In September 2024, Microsoft signed a 20-year contract to buy all the electricity produced by the Three Mile Island Unit 1 nuclear reactor to help its data centers keep up with soaring demand for artificial intelligence (AI) tools. This deal is the latest example of how AI affects electrical contractors and design firms even when they don’t use the technology.
Electrical design firms and contractors benefit from AI’s use by other industries simply because all of the necessary computing power — from employee PCs to the data centers for cloud computing and storage — requires enormous amounts of electricity. How much? Three Mile Island's Unit 1 reactor is capable of generating about 837 MW, enough to power more than 800,000 homes. Shortly before the contract was announced, Microsoft and other major AI vendors met with the White House to discuss the need for several new data centers around the country that each would use 5 GW.
The Biden-Harris Administration’s resulting proposal emphasizes renewables, which highlights how AI can drive more business for electrical contractors and design firms that specialize in solar, wind, and battery energy storage systems (BESSs). The proposal also calls for using streamlined permitting to expedite the construction of data centers and grid upgrades. Those upgrades could benefit more than just AI data centers if, for example, the new transmission lines mean more businesses have the grid infrastructure necessary to electrify their fleets or add public EV chargers.
Even if those handful of 5 GW data centers never materialize, hundreds of smaller ones will continue to be built, such as Google’s $1 billion facility underway in Kansas City that will use 400 MW of renewable energy. Those projects will mean plenty of work for electrical contractors and design firms that target the data center market.
“Gaylor Electric is already at the forefront, taking on some of the country’s most complex and robust data center projects,” says Chuck Goodrich, president and CEO of the Indianapolis-based firm. “We believe AI is a key driver in the expanding data center market. Over the next decade, this trend will significantly enhance data processing, storage solutions, and overall efficiency.”
Work smarter, not harder
In addition to designing and building AI-related infrastructure, electrical contractors and design firms are starting to use the technology to increase employee productivity and efficiency. One example is having AI speed up the process of reviewing contracts and assessing risk by doing a first pass and flagging things for humans to scrutinize.
“We currently use ChatGPT but will likely migrate to Document Crunch,” says Gaël Pirlot, vice president at Mableton, Ga.-based Inglett & Stubbs. “We have baseline exhibits with our general contractor partners that have been scrubbed by our executive team and lawyers and discussed with the respective GCs. These documents are used as the baseline for comparison with new project documents.
“For new GC partners, we typically start by scrubbing for financial exposure (fee, liquidated and consequential damages, retention, non-billable items, etc.), staffing requirements (safety and QA/QC minimums), BIM requirements (LOD), and schedule. These AI scrubs are the first pass but not the final pass. AI saves us approximately four hours per initial contract review.”
AI also can streamline the design process for engineers.
“[Over the] long term, AI will revolutionize how design, construction, and prefab are being approached,” says Tony Mann, CEO at Long Island City, N.Y.-based E-J Electric Installation Co. “AI-driven generative tools use algorithms to generate and optimize building component designs based on predefined constraints. This will allow for more efficient, cost-effective, and innovative solutions on projects.
“All of this is in the early stages here at E-J, but long term we envision that AI can handle preliminary design in a fraction of the time. It can do things like run multiple models for optimal routings, which is key for fast-track project delivery, and the projects will see the most cost-effective solutions.”
This efficiency also highlights how AI can help electrical and other trades address the chronic shortage of skilled workers.
At a time when an estimated 40% of the construction workforce will retire by 2030, AI can alleviate the burden on the remaining employees by reducing nonoptimal work, says Associated Builders and Contractors’ latest construction technology report, which focuses on AI.
Waste not, want not
AI also could help address equipment and material shortages.
“We are exploring predictive analytics where AI can help predict the exact quantity of materials required for prefabrication, minimizing waste and keeping a near-exact inventory,” Mann says. “AI will do this through analyzing historical data and forecasting supply chain disruptions to optimize material procurement.”
Ferreting out waste is a use case that vendors are increasingly highlighting in the marketing for their AI and building information modeling (BIM) tools. By some estimates, up to 30% of building materials — and 6% of a building's budget — are wasted due to misorders, errors, and rework. How can AI help? By automating the process of creating multiple design options, which humans then review to pick the one that best meets the project requirements. For example, one design might have a low material cost but a higher labor cost, while another might be the reverse. Having multiple designs to choose from helps identify the Goldilocks one.
“If you don't know what options you have and the consequences of your choices, you're never going to make the right choice,” says Francesco Iorio, co-founder and CEO of Augmenta, whose initial AI agent is designed specifically for electrical.
But to offer all of those options, the AI tool must have access to all of the necessary information, including codes like the NEC. The internal information is somewhat easier because the electrical contractor or design firm already has it, whereas the external data might not be accessible if it’s in a format that the AI tool can’t use. For example, electrical equipment manufacturers will need to put their product catalogs and other information into formats that tools can use. They’ll also need to create application programming interfaces (APIs) so the tools can connect to those databases. And design firms also will need to convert their internal unstructured data into forms that AI can work with.
