Agentenbasierte KI für datenzentrumsfähige Versorgungsnetze | IQGeo

Welcome back to Bite Size Electric, where we unpack the ideas and innovations that are reshaping how utilities work, one bite at a time. I'm your host, Brandon Kirkin, and today we're diving into one of the biggest shifts happening across the industry, geospatial work execution. Geospatial work execution is a process in which utilities connect their field crews, office teams, and enterprise systems around a single live view of the grid. It's about moving from static digital twins to living, adaptive networks where every job inspection or asset update happens in the real world context of the grid. To help us unpack this, I'm joined by Adrian McNulty from IQGeo. Adrian has been working with utilities around the world to help them transform their operations using geospatial work execution and AI driven workflows. Adrian, welcome to Bite Size Electric. Oh, thank you for having me, Brandon. So Adrian, I think the best place to start would be actually defining what we're talking about. Can you explain what geospatial work execution means in practical terms? Yeah. So, geospatial work execution, it's really about performing the work in the field, but think of it as having full geospatial awareness, awareness of the network model, but also workflow intelligence. You know, there's a lot of traditional systems out there that they're good at maintaining asset records or work management systems that tell you where to go and who to go to. No one's really combining it all together. So think of combining all in one application, the context of a network model. So the underlying data model, the attribution, Why it's really profitable. massive push around electrification and renewables. All of this is leading to extra work in the field, whether it's design work, as-built work, inspections, response to unplanned work. When I talk to utilities, it's not uncommon that we see several different applications covering different types of work. It could be as simple as planned versus unplanned, but even further breakdowns, very siloed work types in different applications. But I think it's really important that we focus on making it easier for the field crews to perform the work. They could have multiple different types of work to do on a daily basis and may have to context switch across different applications. But I also think why it should be now is there's going to be further technology transformations that the future platforms need to be enabled for. And by that, I mean AI and how's that going to work in the field in the future. So I know that utilities already have a lot of existing systems like GIS, work management, asset management, but it sounds like these tools aren't really enabling geospatial work execution. So what's missing from these traditional tools that's causing these challenges? Yeah, that's a great question. I might answer in a slightly different way if it's okay. I don't think it's what's missing in the existing tools, but I think it's more of a case of combining or consolidating different work into one application. You know, it's not uncommon that you see multiple different mobile applications and many of them were or are extensions of a desktop application, right? And then mobilized. I think what's missing is having a true mobile application, one that's designed for the field. Easy to use, works online, offline on any different device. You can give it to contractors, but combines all the different things that you need to do in the field, whether it's located an asset and being able to trace upstream and downstream to see what's connected or what customers are connected, right? Whether it is, hey, I'm in the field and the network model said there should be a poll here, but there's no poll here and doing data edits and corrections and making those get back in. But also then the different types of work that you could be assigned during the day and all having a view of it, all the steps and links back to other types of information, whether it's safety documents, whether it's inspection history, whether it's links to the actual asset documentation itself, all in an easy to use app and completely digital workflow. So everything you do is fully tracked. I think that's what's missing today and the challenge of having too many applications that potentially could have different data. And I think one of the big challenges, as you know, Brandon, the systems that we're putting in place today to actually control the grids require high degree of accuracy and up to date of the actual data itself. And the more different systems we have, the more prone we are to errors. So we can combine all these, digitize all of the work steps, make it easier for the people in the field. The back office systems will have better data as well. So I'm understanding that geospatial work execution is really a process. What would you say are the core components of geospatial work execution that makes it different from traditional processes? Yeah, I think you're right. It is a process. There's inter-end workflows that extend from different source systems into the field and back again. And if they think of the actual impact of the field, the different components, I think there's five. So first of all, it's the concept of having the full network model, being able to edit in the field, but having it up to date. If you make an edit in the field, being able to push it back into the source system as it happens. So reducing the lag of things that happen in the field, but into the back in system and then back out to the field again. Native mobility, I think, is understated, right? So having the ability to deploy it on different mobile devices, whether that's an iOS device, an Android device, or Windows device, but also being able to work through online and offline connectivity, or even low connectivity. We need the field crews to be able to continue their work regardless of what the connectivity is. The next piece is then the configurable workflows. So there's different workflows and steps of a workflow for different types of work, whether you're looking at a design or an as-build or an inspection or response to unplanned work. These all require different workflows and steps. So it's important that we're able to configure those. I think it's also now, I think AI is like the next frontier. It's important that whatever we deploy in the field enables AI in the future, or actually today, not just today, but in the future as well. I think the final piece is integration to make sure we've got the end-to-end processes. So generating the work orders, all the information around the work order, getting that into the hands of the field worker. The field workers perform the work in the field, but making sure that goes