IT Portfolio Management Across the Group
DVV is a group with many parts: Stadtwerke, DVG, DCC. You’re responsible for the IT portfolio. How do you manage to shape a consistent IT portfolio in such a complex group structure?
DVV Holding has just under 500 employees; the entire group, with 27 subsidiaries, has more than 4,500. We’re the third-largest employer in and around Duisburg. Our central IT department has a good 200 employees, split into IT operations and IT service management, which includes my area.
We’re set up as a matrix organization: central IT is responsible for the entire group, and IT portfolio management steers across projects and products. In personal union, we’re also Duisburg City Com, or DCC for short, the infrastructural backbone of DVV, with our own data centers, as an internet service provider, and as a driver of broadband expansion. From the group’s strategic goals, we derive a portfolio that’s built on the same infrastructure and architecture across all subsidiaries.
Rüdiger Strelow
Process automation plays an important role in both of your areas. Do the two areas work together on this? Is one further along than the other?
My department has two teams. One provides the tools: Bizagi for workflow automation, UiPath for robotic process automation, plus ETL processes for the data warehouse and BI with SAP Analytics Cloud and Power BI. We also centrally manage the ITSM platform, currently Omni tracker, moving to Matrix42 in the future.
The other team steers the group’s major IT projects: prioritization, project manager resources, and large transformation projects like a SAP HANA ISU rollout. Both teams work closely together, for example on the ongoing migration from Omni tracker to Matrix42, where my second team is already taking on subproject responsibility.
I see many municipal utilities that, without a holding structure like yours, must orchestrate all their individual automation and AI initiatives themselves. Is your structure actually rare in the municipal landscape, or do you see it elsewhere too? Do you have a best-practice structure?
We confidently belong to the top group here. As DCC, we operate our own infrastructure, which isn’t that common. We run a hybrid cloud strategy: first we check what can be run in our own data center, and only under strict regulatory requirements do we move to cloud providers.
Being able to offer everything from a single source, from physical assets to the application layer, gives us speed and synergies. Others must commission various service providers in a fragmented way; with us, almost everything comes from one hand.
Do you have a flagship project where AI and automation haven’t just been implemented, but genuinely used and made a real difference?
Two examples. First, DVV-AI, our own platform, which we’re currently rolling out productively across the group. A real game changer: with an eye on data protection and BSI requirements, everything runs within our own group boundaries: our own data center, our own servers, our own lines. Netze Duisburg already uses it too, with a mirrored instance due to unbundling regulations. More subsidiaries will follow.
Second, the automation project R2D2 with UiPath, which started a year ago. What began as one or two planned example processes turned into 17; by now we’re implementing just over 100 processes across the group, with the number still rising. It was important to us to optimize processes along the way, not just automate them one to one. We compare throughput time and effort before and after and translate that into economic metrics. We now use the same approach for AI requirements too, for example efficiency analyses for our project management tool Asana.
Data quality is currently one of the biggest obstacles to automation and AI, and without the right skills on the team, even the best platform doesn’t get you far. How have you experienced the cultural shift over the past few years, not just at the holding level but also within the individual companies? Is the acceptance there, and how have you supported it?
Yes, absolutely. We’re confronted daily with requests for GPT licenses, cloud AI licenses, and automation, which is a good sign. We’re also seeing more young colleagues joining us with their own skill sets. In the past, it was often just, IT will automate that’; today we experience real sparring partners.
We channel this demand into clear governance structures; we’re a critical infrastructure company and can’t hand out AI licenses indiscriminately. That’s why we offer training: through the DVV Wissen platform and with office hours on developing your own AI skills. One concrete flagship project: using Bizagi, we digitized work slips for our service provider Okteo; around 15,000 a year no longer run on paper.
And: in IT, we openly celebrate both successes and failures. If something doesn’t go well, we say so, get back up, dust ourselves off, and keep going, learning as we do.
Do customers notice this too? Do DVV’s partners feel friendlier, more optimized, faster to end customers?
According to our customer surveys on IT: yes. Whether that shows up the same way for end customers of the Stadtwerke or the grid operators, I can’t judge, though I’d like it to. Our main priority, though, is that our initiatives benefit the group.
I think we’re currently experiencing a paradigm shift: going forward, companies won’t be asking which processes to automate, but deliberately which ones not to, and for that they’ll need AI competence, machine learning competence, and energy-industry data competence. How do you see this from DVV’s perspective: what are the decisive strategic decisions for the next ten years?
First: agile portfolio management instead of rigid five-year plans, regular prioritizing and adjusting, even within annual business planning. Second: an independent digital infrastructure, real control over critical infrastructure. And for AI and automation, you need a data platform and data literacy: knowing what data you have and clearly defining who’s the process owner and who’s the data owner. Bad data is the foundation for bad automation and bad AI; good data is the foundation for good work with AI models.
Thank you, Rüdiger, for the conversation and the deep insights into your AI work.
