Building Technology the Same Way You Build a Job Site

In construction, the sequence of operations is non-negotiable. You cannot pour a foundation before the site is prepared, and you cannot run conduit before the trench is dug. ChaChi Gallo, Vice President of Information Technology at Michels Corporation, applies exactly that logic to the way technology gets built — and he has a name for it: constructing intelligence.

Michels is one of the country’s leading horizontal infrastructure contractors, doing the foundational work that precedes every building: fiber, gas lines, power, road construction, and the underground infrastructure connecting communities to energy and data. The company is currently on some of the largest data center contracts in the country, supporting projects for Google, Oracle, and OpenAI. ChaChi leads IT across an environment that is geographically distributed, operationally demanding, and physically unpredictable in ways that most enterprise technology frameworks were never designed to accommodate.

Why AI Fails Before It Starts

The construction industry is moving fast on AI and automation — often too fast. ChaChi points to a measurable problem: Gartner and Forrester have both flagged a 60 to 70 percent failure rate on AI deployments, and he traces that failure directly to sequencing. Organizations rush to deploy tools before the data foundation is in place. The model gets trained on incomplete, inconsistent, or inappropriate data, and the outputs become unreliable.

He draws the comparison to a construction permit: 

“Just like you have to get a permit to do construction work, you have to go through cyber and risk to make sure that this product is okay for us to use. We have to do the same checks and balances that we do for our customers.” 

The governance structure exists not to slow things down, but to protect the outcome.

His experience at Generac made this concrete. While building an AI-assisted call center knowledge base, the team initially loaded conversational history alongside structured technical documentation. The model began imitating the informal tone of support agents in ways that were actively harmful to the customer experience. The fix was to go back to clean, validated, expert-reviewed data, train the model on business rules, and validate outputs before deployment.

Connectivity as the Foundation Layer

Construction job sites are not like factories. Every site is different. Conditions change. Critical infrastructure is underground, invisible, and subject to surprises. And in many of the areas where Michels operates, cellular coverage does not exist — because Michels is actively building the network that will eventually serve that area.

That constraint changes the infrastructure calculus entirely. Satellite connectivity, and Starlink specifically, became a genuine operational breakthrough. Where crews previously had to drive fifteen miles to their hotel just to upload the day’s data, they can now move information from the site. The workload does not require millisecond latency. It requires reliable daily data transfer — drone footage, equipment logs, progress reporting — and satellite delivers that in environments where no other option exists.

ChaChi is equally clear about what wireless cannot yet guarantee. A visit to an AWS distribution center to evaluate 5G revealed that even at scale, Amazon runs Wi-Fi as its primary facility network, with 5G serving only as a tertiary backup. For teams facing pressure to adopt 5G across job sites, that lesson offers a useful reframe: match the technology to the workload, not the other way around.

Autonomy with Oversight

The future of construction includes autonomous equipment, but ChaChi is deliberate about where that line sits today. On flat, controlled routes, autonomous dump trucks and drilling support can operate reliably. In environments with elevation changes, hidden underground conditions, and unpredictable terrain, the safety case for full autonomy does not yet hold.

The same principle applies to network operations. AI can take on documentation updates, change control tasks, and faster root-cause analysis. What it cannot yet do is replace the experienced engineer who knows which conditions warrant a human decision. As ChaChi puts it: 

“It should think like us, not think for us.”

Building the Foundation Right

ChaChi’s closing message is the one that defines his entire approach. Technology is not the hard part. The hard part is the relationships, the presence, the willingness to leave the office and understand what the work actually looks like. IT leaders who operate only from dashboards lose the operational context that makes technology decisions meaningful.

For construction organizations navigating the pressure to modernize, his framework is clear: build the data foundation before you scale. Design connectivity for real conditions. Keep humans in the loop until the system has earned the right to act on its own. And measure success not by how many tools you have deployed, but by whether the people doing the work trust what the data is telling them.

• Why 60 to 70 percent of AI deployments fail and how to avoid that pattern
• How satellite connectivity eliminated a daily 15-mile data upload drive for Michels crews
• What a visit to an AWS distribution center revealed about 5G’s real-world limitations
• Where automation belongs on construction sites today and where it does not yet
• Why IT leadership depends on relationships and field presence, not dashboards