“We’re in the data governance and accumulation mode on this,” says Inglett & Stubbs’ Pirlot. “We have archived build types (build books) for each project completed over the last five years and formatted the data in a uniform and searchable manner. This data will be merged with defining project attributes and presented through Power BI. We are assessing the right project attributes to review. Our goal is to eventually use AI to predict and align prefab/modular capacity based on market outlook.”
Predicting the future
Another use case involves collecting and analyzing data about an electrical network, such as the one in a building, a factory, or an office campus. This application could use machine learning (ML), which is a subset of AI that analyzes normal behavior so it knows what anomalies look like. Depending on how the system is designed, the AI either alerts humans or takes corrective action on its own, such as shutting down a circuit and rerouting power.
“It’s more prevalent in the [electric] utility side for optimizing the energy use of distributed energy resources (DERs), but we have seen it applied in microgrid applications on commercial facilities, as well,” says Dan Webb, integrated automation technical director at Lenexa, Kan.-based Henderson Engineers. “[An example is] training a model to assess and predict control of the DERs."
“Monitoring out-of-range or out-of-tolerance variables is more in line with what we would refer to as advanced fault detection and diagnostics,” notes Webb. “It’s a fairly common use of ML in predictive analytics in regard to electrical systems, monitoring the various attributes of a system to create a model. It may be voltage, ampacities, or energy consumption, for example. From that model, equipment maintenance (predictive maintenance) schedules and equipment failure could be assessed.”
The AI monitoring these networks could be provided by an electrical contractor or design firm as part of a managed service contract.
“Predictive maintenance leverages AI to monitor and analyze the behavior of electrical systems, identifying problems before they lead to failures,” says Gaylor’s Goodrich. “This proactive approach can enhance system reliability and open new revenue opportunities for contractors by offering predictive maintenance as a service.”
Others agree.
“I think this is a definite possibility,” Pirlot says. “It requires us to set up an early relationship with the end user to install and have access to the proper monitor points. These points are available on smart grids by default but may not be available on the majority of our other builds. A digital twin is the ideal means of doing this, but there are easier means to get there.”
Monitoring and managing networks requires accurate, granular information about each component. This information can be used for additional applications.
“We've been involved in testing and exploring different kinds of algorithms to monitor network performance, predictive analytics, and things of that nature,” says Ed Sutton, enterprise evolution director of AI and innovation at Overland Park, Kansas-based Black & Veatch. “But I think it's even more basic than that: Where are the assets? What condition are they in? Do I have good data? The scale and scope of these networks are growing, and the devices are getting changed out faster. You have environmental context with increasing large-scale weather events, and that compounds stuff. So we're really seeing AI coming in helping make sense of the basics, such as asset management, and our testing is showing many downstream benefits from better supply chains, optimized maintenance, and data-driven capital planning; that really represents the pulse of the organization and its real-time operations.”
This information can inform the design process.
“Interconnects for solar and battery storage and stuff like that, you need grid data, and usually from multipe sources,” Sutton says. “’Do I have to upgrade this feeder because I want to bring X megawatts on? Can I trust the information I am using in my design?’”
Questioning the technology
Generative AI uses “natural language” or “large language” interfaces. These let users type everyday terms to tell the AI tool what to do, such as: “Show all of the options for running conduit on this floor.” There’s no shortage of off-the-shelf generative AI tools, such as ChatGPT, but some electrical firms are developing their own.
“We’ve developed a quality management system (QMS) chatbot that basically brings to life typically 'dry' process and procedures into an interactive experience,” Sutton says. “[It’s to] help our employees have the right information at the right time so they can quickly look up: ‘What's this process? What's this checklist? What's that procedure? Empowering professionals with tremendous amount of knowledge and trusted information in a way that is opening up a lot of potential for even more innovation.’”
Besides being highly customized for a firm’s particular needs, these homegrown AI tools also can improve cybersecurity by ensuring, for example, that company and client data don’t wind up in a public cloud or training a vendor’s AI tool.
“We generally use a RAG process, which uses OpenAI for the service, but the data is all private to our company,” says Brian Melton, Black & Veatch technology innovation lead for governments and communities. “We're trying to leverage the technology, but the privacy and security piece is always in the back of our mind.”
Whether the tools are off the shelf or developed in-house, we need to think about integration with other platforms and technologies, such as building information modeling (BIM), which will unlock new types of capabilities and workflows.
“One of the things we found is AI becomes almost a commodity,” says Greg Tanck, Black & Veatch project manager for operating assets data analytics. “That's not really what makes a piece of software or an organization successful. It's more about everything that goes into supporting that software, [such as] having a good workflow that makes it easy to interact with and to get out of it what you need so it's not: ‘Stop what I'm doing, go do this AI thing, and then come back.’”
That’s one more example of AI’s learning curve.
“AI has been a hot topic, and its impact on the electrical contracting and design industries has been encouraging,” says Gaylor’s Goodrich. “Companies are reporting notable gains in efficiency and productivity. There are certainly instances where AI has streamlined construction processes. But, in my opinion, there needs to be more test results and white papers providing solid documentation. Although we’re still in the early stages, the opportunities for growth and innovation with AI in construction are immense.”
