You write it down. Where I stand up and go, go on Rob, you can do value today. Go on. It's like, yeah. Lipstick on the mirror. Yeah. Yeah. Anyway. God. Very imaging. So I was at the conference this week and obviously they all want to convince us of the power of agents.
Right. And really showcasing it is sitting behind the laptop and being able to build an agent within five minutes yourself. [00:47:00] Which is fantastic. Absolutely. And I loved it doing it myself, of course, as well, but it's all smoke and mirrors, to be honest. And we all know that in terms of demos and, and showcasing technology, we've been doing that like forever.
And I think it's also one of the reasons why some people have like this love hate relationship with it in general, you know, we paint the bigger picture, you know, please sign here, you get the golden tickets. We see the demos with every details in place and it's amazing. And, you know, reality hits us hard.
So I'm just wondering, and it's going to be tricky, I think, with this agent story that we're now in the midst of, because a lot of people are talking about business cases. But I think everything that Johan also explained, it's really understanding your current business as well. We do not need business case, we just dive into how do we work and do our business today and how can we really increase value end to end instead of now starting to look up for business cases in the organization.
I'm [00:48:00] just wondering how you feel about that because it feels a bit like we know what we're doing right and we should become better at what we do every day. So I think, I think for me It's, it's always that there's somebody always wants a business case that's tangible. P times Q. Yeah. Blah. Right. But there's these other things where it's like, you know, it's going to add value.
You know, it's the right thing to do. It's the leap of faith moment, which is we're all sitting here going, I'm pretty confident that the agentic concept is going to make massive improvements in how we operate as a business. And Joanne, you described the Yeah. The sales process one extremely well. It is that we just need to get into this to know that it is probably a foundation of our future.
Don't stop, just go. I think if we get, if we get wrapped around the axle of the business case and it has to perfectly add up before we start on this new technology journey, then I think you're going to get left behind. Basically, I think there are certain nodes. You just go, it's just going to work. Just get on with it.
Everyone else is buying power. It's the future. How's that for you, Johan? Are you using demos to, to, to [00:49:00] have that business convinced in let's try this? It's very good with the product oriented model and agile, right? Get your organization to do demos and there's a lot of demos and yes, we do. There is this case that we could have talked about when we want to deploy and try out this for SAP and the transformation to rise.
I think there is a lot of opportunity there. I don't know if we have time, but If, if we talk about use cases, I think there are also known use cases. So as you innovate, don't try to innovate. to blow the market out. If you're an IT organization, look what others have done. We know Service Desk is a perfect example to start at.
We know infrastructure provision is super good as well. You know, if you do things, if you still have teams configuring and not codifying their configuration and provisioning of infrastructure. super good way to start. And the third one is where I think agentic really will come in is when you transfer knowledge to different languages, [00:50:00] right?
So if you have an old programming language, you don't need to do a modernization. You know, the bridge of transferring from a language to another language, infusing knowledge is also a very, um, tested and validated area for this. So as soon as you have legacy. software that you need to modernize, I would go hard on looking if there's not an ability to use agents for that.
I want to go back to the point you made around business cases. And I just wonder, is there a more abused and under leveraged and underdeveloped middle management tool than the enterprise business case? To me it just feels like people using it to create some illusion of certainty and an illusion of control and using it as a, like I said, a middle management tool rather than something is sort of genuinely [00:51:00] trying to drive an organization forward.
And I'm not saying that the intent of the people that are trying to develop them is not good or trying to move an organization forward. But it seems to me that the governance around business cases is entirely stifling. To your point, Rob, You know, the decision making is as good as the decision makers and whoever wrote the business case.
So if you ain't got somebody writing your agentic business case that genuinely can express a future that's radically different from what you've got to do today, and then mathematically try and prove that, which actually could look ridiculous because you could go, Well, actually, there's 30 percent of what we do today is going to be agentic.
So, like, this thing writes itself, but you'd put a series of numbers down, then any middle manager's going to look at it and be like, I don't believe that. You know what I mean? I don't believe it. But it's a very good example, actually. If you look at digital organizations like we've described, get it, and they plow forward.