in to the back office applications or the source systems quickly and accurately. So there's no data lab and the data quality increases. Awesome. So this sounds really impactful at a high level, but can you describe what this looks like for a utility crew in the field? Like how does this actually change their day-to-day experience? Yeah, I think the best way to do it, let's think of a day in life, right? So imagine, Brandon, you turn up at work, you get to the utility crew for the day, and you sync your IQ Geo app. You open it up and you've got an up-to-date network model overlaid onto a map and all your assigned work for the day. When you get into your individual work orders, you've got access to all different types of information that could be specific to that work order, whether it's a permit, whether it's a safety document, whether it's different steps that you have to do to execute that work. But you've got a guided process within the application to execute the work from start to finish. You can take images. In some instances, you may have AI-enabled workflows where the images you're taking are validating work while you're in the field to give you the confidence that you've done the work right first time. And then everything you do is logged. You've got a full audit trail of every edit you make, every piece of work you execute, but a full digital audit trail that's integrated back into the back office systems. Great. Okay. So I'd like to dive a little bit further into something you mentioned earlier as one of the core components. And then again, about the field crew is the role of AI. AI is pretty much everywhere today. And it sounds like it's also a big part of this shift as well. How does AI actually enhance geospatial work execution? Yeah, I think that AI is the next level of transformation when it comes to workflows in the field. There's been a lot of work around digitizing current workflows in the field. So replacing paper with digital processes, going away from paper forms to digital forms. And don't get me wrong, these are tremendous value, but there's so much more we can do. Putting the network model, native mobility, and the workflows into the app for the field user, I think it builds the basis to add AI-enabled workflows and in the future, agentic workflows to support all different types of tasks in the field. So what does that really mean? And I think AI transformation can happen in a number of different ways. We can look at real-time validation, and this is available today. So using or introducing images into the workflow. So imagine you're installing a piece of equipment and the workflow requires different images to be taken along the way. When you get to the job site, maybe you're doing a meter install, you take a meter read, you take an image of the pit to make sure it's clean, you install the meter, you check the connections with an image, you check the asset number, all of this is done automated in the background and the model locally in the field, and it gives you that instant validation. The next step there is guidance. So we're starting to move into AI-guided steps along the flow. This could be flagging issue before they become a problem or reducing all the manual steps. Over time, I think we'll get into predictive workflows. So the AI models learn what's happening and provide guidance on what we should do. And we get into the future of true agentic automation. Imagine you're doing an inspection. The image identifies a crack in a pole, for example, and the AI agent automatically creates a referral to come and replace the pole based on the condition. So I think there's a lot of areas where AI is going to help. But I think what's important is to do it in a way that actually helps the crews complete their work on a daily basis rather than being intrusive. Awesome. Okay. So it sounds like AI can be helpful both in validation and also in the actual decision making process for some of these workflows. Can you give us an example of maybe a common workflow that you see utilities doing nowadays that you think AI could really help make much more efficient? Yeah, there's actually two I can think of. First one is we do quite a lot of inspections. Think of right-of-way or transmission structures. Even gas utilities do a lot of regulatory inspections. What we're seeing is a high interest in many of our customers actually are taking imagery of the, let's use transmission right-of-way as an example. They're taking imagery with drones as they do the manual inspection process. So they still go into the field with ITG on an iPad as they've got a set of forms and they're using their own eyes to check what's going on. But in parallel, they're taking images along the way of whether it's the span or the structure. So we're building models that will automate the inspection process. Initially, that will be done in a batch process. So allow them to continue with a manual process, take all the imagery, load it all in at the end of the day, and the models verify what was recorded locally. In the future, that could just automate the process so we could really get towards a drone-driven inspection process. The other one is around installation of equipment. So I mentioned meters earlier because I think it's a fairly common one right now. So using AI models to verify the installation of a meter has happened correctly and safely with the right permits or the meter routes. Those are things that we're seeing quite common today, but I think it'll go to other use cases in the future, such as as-built, such as damage assessment. I think there's a whole area where we could add this into. Great. Wow. Yeah. That really paints a picture of how powerful that could be. It sounds like geospatial work execution, this process is really setting up utilities for much more efficient workflows in the future. Is that where you see this heading? Could we eventually reach a point where the network becomes self-updating, like an intelligent grid that can detect and validate everything automatically? I think there's two questions there, I think. So one around the efficiency and productivity of the crews. So let me take that one first. So I'm going to give you an example. I sat with a utility a couple of months ago, and they were walking through how they do their current response to outages. And they actually showed me that they jumped between two or three different applications. One to get the network model and spatial awareness, and then one to fill out the forms related to the actual outreach ticket themselves. And what we've done, we've built a solution that basically all inside IQGL, all inside one application, but a fundamental network model, and WordPress specifically designed for outreach management. But then in the same application, there could be