Computer vision is an artificial intelligence (AI)-powered technology that enables machines to “see” and analyze objects in the real world via cameras and sophisticated algorithms. Through machine learning, computer vision systems learn to recognize features and patterns in visual information, ultimately understanding the contents and context of an image the same way a human can.
However, unlike humans, AI computer vision solutions can analyze thousands of images in minutes and detect relevant information, like the presence of specific cables in a trench or the location of a newly-installed smart electricity meter. AI computer vision capabilities include optical character recognition (OCR), where the computer can read words and numbers that are present in an image and translate them into accessible data points.
Jay Cadman, Chief Revenue Officer, IQGeo
Telecom and utilities operators manage increasingly complex networks and decentralized assets. They’re tasked with modernizing expanding asset inventories with new technology while facing challenges around skilled labor and rising OPEX. They need operations software that enables them to address the demands of deploying, upgrading, and maintaining massive, next-generation networks.
AI computer vision embedded into an existing geographic information system (GIS) or geospatial network management software like IQGeo enables operators to accurately capture detailed information about their networks that can be used to build a robust digital twin and gain insight into the quality and performance of their assets. It also enables quality control (QC) automation to ensure that field workers complete installations and maintenance work right the first time. Computer vision improves operations and reduces costs across the network lifecycle from construction to asset management and ongoing preventative maintenance.
In this section, we explore how telecom and utility operators can leverage AI-powered computer vision, including the following use cases:
With AI computer vision integrated into field workers’ digital documentation workflows, telecom and utilities operators can control the quality of their network infrastructure and assets from the start and build an accurate picture of their network as it expands.
Rather than relying on time-consuming manual reporting and in-person inspections (that call for expensive truck rolls) operators can leverage computer vision to automatically check 100% of their network construction activities in real time. As trenches are dug and assets are deployed in manholes, on poles, or in cabinets, AI computer vision can assess the quality of equipment installation and engineering work using photos taken by workers on the ground and provide feedback in seconds.
Field workers can take actions immediately to correct errors and ensure their work meets the operator’s standards. This also allows operators to enforce quality standards within growing field teams and even across their contractors’ organizations. Operators get visibility into whether contractors have done their jobs correctly, expedite validation, and streamline the payment process to contractors.
Leveraging AI for quality control automation does not mean that there are no longer humans in the loop. Rather than eliminating roles such as quality control managers, AI supplements their work. In this white paper, IQGeo explains how, and addresses some of the key common misbeliefs around the use of Visual AI in field operations.
AI computer vision enables network construction teams and contractors to deliver accurate, complete, and verifiable as-built documentation. Accurate as-built documentation is especially critical for compliance and computer vision makes it easy to create granular documentation of the network at the outset.
With as-built information captured and validated by AI solutions, operators can also get deeper visibility into the locations of assets and their surroundings. They can see whether the network build deviates from the network design to validate completeness and ensure that their digital records of the network match reality. This data is critical for tracking performance and maintaining the network as time goes on. It also creates the foundation for a highly accurate digital twin of the entire network
Listen to the podcast to discover how AI-powered computer vision is revolutionizing fiber construction - automating inspections, reducing rework, speeding approvals, and guiding crews in real time. A must-listen for telecom leaders driving smarter, faster network builds.
If they want to start winning subscriptions and driving revenue, fiber operators have to get last-mile connection right the first time to deliver on their promise of high-quality connectivity for customers. With AI computer vision, operators can offer exceptional customer experience while managing an exponentially-expanding asset inventory.
Last-mile fiber-to-the-premises (FTTP) installation is complex. Every single job site is different and technicians need to understand how to apply quality standards under a variety of circumstances. Techs are pressed for time: customers expect fast service that doesn’t interrupt their day. Slow service or a poor connection can lead to complaints, refunds, and churn, costing operators time and recurring revenue. They also face pressure to work fast from the top down: operators need technicians to complete installations as fast as possible to speed time-to-revenue. AI computer vision enables techs to check their work as they go using the software they already use to document their work and receive instant feedback. They don’t have to spend time on additional reporting during or after each installation, which means they can be on time to their next appointment.
QC teams also get visibility into every installation. Instead of manually checking reports, the back office is automatically notified about anomalies and issues that might require additional intervention. The team can take proactive steps to improve the quality of installations as deployment unfolds.
With a photo, technicians can quickly and correctly document the equipment that’s present at each installation location. OCR eliminates the need for techs to manually enter tedious details like optimal network terminal (ONT) models and serial numbers. They can also document where they run fiber cables to the premises, whether that’s via underground ducts or poles. This information gives 360° visibility into the full context of the installation, making future tech visits smoother and providing operators with a detailed view of every asset deployed.
