FROM THE INDUSTRY
How is AI affecting IQGeo’s direction of travel? AI is reshaping how we think about our role. Today, IQGeo is the system of record. We show operators where their assets are, how they connect, and what work needs doing. But as AI becomes more capable, our vision is giving operators’ infrastructure the ability to “speak” for itself. A utility pole or street cabinet can’t literally report its status, but if we collect enough data from photos, sensors, and other systems, AI can start to infer what’s going on, surface actionable insights to the operator, and eventually self-manage. Sounds like a lofty ambition. Well, yes and no. Sure, no one’s ready to let an AI push buttons that spend money or dispatch people without human sign-off. But that’s the direction we’re heading. Over time, as trust builds, we’ll move from supervised automation to systems that just get on with the job. They’re learning models; we will soon reach the point where it doesn’t need to involve a human. Correct. In cars for example, you’re relying on satellite navigation and lane-keeping assistance. We wouldn’t have trusted these 10 years ago, but now we do. The same applies with workflows. We wouldn’t trust an AI today to send a repair crew out because we’re just not used to it. But we will get used to it, and in five years we will wonder what we were worried about. Do you feel this is just an issue of trust or that the tech isn’t quite there yet? Well, it’s both, isn’t it? AI models are not necessarily good enough yet to be correct close enough to all the time, but they will get there. Like self-driving cars. There’s no way you’ll get me having a lie down on the motorway. Then again, when you get on an aircraft, you don’t think twice about the pilots engaging the autopilot —even though it
Dated formats, jumbled naming conventions must be a big issue. It’s a headache for anyone who has used all the different generations of word processors over the years. The modern malaise. It’s always the same story. The tech team wants to clean up the data, the business says they’ll sort it later, and it never happens. But that’s changing. Deepomatic can be used to flag inconsistencies as soon as something’s plugged in. When the AI processes a photo of a street cabinet might show a red fibre in port 20 and a green in port 27, while the system of record expects the red in port 6 and a blue in port 7, with no green at all. The image is the truth, so the system of record flags the mismatch and suggests next steps, like updating the record or sending a crew. We can help telcos fix their decades-old data problem. Will AI resolve all of this? There’ll be some happy medium with a human in the loop, at least to begin with. But the key shift is that we can now say, “here’s how reality doesn’t reflect your documentation.” Now you can start fixing up 20 years of legacy data and, from there, start thinking about using AI to parse and resolve data discrepancies autonomously.
can, I believe, land the plane. Or driverless trains, if you’ve been on the Docklands Light Railway in London. But back 50 years ago, driverless trains? We all would have said anyone risking that was nuts. How else are you employing AI at IQGeo? AI will be key to recognising and automating workflows. Today, there are so many exceptions that need a human in the loop to review, and this makes reliable automation difficult. AI agents will sort this because they can analyse patterns in the sort of edge cases that crop up and learn the steps humans usually take to resolve them. But AI needs to make these decisions based on reliable network data, so the accuracy of digital twins is paramount. Otherwise, it may spend money, or create work for field engineers, based on outdated information. Presumably your partnership with Deepomatic is a big part of that. Absolutely. Deepomatic’s computer vision software is helping operators to capture and maintain high-quality data that forms the basis of AI-driven workflows. This is especially crucial for tier-one operators, many of which have acquired maybe 15, 20 companies over the past couple of decades — and the data they’ve inherited is usually awful. It’s not just about verifying new work, but turning decades of poor data into something operators can trust and use. Deepomatic can therefore check what’s been done. That will add a welcome level of integrity. You know, it adds integrity on multiple levels, even with administrative issues like payments. There are all sorts of games contractors play to get paid, such as doing work to connect customers at the network build stage but not declaring it. Then, months later, a request comes in to connect a customer, and the contractor then claims the work again, this time at a higher rate. Deepomatic’s AI checks on the images act as a digital audit trail that proves when the connection was done, and captures valuable metadata to add to the system of record.
What is the long-term strategy?
We think of our strategy in three chapters. Chapter One is where we are right now, working with Deepomatic to help operators capture their data right first time. Chapter Two concerns fixing up legacy data. Chapter Three is where it gets really interesting: using photos taken to detect what’s changing over time. If we can see gradual wear or shifts in the infrastructure, we can predict when something needs replacing and trigger the right workflows. No more wasted truck rolls because the wrong kit was packed or the job wasn’t done right. It’s the move from reactive to predictive maintenance and the savings from that are massive.
SEPTEMBER 2025 Volume 47 No.3
51
Made with FlippingBook - Online magazine maker