other work order types. So they can stay in one application to all of the works, driving much more efficiency, but also the data quality. Because sometimes you've got different data in different systems, so having it all in common will improve data quality, efficiency, and get the data back into the other systems. And Brandon, to the question around self-updating networks, I think absolutely that's where we should be heading. So as utilities start to adopt geospatial work execution and layer in AI, I think where we evolved from is more static data into an intelligent living system that helps us on a day-to-day basis. And that takes a number of different forms in my mind. So we've touched on it in different questions you asked me earlier, but like the real-time validation, as you do work, performing the field, using AI models and imagery to make sure what we're doing is done right first time, done safely. You can even take that a step further, having validation from the field before you go out. Imagine a customer wants a new meter connection and they take an image or a video of it first, they send it in, you know exactly where it's located, you know what's going on before you get to the job site. Then we get into like automation and detection. As you know, there's more and more sensors getting put on the grid today, inside the substation, outside of the substation, or automated equipment. So there's more and more knowledge of what's happening across the grid, all the way down to the edge. And the next area I think is around agentic workflows. The AI agents don't just flag issues, but can actually create what needs to happen next. So if it's a follow on piece of work, get it back into the source system and automatically generate, whether it's a tree trimming or a pole replacement. Leading on from the sensors and all the new measurement devices we have on the field, that enables us to be more predictive and to take more prescriptive actions around replacing equipment and different types of actions. And then continuous learning. The system will learn the more data we put into it and every job it executes, every image it reads. So over time, it's going to know how to adapt and make it smarter. So I think absolutely, I'm not sure how long it will take us to get to a full self updating network in the field, but I think that's the path we should aspire to. I just wanted to ask a quick follow up to what you had mentioned about how AI learns and gets a little bit better as you put more data into it and as it's exercised. How long would it really take then for this to be useful? Is this the kind of thing that, you know, you get this in the field, but it's not really useful for a bit until you start using it more and more and then eventually it becomes actually useful? Or is this something that you can actually use right away on day one? So that's a really good question, Brandon. So for our listeners, IQgeo acquired a company about Deepomatic early this year. They focus on visual AI, but specifically for utilities and telcos, more interesting is in the field as well. So validating the work that's been done. And traditionally they have been in the space for almost a decade. I think last year they processed something like a hundred million images so they can work at scale. And traditionally they've been more of a discriminative AI. So using very specific images to train the models over time. And that can take some time. So you're using the individual customers imagery of a good image versus a bad image and training the models. And they just get smarter and smarter and smarter, but it does take time. The benefit though, they're highly accurate. What we've done recently is blend generative AI. So think now large language models. They've got a mass amount of data that can help accelerate the training of the model itself in parallel to feeding the specific images from the customer's data set as well. So as a long winded answer of saying, we're now looking and deploying ways of accelerating the learning curve. What's important though is that we keep the accuracy for certain types of inspections. Adrian, thank you so much. This has been great. I really appreciate the details you put into your answers here. I have one last question for you, just kind of a quick wrap up one. For any utility executive or field leader out there who's listening and thinking this really sounds like the next big step for us. What's one piece of advice to give them before starting their geospatial work execution journey? Adrian, I think my recommendation is pretty simple is start small, but scale fast. And what I mean by that is, I think we need to look at a broad scale around how do we make field operations more efficient, both now and in the future. But trying to do it all in one project is going to be very, very hard. Where I've seen success with other utilities, they pick one use case that's problematic or an aging application and focus on that, but really focus on getting it into the hands of the field users very, very quickly and then iterate, make it better over time. It doesn't have to be perfect from day one. And then from there, migrate workloads into the same application. So rather than having a single project that lasts several years, you have maybe these quarter releases where every release is either an iteration on the workflow itself or a new workflow, new work order type gets put into the application. I think it's also important to start to look to introduce AI. Like it or not, I think it's going to be around and it's going to stay. So I think we could look at, or I'd recommend looking at workflows where you can introduce AI technology today in a non-intrusive manner. And I'd mentioned earlier in some of your questions, the inspections or the meter installation process, those two to me seem pretty good candidates to do an initial use case because the models are fairly easy to build and to train against what you would look for. And you can do it in line with the current workflow today without being intrusive and you'll get a fully automated QAQC process along with it. Great. Adrian, thanks so much for joining us today and for breaking down what geospatial work execution really means for utilities. No, thank you for having me, Brandon. It's been great to be on. And as we've heard, it's not about better maps or new applications. It's about connecting people, workflows and data in real time, and obviously using AI to make every job faster, safer and more accurate. To learn more, visit iqgeo.com and stay tuned for our upcoming episodes where we'll explore how visual AI is transforming each stage of utility operations, from design and construction to inspections and storm recovery. I'm Brandon Kurgan, and this has been Bitesize Electric. See you next time.