Globally, utilities companies are under increasing pressure to make the switch from analogue to smart meters across their existing grid while expanding their services to more areas. At the end of 2023, North America led the global switch to smart electricity meters with a 77% market penetration rate and is projected to install 182.9 million meters by 2029. Meanwhile, Europe is projected to install 335 million meters by the end of the decade.
Operators that want to lead the smart meter transition can use computer vision to ensure high quality installations and collect accurate documentation as they go without compromising on velocity.
Smart meter installation is an essential part of the global transition to green energy. As smart meters quickly become the standard, operators need to ensure compliance through rapid rollouts. Like last-mile fiber network deployment, utility meter installation takes place in a wide variety of locations under variable circumstances. Technicians need to deal with installing meters that might be located in different locations in each customer’s home or business. They might have to troubleshoot issues with unexpected configurations and outdated or damaged components. They also have to ensure that the meter is installed safely, protecting themselves and the customer.
AI computer vision automates the QC process, checking that the installation meets both internal quality standards and compliance regulations. AI solutions provide instant feedback, so technicians can correct issues quickly, ensuring a successful, safe installation the first time. This speeds up deployment and makes it easy to maintain compliance without compromising on velocity.
Beyond smart meters, the utility grid is composed of many assets that require inspection throughout their lifetime, not just at the point of installation. By embedding AI-powered image recognition directly into field workflows, organizations can automate the field verification of infrastructure, detect anomalies in real-time, and maintain a trusted record of network integrity – all from a mobile device. To see how AI photo capture is revolutionizing utility asset inspections, check out our detailed blog article on the topic.
OCR enables utility operators to get a detailed view of their meter inventory. Field workers can seamlessly document meters during installation and inspections, capturing serial numbers, model information, and location in a single photo. AI computer vision eliminates the chance of error that comes with manually entering long serial numbers. Having this data enables operators to better manage every meter through its entire life cycle. Should meters need to be repaired or replaced in the future, technicians will have all the information they need, plus an image of the meter in its original condition, at their fingertips.
Maintenance and repairs are a major cost center for telecom and utilities companies. AI computer vision solutions provide proactive insight into the condition of the network so that operators can perform preventative maintenance. This reduces the cost of labor and materials, decreases truck rolls, increases efficiency, and lowers the risk of expensive downtime.
Jobs done right the first time are less likely to lead to issues down the road. AI computer vision can verify whether assets are at risk of damage or degradation every time a technician logs a photo during routine checks. Data captured and analyzed by the AI gives infrastructure managers a view into each asset’s history and automatically flags updates or fixes that need to be made during a maintenance visit. If an asset is shared between multiple operators and maintained by different subcontractors, wholesale carriers can follow its state over time and track which parties are responsible for repairs if damage occurs.
When techs return to a job site to perform maintenance, they can proceed through the same AI-powered photo capture workflow that validates their work in real time and sends insights to the back office. QC teams can ensure that maintenance jobs are complete and compliant, reducing the chances of failure and eliminating the need for a revisit.
As field workers document the assets and infrastructure they work on, AI computer vision can actively analyze images for signs of damage and predict potential failure before it happens. Teams are automatically notified when the AI detects issues, so operators don’t have to spend on additional truck rolls for inspections. Rather than reacting to issues when they cause problems for customers, operators can strategically schedule maintenance activities and eliminate downtime. This not only increases operational efficiency and cuts costs, but it also improves customer experience and elevates the operator’s reputation.
Computer vision solutions can be customized and adapted over time to address the evolving needs of your operations. Here’s what the process of building and deploying AI-powered computer vision looks like:
Phase 1: Data Collection
The first step in creating a robust dataset is to determine what you want your field workers to photograph and which features or objects the AI should check for. You’ll need to start collecting images that capture the circumstances your teams are likely to encounter. This includes any assets you want to track and assess, like trenches, fiber chambers, transformers, transmission or gas lines, and meters. You’ll need photos that show equipment in different lighting and in different working conditions. Most importantly, you’ll need photos that show both the correct and incorrect state or configuration against which the AI can check all future images.
The best way to collect images to use as training data is to get technicians out in the field to start taking photos. If they’re using a field service management (FSM) or GNMS solution, this requires a new workflow that prompts them to take photos of specific equipment at certain checkpoints. Depending on the visual complexity of the circumstances you need the AI to be able to analyze, such as detecting single cables in a bundle, you will need anywhere from 100 to thousands of photos to achieve reliable responses.
Phase 2: Machine learning model training
The next step is annotating the image data coming in from the field so that the AI machine learning models can train on it. Training refers to teaching the AI how to spot relevant patterns in the data. In the context of AI computer vision, annotation is the process of describing or pinpointing to the AI the information it should look for in the photos.
Once the data is annotated, it’s fed to the machine learning models. Through this process, the AI learns to identify what elements should be present in the photos techs take in the field and what output to provide when the image matches its expectations, and when it doesn’t. At this point, the AI will learn how to assess technicians’ work, from reading meters and component labels to identifying installation errors. Training is an ongoing process. As new assets and standards are implemented in the organization, the AI models need to be continuously updated to reflect what’s happening in the real world.
