Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

How to solve the utility industry’s black box problem

Technology now enables us to capture the box makeup and information, sidestepping manual efforts.

[Photo: Getty Images]

Profile photo of Todd Slind

Vice President of Technology · TRC Companies

There are more than 55,000 electrical substations in the U.S. power grid. Every one of them is critical infrastructure. And to the digital systems utilities rely on to manage power delivery, nearly every one is a black box.

The digital information about these substations is limited, and often nonexistent. Utilities are steadily capturing more of it, usually in the course of engineering work or design upgrades. But for the most part, what sits inside a substation—the specific equipment, how it’s performing, its maintenance history, the issues just starting to emerge—never makes it into the digital record.

This isn’t an oversight, it’s a massive logistical challenge. Until recently, the only way to see inside one of these sites was to put a person in a truck and send them out to look. Across 55,000-plus locations, that’s a decade-long initiative. And utilities already have their hands full keeping the lights on, managing daily operations, and responding to urgent events like major storms. Asking them to also digitally inventory every substation from scratch was never realistic.

Utilities don’t have a monopoly on this problem, either. Every industry that runs large-scale infrastructure faces some version of it: critical assets that have never been fully inventoried or digitally connected, visible only when someone physically shows up on-site. The cost of that gap is real. It touches reliability, safety, and operating expense, and it leaves valuable data stranded, out of reach of the IT and AI initiatives that increasingly depend on it.

SOLUTIONS ARE EMERGING

So why has the black box problem gone unsolved for so long? Because the traditional approach simply doesn’t scale. Done the old way, closing the gap competes directly with the work of actually running the grid—and the work of running the grid wins, every time.

Two technologies are changing that math: 3D computer vision and purpose-built AI models designed to understand complex infrastructure.

3D computer vision captures ultra-high-resolution imagery through portable cameras and sensors, producing a highly detailed digital scan of a physical space. Because the equipment is small and portable, crews can carry it with them as part of work they’re already doing—capturing a substation while they’re on-site for a work order, not on a separate trip. They can also fold in information that has only ever lived on paper or in someone’s head: service history, field notes, the observations that never made it into a system before. In a single visit, a black box becomes a precise digital model.

Then the second technology takes over. Purpose-built AI models—what many people now call vertical AI—analyze the scan and map what’s inside it: equipment, pipes, conduits, and the relationships between them. What once took days or weeks of manual work happens in moments. These models can identify remarkable detail, down to the specific model of a pipe elbow or conduit, and flag conditions like corrosion, leaks, or maintenance that’s coming due.

This is work generic large language models can’t do. General-purpose LLMs are built and trained in pursuit of artificial general intelligence—a goal that often has little to do with the specific demands of running a business. Vertical AI is the opposite: trained on industry- and company-specific data, and engineered to solve a defined problem, like turning a 3D scan into a live model of infrastructure that used to be invisible.

A PATH FORWARD

Together, these two technologies turn an organization’s existing workforce into its most effective data-capture engine. They close black boxes systematically, as a natural extension of work that crews are already doing, rather than through a separate program that takes 10 years to finish.

That’s a practical path forward, not just for the 55,000 substations the U.S. grid depends on, but for every industry that needs to finally see inside its most critical infrastructure.

Todd Slind is VP of technology at TRC.