With labor shortages and project overruns plaguing broadband construction, agentic AI is emerging as a tool for discipline and timely execution
In a recent episode of The Ezra Klein Show, the New York Times columnist said, “I think that period in which we are always talking about the future, I think it’s over now.”
He was speaking in the context of AI. AI models that sound like science fiction in their ability to outperform humans, self-correct and self-improve, and do actual work that is useful and of value — are no longer a forward-looking vision. They’ve arrived, and they’ve a new name: agentic AI.
Agentic AI rings a new era of artificial intelligence in which experts believe the tech will finally catch up to its promises.
Unlike traditional AI, agentic AI does not need prompts to function. Once programmed, it can decide and act on its own. And that has opened a floodgate of opportunities for industries typically slow to adopt technology.
Broadband construction is one of those sectors where AI adoption is gradually picking up, automating bloated manual-heavy processes and shortening operational lifespans.
Motivators of adoption
Stephen Rose, CEO of Render Networks, a major player in the space, says there are two major forces behind this. While it is the natural curiosity around AI which gets companies started down the path, the real catalyst is economic imperatives.
“The problem that you see today in the buildout of networks, 98% of them come in late or over budget or both. So there is a massive economic imperative,” he said.
Alluding to Marc Ganzi’s keynote at Metro Connect USA earlier that day where this interview took place, he echoed that labor shortage remains a major constraint in construction — and that presents another big opening for AI.
“Those labor shortages mean that you don’t have time and resources to redo things that you can get right the first time; so using AI and automation ensures that there is a very simple but effective handover,” he said.
Rose broke down the workings on agentic AI, and how it introduces speed and economy. “The way we think about [agentic AI] is that there will be a system of modular, small models or agents. And those agents will then be able to be stitched. So, you’ll be able to interrogate the data and use the agents to actually think about what types of problems you want to solve for or what kind of analysis that you do. So, it’s far more intuitive and far more intelligent,” he explained.
This modular design he said allows agentic systems to run on relatively low compute. But an even bigger perk is autonomy. “The really interesting thing is you can get the agent to figure out for itself what is the next best action for you. And then you can program it to take that action if you want to — because if you’ve faith in the system and your testing says, every time it gave me an answer that was 98% correct, you’d take the 2% risk if it was something that wasn’t going to be harmful in any way. So that’s a completely different way of being able to use AI.”
AI-driven project execution
Rose, who stepped into the position of CEO a year back, views agentic AI as a game changer for modernizing construction processes both in the field and in the office.
An agentic AI mesh has many capabilities. As noted earlier, its composable, distributed design requires very light-weight compute which is perfect for the edge; similarly, its ability to execute multistep processes can eliminate repetitive administrative tasks, optimize scheduling, enhance on-site safety, and reduce planning and design time by cutting down errors through real-time monitoring.
A Science Direct paper argues that LLM-based agentic AI holds the potential to address long-standing complexities in built environments originating from expansive spatial scale, fragmented architectures, and prolonged operations. This is evidence-based. The paper draws on multiple experiments exploring application in domains like modeling, design verification, data mining, and system operations. In each of these applications, agentic AI’s ability to set goals and achieve predefined objectives through task completion showed clear benefits.
AI’s ability to parse through volumes of data is one of the greatest capabilities of all times. AI agents take that a step further.
“What’s really interesting about the new advancements in AI — and particularly agentic AI — is that…instead of..actually interrogating an entire database at once, you essentially break it down into modules…you atomize it,” Rose said. “And by atomizing it, you actually then are able to interrogate those different discrete areas and then combine using interesting prompts. You can actually interrogate that data in a dynamic way and then get the types of answers that you’re looking for.”
This is especially useful for making sense of fragmented data, such as quality of soil, satellite imagery, build processes, etc. coming in from various sources. Using AI’s pattern recognition and trend analysis capabilities, and the agents’ autonomy and speed, construction firms can move from data to insights much faster. Use cases also include processes like invoicing and auditing that demand speed with no room for error.
Another use case, Rose highlighted, where different forms of AI can be used in concert to deliver better results is reducing manual measurement time and human error in as-built diagrams. AI tools allow workers to record adjustments and changes made to a build in real time, ensuring greater accuracy and fewer reworks.
“When you bring those things together, what you end up finding is that the accuracy of your completion process or your closeout process is far greater,” he said.
Asked if AI could also expedite the long-drawn broadband permitting process, Rose said, “It is more about trying to make sure that the ecosystem plays friendly…that is still an area for us to work on together.”
Rose recommends adopting a cohesive and consistent approach for achieving best results, which involves implementing AI in pre-construction, through the build process, and then at closeout, so as to ensure “that you actually have an understanding of the state of the build at any time.”
But could such deep integration of AI mean fewer jobs? Rose responded that, contrarily, adoption of AI will release humans from the burden of standard supervision work, instead moving them into leadership roles that involve actively thinking about the next problem rather than figuring out where the problem is.
“The machine or the AI does the thinking for you,” he said. “That’s a huge change, and it just means that the work quality goes up and the stress goes down.”