Phase 3: Scale deployment in the field
Deployment starts with fully integrating photo capture and AI computer vision checks into field workers’ existing digital workflows. Ideally, technicians are guided through a set of checkpoints that tell them when to take photos and what those photos should contain. The AI can then validate the images in the moment and provide real-time feedback so technicians can get the job done right the first time.
In the back office, quality control teams can use the incoming data to assess whether there are gaps in the information coming in. At this point, you can determine whether you have enough, or too many, checkpoints. The data can also help teams understand where common issues tend to arise so they can provide coaching or make operational changes to get ahead of potential problems. With deeper insight into the quality and completeness of work, teams can quantify the quality of their networks and grids and effectively measure it against efficiency and cost.
Rolling out your AI computer vision solution in the field should be iterative. As more photos are captured and analyzed, the solution will become increasingly adapted to your operations. Along the way, you’ll update the models based on changes in your business to ensure that your evolving quality standards can be carried out at scale, eliminating the need to train and re-train team members as you grow.
This guide provides in-depth insight about the step-by-step approach to computer vision implementation in your field operations. Discover the recipe for a successful visual AI project that will generate business value in less than 3 months.
Operators that deploy AI computer vision solutions today can start driving operational and business outcomes very quickly. They can create detailed documentation of their network faster and more accurately than ever before, eliminate human error from the data capture process, and speed up reporting for their techs. Technicians get coaching in real time and operators can enforce quality standards without spending time and money on additional training. In turn, deployment happens faster and fewer installation and maintenance errors reduce costly revisits. Operators can cut time-to-revenue during deployment and keep costs down throughout the network lifecycle.
Interested in implementing visual AI, but not sure about the ROI for your organization? Take one minute to input your operational data. Get the result instantly!
AI-driven computer vision also lays the foundation for an accurate digital twin of networks and grids. With high-fidelity data, operators can build advanced automation to increase efficiency and reduce labor costs moving forward. Creating a digital twin prepares networks for the future by making it easier for operators to understand performance as the demands of the market evolve in an increasingly connected world.
Raf Meersman, Chief Customer Officer , IQGeo
Incorporating AI computer vision into reporting workflows enables contractors to reduce operational costs, get paid faster, and win more business. Contractors shoulder the high cost of technician labor, administration, and equipment that revisits require. Errors slow down operational efficiency and hurt contractors’ relationships with operators. AI computer vision ensures field workers get the job done right the first time so contractors can work quickly and deliver exceptional services.
It also improves field workers’ experience: with AI computer vision, they don’t have to take extra time out of their day for reporting and can focus on the skilled work they’re trained for. They also get real-time feedback so they can improve the quality of their work and deliver results.
AI computer vision can automatically validate the work contractor teams perform so operators can pay them as soon as jobs are verified as complete and correct. For contractors, getting paid faster means better cash flow and more capital to invest back into the business. This validation process also allows contractors to showcase tangible value. Operators can measure which contractors are doing the best work and reward them with more projects.
The success of a visual AI project hinges on adoption by contractors and their technicians. In this whitepaper featuring real-world insights from telecom and utility leaders, learn
At IQGeo, we believe that your geospatial network management software should provide visibility into the full context of the network through a single pane of glass, bringing information out of data silos to give operators real-time insights into network performance. To deliver unmatched geospatial asset intelligence for operators, IQGeo acquired our existing technology partner Deepomatic, a global leader in AI computer vision software for the telecom and utilities sector.
IQGeo’s AI-powered computer vision software is already deployed globally and has processed more than 20 million jobs in 2024, including over half a billion transactions from more than 30,000 daily field users.
See how operators like Swisscom, CityFibre, and Unit-T use IQGeo’s computer vision software to get a detailed view of their assets and networks and strengthen operations.
Quality controls in network construction using AI are a central lever for Swisscom. AI computer vision not only improves construction quality, but also enhances the customer experience thanks to a higher rate of correctly-completed work and fewer interventions on site.
Ralf Gugelmann, Head of Rollout, Maintenance & Single Project , Swisscom
Our priority is the quality of customer experience. With AI computer vision, we can achieve this, even when performing thousands of work orders every day. It enables our personnel in the field to focus on adding higher value.
Carmine Muscariello, VP Executive Director Wholesale & Operators Relations, Bouygues Telecom
By obtaining feedback on the quality of their work, teams gradually improve their techniques. Moreover, the reduced error rate eliminates the need to plan additional visits. This reduces the risk of delays, contributing to a smooth and effective fiber deployment.
Philippe Jaspart, Quality Manager, Unifiber
Discover more stories of customers using real-time visual AI.