Also, IQGeo, wir sind ein Softwareunternehmen und bieten eine Geodatenlösung an, die sich in erster Linie an Außendienstmitarbeiter von Versorgungsunternehmen richtet. Wir haben also im Grunde alle Daten aus GIS- und EAM-Systemen neu aufbereitet und in eine einfache, benutzerfreundliche Anwendung integriert, die sowohl online als auch offline funktioniert und die Arbeitsaufträge den physischen Anlagen zuordnet. Im Versorgungssektor ist derzeit viel los. All die Veränderungen, die Sie erwähnt haben – die Modernisierung des Stromnetzes –, all das erhöht das Arbeitsvolumen, das von den Mitarbeitern tagtäglich bewältigt werden muss. Wir beobachten bei vielen unserer Kunden, dass sie entweder veraltete Anwendungen haben, die ersetzt werden müssen, oder dass die Außendienstmitarbeiter im Laufe eines einzigen Tages Anwendungen hin und her. Viele der Anwendungen verfügen nur über eine begrenzte räumliche Wahrnehmung oder sind nicht direkt mit dem Netzmodell verbunden. Mit IQGeo verbinden wir das physische Netzmodell mit den Arbeitsaufgaben und verknüpfen sowohl die Anlage als auch die Arbeit miteinander – und das alles in einer benutzerfreundlichen Anwendung. Dadurch wird die Arbeit für die Außendienstmitarbeiter vor Ort optimiert. KI steht derzeit im Mittelpunkt jeder Diskussion, sei es im Backoffice oder im Außendienst. Viele Versorgungsunternehmen machen bereits heute Fotos bei ihren Inspektionen, egal ob sie Stromausfälle beheben oder welche anderen Aufgaben sie gerade erledigen. Was sie jedoch nicht tun, ist, die in den Bildern enthaltenen Daten zu nutzen. Mit visuellen KI-Techniken kann man also tatsächlich automatisieren, was ein Mensch sonst manuell erledigen würde, und dies direkt mit der jeweiligen Aufgabe verknüpfen. Meine Empfehlung wäre, klein anzufangen. Man sollte den Arbeitsablauf finden, in dem bereits heute Bilder verwendet werden. So verändert man die Benutzererfahrung im Außendienst nicht, sondern automatisiert und vereinfacht den Prozess und mit weniger Aufwand mehr Arbeit zu erledigen. Wenn ich heute mit Versorgungsunternehmen spreche, ist meiner Meinung nach eine der größten Herausforderungen, dass ein Außendienstmitarbeiter möglicherweise mehrere Anwendungen nutzen muss, auf die er zugreifen muss, um seine Arbeit zu erledigen. Gleichzeitig versuchen sie aber auch, sehr komplexe Daten zu verbreiten. Daher glaube ich, dass es entscheidend ist, es den Außendienstmitarbeitern leicht zu machen, ihre Arbeit zu erledigen – und zwar sicher. Also, integrieren Sie alles in eine Anwendung, die online und offline funktioniert, einfach zu bedienen ist, ihnen aber Zugriff auf alle Daten gewährt und sowohl das Netzmodell als auch die Arbeit, die sie ausführen möchten, in einem benutzerfreundlichen Arbeitsablauf miteinander verknüpft. Also, schauen wir mal.