In an interview with CRN, IDC analyst Carlos Gonzalez talks about how AI development is driving demand for IoT technologies and how IoT vendors are taking advantage of generative AI and agentic AI capabilities to introduce new features into their products.
While much of the AI conversation has been focused on massive data centers being propped up to enable cutting-edge capabilities, the demand for such offerings is also propping up demand for IoT technologies.
This is according to Carlos Gonzalez, research manager of industrial IoT and intelligence strategies at research firm IDC, who said this comes down to the simple fact that AI applications need data to unleash new levels of automation and analysis.
[Related: 10 Cool IoT Products And Services That Debuted In 2025]
“It can’t do any of that without the data that was being collected by Industry 4.0 initiatives,” he said in an interview with CRN last Friday. “So IoT is how we collect information. That is the backbone of anything that we want to do today.”
This taps into “physical AI,” a term popularized by Nvidia in recent years to describe what Gonzalez called the infusion of AI into traditional operational technology environments, whether that’s a manufacturing plant or a power grid. The aim is to “collect data from those areas to help facilitate and automate processes that would be done physically,” he said.
While Nvidia has mainly focused on robotics for physical AI applications, Gonzalez sees the field encompassing a broader set of use cases.
For example, he cited a startup he recently spoke to that is working with a utility company to speed up the process of inspecting power lines. This kind of work typically requires an inspector to drive hundreds of miles and look at power lines along the way to ensure there is not any rot within the wooden poles or any downed wires.
“With physical AI and using sensors and connectivity, you can install hardware to collect data from those areas, [monitoring] humidity in the ground to make sure this has not reached the level of humidity that we need to be concerned about,” he said.
The solution could also involve systems that can perform visual inspections, using AI to determine whether maintenance is needed, the analyst added.
“It’’ a combination of technologies that have really been around for a while, but now, because we have this new compute layer with artificial intelligence, we can start doing more of what we already have,” Gonzalez said.
How IoT Vendors Are Using GenAI And Agentic AI
While AI development is helping drive demand for IoT solutions, vendors are also introducing generative AI and agentic AI capabilities into these offerings to improve the way information is accessed and automate various processes.
For generative AI, this is showing up in IoT as a “natural language integration between the operator and the data,” letting users more quickly get insight, query dashboards and fetch information such as maintenance procedures, according to Gonzalez.
“Instead of having to have an operator go search and hunt down for a particular maintenance manual, well, they can just use a GenAI query within a platform and find what’s needed,” he said.
Agentic AI, on the other hand, represents “what a lot of IoT platforms are driving right now and really pushing to their customers,” said Gonzalez.
For instance, he pointed to industrial IoT vendors like Honeywell, Siemens and Aveva, which are all using AI agents to connect data points and run other types of processes autonomously.
“It’s about facilitating autonomous operations, providing recommended actions and analyzing the data on its own,” he said.
While Gonzalez said vendors are “seeing success” with such features, adoption has been very limited so far because customers need to have a “good data strategy” in order to take advantage of agentic AI for “this sort of autonomous level of operations.”
“They want to push this offering. Whether or not end users are ready for it is the question,” he said.
But even if agentic AI features gain more traction in the IoT market, Gonzalez does not expect such things to ever enable 100 percent autonomy.
“No agentic AI is completely operating on its own, and I don’t believe that, in my optimistic view of the human race, we will ever get to that level because at the end of the day, you still want to trust the human operator to make the final decision,” he said.
