See also: The last unconnected asset on the plant floor: The case for ‘smart’ PPE

SI: NEXCOM's survey data, which concluded that about 95% of organizations prefer localized or edge-based AI inference over public cloud deployments, suggests that control is a key concern for manufacturers when it comes to AI. Can you explain what “control” means in the context of an industrial operator, and why edge AI offers more of it?

TJ: For industrial operators, control means having certainty that critical systems will respond as predicted. It also means maintaining visibility into operational data, deciding where that data is processed, and making sure AI supports existing production requirements instead of introducing new uncertainties.

Running AI at the edge gives manufacturers direct oversight into how decisions are made and allows those systems to continue operating, even if network connectivity changes. In industrial environments, that level of consistency is often just as valuable as the intelligence itself.

See also: Industrial IoT starts with the physical layer: Rethinking cables, conductors and connectivity

SI: Most of the AI market is currently in favor of cloud-first solutions, with industrial AI being the outlier. Do you think that preference gap will continue to widen or narrow over the next few years?

TJ: I expect the gap will narrow, but I don't think industrial AI will become entirely cloud-first. As manufacturers become more comfortable deploying AI, I see them building architectures that combine the strengths of both environments.

Real-time inference, robotics, and machine control will continue to benefit from edge computing, while the cloud will remain better suited for model development, fleet management and analyzing data across multiple sites.
I think we'll see a more balanced approach over time with organizations placing each workload where it delivers the most value.

See also: Growth in IoT sensor market points toward strong momentum for DX

SI: Where do you see enterprises in the industrial space prioritizing AI spend? Is this yet another area where manufacturing and industrial sectors are diverting from the enterprise AI status quo?

TJ: Manufacturers are generally investing in AI where it can improve daily operations. This includes areas like machine vision, robotics, predictive maintenance, and production optimization, where organizations can measure improvements in quality, uptime, and efficiency.

In comparison to many enterprise AI deployments, industrial companies tend to spend more on strengthening the underlying infrastructure, including edge computing, data collection, and system integration.

This ties back to a common perspective and practice across the manufacturing industry that reliable AI depends on a reliable operational foundation.

We’re seeing this priority particularly clearly among small and midsize manufacturers. Rather than beginning with large, enterprise-wide AI transformation projects, many are looking for practical entry points that can deliver measurable operational value quickly.

With this approach in mind, industrial leaders are allocating AI spend to platforms that can connect existing CNC machines and production data to help manufacturers monitor utilization, identify abnormal conditions and uncover production bottlenecks.

The objective is to build a reliable data foundation first, then introduce AI applications that can support daily decisions and continuous improvement. This allows manufacturers to start with a focused deployment, demonstrate business value and expand over time without replacing their existing operational systems.

See also: What to do when ‘rip-and-replace’ is financially impossible

SI: Are there specific AI workloads that manufacturers believe should remain at the edge, and are there others where the cloud still makes sense? How do you see companies making those distinctions?

TJ: The distinction usually comes down to timing and operational impact. Workloads that involve machine control, robotics, quality inspection, or safety-related decisions are typically better suited for the edge because they require immediate responses and uninterrupted operation.

The cloud is valuable for tasks such as training AI models, analyzing production trends across multiple facilities or managing large amounts of historical data.

Rather than asking whether AI belongs at the edge or in the cloud, manufacturers are increasingly evaluating each workload based on its latency requirements, reliability needs and the consequences of delayed decision-making.

See also: For sweeter AI adoption, Hershey and KDP utilize Augmentir's connected worker platform

A practical example is production monitoring, where AI continuously evaluates equipment status, machine utilization and process conditions directly on the factory floor. Processing this information locally allows operators to identify abnormalities and respond immediately, even if network connectivity is limited.

For workloads that directly influence production, reliability and response time often matter more than centralizing every piece of data.

The cloud still has a valuable role in consolidating data across facilities, retraining models, managing software updates and identifying broader operational trends. Rather than choosing one environment over the other, manufacturers are increasingly matching each workload to the environment where it delivers the greatest operational value.