Inside the Electrical Industry’s AI Learning Curve

One day soon, you’ll catch yourself telling interns and apprentices, “Before we had AI...,” just like Ryan Elbert reminisces about a time when there wasn’t even a computer on his desk — let alone one connected to the internet and the cloud and running AI. He wasn’t replaced by a computer, and it’s unlikely that you’ll be replaced by AI anytime soon. Why? As powerful as it is, AI is still just a tool — one that’s far better at answering questions rather than knowing what to ask.

“We see AI as a powerful tool, but its true value is realized when used by skilled technical professionals,” says Elbert, a Black & Veatch executive vice president and global director of engineering and development services. “While it can be used to automate repetitive tasks and streamline workflows, having a human in the loop is really essential, especially when it comes to construction and design.”

In electrical, AI use cases run the gamut, from producing multiple iterations of a design to creating slide decks to present those design options to a client. A common denominator — and a big part of the business case — is that AI does grunt work that otherwise would tie up highly skilled employees.

“A lot of our integration has been using Copilot to simplify routine tasks and enable our professionals to devote more time to higher-value work,” Elbert says. “I’ve used it to turn emails into PowerPoint presentations. It’s really useful for people like me to increase our skills and utilize Excel more effectively.”

Henderson Engineers is using AI for automating design work in some cases and optimizing it in others. An automation example is AI creating a basic design using criteria for where to place receptacles, luminaires, and Ethernet drops in a room full of workstations. An optimization example is having the AI analyze multiple design iterations to pinpoint the one that best meets the client’s requirements.

“If they said, ‘Optimize this for cost,’ you might get something that was easier to build, faster to build, and reduced labor,” says Dustin Schafer, CTO.

Although ChatGPT made AI an overnight sensation when it was released in November 2022, it’s just one type of AI: generative, which means it uses raw materials such as text and images to create things such as the design for an electrical raceway or a PowerPoint deck. Some electrical firms have been using other types of AI for longer.

“Rosendin has been using AI in one capacity or another for the past six years, starting with more traditional AI use cases, like predictive analytics, and leading up to using generative AI in its current form starting in late 2022,” says Jad Chalhoub, senior director of innovation. “We have also deployed a lot of automation software that I wouldn’t necessarily classify as AI, but that has had massive impacts on repetitive tasks.”

No silver bullet

Black & Veatch, Henderson, and Rosendin are part of a trend in the overall construction and engineering (C&E) sector. According to a 2025 survey of more than 300 senior C&E executives, 91% plan to spend even more on AI in 2026. Another, larger survey conducted by the same company — IFS, which specializes in industrial AI software — found that 58% of those in C&E were creating departments dedicated to implementing AI.

“One thing I would like to point out is the importance of focusing on incremental enhancements,” Chalhoub says. “We are not looking at a silver bullet solution, but rather at utilities that will make [something] 1% better, over and over again.”

Henderson also isn’t looking for a silver bullet.

“It doesn’t do everything, but it does a few things really well,” Schafer says. “There are 2,000 places you could apply those few things and have a big impact. We’re focused on the orchestration of existing workflows and calculations.”

As EC&M explored in a June 2024 article, “Is AI the Future of BIM?”, chatbots and other AI-powered tools can serve as a user interface for multiple software platforms and databases. This also helps minimize hallucinations, where AI basically makes up responses to compensate for incomplete information and prompts.

“We have a lot of design tools and a lot of design standards, and we’re starting to use agents to orchestrate those automations in a way that gives us deterministic outputs,” Schafer says. “It reduces the hallucinations. We’re not counting on the agents to give us an answer. We’re using the agents to sit between the users and the tools to automate streams of calculations so we can do things faster but still get the same good answers we were getting before.”

This is another example of why AI is unlikely to eliminate skilled jobs.

“You really have to have a human in the loop that understands exactly what the design requirements are and how you get to the right answers to validate what AI generates,” Elbert says. “That’s what our licensed engineers do. They have to understand the math and the theory behind the output. The other challenge is the codes and standards. There’s so much interpretation that has to be done. That’s best done by a human.”

Like any other tool, AI’s effectiveness depends on the skill of the people using it.“There aren’t a lot of buttons to learn like traditional software, but knowing how to talk to AI and what to expect out of it is incredibly important to harness its power,” Chalhoub says.

Black & Veatch agrees.

“We’ve learned that effective prompting and experimentation are really the key to unlocking its true potential,” Elbert says. “To become successful in AI, I encourage professionals to share prompt strategies, to collaborate, and even leverage AI to build stronger prompts. You can ask it to help you with prompting.”

DIY professionals

There’s a large selection of off-the-shelf AI tools for general business applications. They’re useful for many electrical use cases, but other applications require AI with industry-specific capabilities.