The supporting functions in an enterprise [00:52:00] organization may not have come up to speed with the new technology. thinking. So commercial processes, financial processes, all of this sort of stuff. And sometimes they're not aware of what technology can do for the future. I think there's an education process in those functions to say, guys, the world's changed.
You need to think about business cases differently. You need to think about commercial models differently. You need to think about how your finances flow. Classic one, legacy accounting over the top of a product based operating model doesn't really work as well. You need to change your financial processes to make a product.
Yeah. Well, it's those I think that's parts of the organization that we need to maybe go give them a bit of a hug and say the world's changed. Let's go give you a bit of an education on it. I think there are tangible things that we could also predict, right? So based on our experience, we can predict if we, you know, if we automate patching and vulnerability scanning and stuff like that, we can easily translate that to the effort that we're doing.
But if you, if you do like we're aiming for touchless operations and we move beyond that into [00:53:00] AI. The big question is, if you are running the organization and you're, I don't know, you put in 100 percent effort to it, how much would that effort reduce if you do all deployments automatically and you start, you know, Resolving tickets with a I is the effort going to be 60 percent 40 percent across the whole estate if you had 400, 000 tickets and you cut that to 1100, 000 tickets, how much of how much of the size of the organization do you still need to retain to run it at the fourth of the workload?
So how do you How do you construct a business case for that? And how much are you willing to stick your neck out in order to say this is what you believe in? Yeah, right. As nobody has really tried it. Nobody's really gone all the way, I think. We're all spearheading down a lane that we think is actually going to continuously evolve and that technology is going to carry us forward on the promise that it's going to be better in 12 months from [00:54:00] now.
So, but when is that inflation point? We don't know. Let's leave that in terms of the conversation today on that note, because I think it's, it's a great note to end on because it's, it is one of the big questions for tech as an industry today. And let's face it, tech in the past, to Esme's original point, is not always fully delivered on its promise, has it?
So, are the humans going to get in the way again? Let's, uh, let's see. Let's see. Johan, thank you so much for joining us today. It's been great spending Friday morning with you and talking about that. Thank you. Fantastic. Thanks for having me. Our pleasure indeed. And now we end every episode of this podcast by asking our guests what they're excited about doing next.
And that could be going to a great restaurant booked at the weekend that you can't wait to go to, or it could be something in your professional life. So, Johan, what What are you excited about doing next? I, I, I just going to take this top of my head. I don't know if it's so excited, but you know, we just got the weather forecast and it's going to snow this weekend and I [00:55:00] put on the winter tires on my bike, so I'm one of those that's going to bike the whole winter.
So I'm just gearing up to how will I survive next week when the first snow comes? I just can't get that out of my head. I don't know if it's that pleasant, but you know, that's top of mind. Have you used it as an excuse to buy a new outfit? to go cycling in, like a wet weather. No, no, I used to buy a new bike, a new bike.
Dave, dream pig, think big man, big pig. That's the outfit all of you. There's a business case here. Yeah, yeah, exactly. Well look, enjoy that. First snow of the season. How exciting. Um, we don't really get snow in the UK anymore. It's definitely one of those things as a child, I remember snow being quite deep, but of course I was much smaller then, Rob.
You could say that. It's like the Curly Wurly argument where you say Curly Wurly used to be bigger and I went, no, you just used to be smaller. Yeah, yeah. Big cow, small cow. No, that one's just closer. It's one, uh, it's one fun with that is, uh, when, when, uh, [00:56:00] before the snow, the roads of with bikes are fully cloaked, right?
Tons of people. First snow comes, you're more or less alone. Yeah. Yeah. It's me and some, one more mad person on the road or something like that. Well, I, I admire that, Johan, the tenacity to go through the difficult weather and still do it. Not like those fair weather cyclists. We get a lot in my, uh, we get a lot in my area and, uh, it drives you mad.
Anyway, thanks again, man. Good to see you. Good to see you. Take care now. Have a nice Friday. Yeah.
Thank you so much, Johan, for this week and bringing your insights. And thank you also for our sound and editing wizards, Ben and Louis R. We actually have our super agent, super producer, Marcel. Thanks again for all your effort this week and of course, thank you all our listeners. We're on LinkedIn. We're also on X, so please connect, DM of [e-mail] and subscribe on the podcasts and see you in another reality next week.[00:57:00]