“A lot of the processes in electrical construction are similar to other office jobs, and leveraging the abundance of tools available for these applications is as important as finding tools for the niche use cases of electrical construction,” Chalhoub says. “We are actively working with a number of providers and startups working on electrical-construction-specific AI tools that we are hoping will address these niche needs.”

Sometimes those niche needs are best met with a tool developed in house.

“There’s no off-the-shelf option that works because all of [our] calculations are proprietary,” says Henderson’s Schafer. “By nature, they’re intellectual property because it is the way you work. Your workflows are baked into your calculations, so you have to build agents to orchestrate those calculations. I’m actually struggling to think of an idea design-wise that is so straightforward I would want to buy a third-party tool to do it.”

Some of these homegrown tools are chatbots, such as Black & Veatch’s BV Ask.

“We’ve been around for over a hundred years, and those hundred years have baked valuable knowledge into our policies, our processes, our procedures, our lessons learned,” Elbert says. “Our professionals can ask a question and [receive] responses with context and citations and point them right to the policy and procedure that applies, to subject matter experts, and things like that. That’s been a really important use case for us.”

Another example is CannonDesign’s Billie chatbot.

“In 2024, we were experimenting with the major off-the-shelf tools available at the time, but we quickly identified a significant gap between their consumer capabilities and our specific professional needs,” says Joel Yow, director of digital products. “The primary shortcoming wasn’t just that they were general business tools, but that they operated as closed systems with limited access to relevant data. Off-the-shelf models could write a convincing paragraph, but they couldn’t reference our past projects, our specific design standards, or the vast institutional knowledge we’ve built over decades. They were brilliant, but at the time, they were blind to who CannonDesign was.”

Developing AI tools in-house also reduces reliance on vendors that could discontinue products or go out of business completely.

“A significant risk with consumer AI products right now is that many are ‘wrappers’: user interfaces built on top of OpenAI, Google, or Anthropic that serve to replicate key functionality but do not fully understand our business processes,” Yow says. “One of our key vendor selection criteria is, ‘Does this product have proprietary data or a unique workflow engine that cannot easily be replicated?’ If we feel the product is fundamentally reselling access to frontier models, especially those models being offered by vendors that are not profitable yet, they are high-risk for the AI bubble pop due to pricing changes or slow enterprise adoption. Getting a company of our size hooked on a product or API that a vendor decides is no longer profitable and shuts down is not an option for us.”

This risk is why CannonDesign gave Billie a nimble architecture that was easily extensible.

“Right now, it connects to many different models, including OpenAI, and we are responsible for the development of our own modularity and features.” Yow says. “If OpenAI were to disappear tomorrow, or if a better, cheaper model emerges from Google, AWS, or Meta, we can ‘swap the engine’ in the background or update functionality with very little friction or negative impact to our users.”

Cybersecurity and job security

Developing chatbots and other AI tools in-house also helps ensure that proprietary company and client data doesn’t inadvertently wind up in a software vendor’s cloud.

“We focus a lot on ‘secure by design,’” says Heath Jeppson, Stanley Consultants senior cybersecurity analyst and author of a forthcoming book, AI Exposed: The Naked Truth. “You have to take every possible avenue to harden the system where you can. We even go beyond the NERC Critical Infrastructure Protection (CIP) requirements by designing these systems so they have that robustness that is going to be required in the future.”

One key to a great design — whether it’s an ultra-secure AI tool or a LEED Platinum building — is experience, which is where relying on AI can actually be detrimental.

“In the cybersecurity field itself, there’s a lot of boring grunt work that goes on behind the scenes that most people have no idea about,” Jeppson says. “It’s kind of a double-edged sword because the industry no longer has junior analysts going through logs and looking for patterns. We’re seeing organizations use AI to do that instead. But that experience is what kind of forges you in the cybersecurity world — spending that time in the trenches.”
One thing is clear: AI isn’t going away, and neither are the people who use it.

“I was talking to some college students, and they’re really worried whether there will be jobs for them when they graduate,” Elbert says. “I go back to when I started at Black & Veatch. Computers showed up, and we’re still here. Like computers many years ago, AI is a powerful tool to assist engineers in efficiently producing quality deliverables.” 
In other cases, new graduates and experienced professionals alike are looking at a firm’s AI adoption when deciding which job offer to take.

“Employees are demanding that we figure this out,” says Henderson’s Schafer. “Employees don’t want to work at a firm that isn’t using AI because they worry that that’s where the jobs will get lost. They feel like if a firm is doing this, that’s where the job will get retained and added. There’s the adage of ‘You don’t lose your job to AI. You lose your job to a person who’s using AI.’ Everyone wants to be the person using AI.”