The Concept of Physical AI Is Growing for the Supply Chain, but What Is It?

Over 2H 2025 and the beginning of 2026, the concept of Physical Artificial Intelligence (AI) has started to be used more in relation to a variety of Internet of Things (IoT) markets. The supply chain is one key market where the concept will start to have more of an impact, and it is worth exploring how this represents an evolution of technology availability and adopter needs.

Broadly speaking, Physical AI refers to devices and infrastructure sensing their physical surroundings, and using AI agents distributed at different physical infrastructure levels to communicate with each other, to autonomously send and execute workflow instructions. It is, in effect, similar to the concept of “automation,” whereby sensing functions trigger rules that result in a real-world physical action being taken without the need for human intervention. Where Physical AI differs from traditional “automation” is the use of AI models and distributed AI agents. Where “automation” enables fairly simple workflows to be triggered based on a limited pre-determined set of rules, Physical AI aims to take this to the next level by applying AI capabilities to highly variable scenarios. Supply chains are an important field where AI can generate value because of their inherent characteristics of scale and variability.

The concept of Physical AI also changes the paradigm regarding where value is generated from supply chain technologies. One key instance of this is the field of supply chain visibility, a market that has traditionally been dominated by suppliers offering cloud-based Software-as-a-Service (SaaS) platforms of varying descriptions, usually providing dashboarding and data analysis for tracking and sensor data, coming from various IoT tracker devices, telematics, or other supply chain sources. Physical AI changes this paradigm by pushing the focus from cloud SaaS platforms to edge execution, with intelligence distributed at various levels of edge and cloud to orchestrate physical actions.

This is significant for two reasons. First, SaaS platforms are increasingly considered less defensible because of the increasingly low barriers to entry to creating dashboards and platforms. Part of this is the result of “vibe coding,” though significant challenges, often around security, still remain before this approach becomes mainstream. Another and bigger part relates to the fact that supply chain visibility solution providers have always struggled to make money off their offerings as data aggregators and dashboard builders, finding it difficult to differentiate one platform from a competitor’s. Increasingly, a trend toward adopters unbundling hardware purchases from software purchases highlights the challenge: while adopters are happy to buy hardware rather than build their own in-house, cloud platforms are a different matter, so adopters have been looking to reduce margin stacking by sourcing hardware directly and building their own platform in-house.

Second, the shift to Physical AI is significant because of what this implies for hardware. Supply chain tracking device manufacturing is increasingly shifting to Chinese Original Equipment Manufacturers (OEMs) because of their very low cost. As a result, Western OEMs are looking for ways to differentiate their products. For some, this means developing their platform capabilities further—with the challenges noted above. For others, this means the sale of an intelligent hardware system, with a focus on edge intelligence at both endpoint and aggregation levels. Key examples of this are OnAsset Intelligence, sSystem Loco, and Trackonomy for which hardware is a differentiator because of its embedded intelligence, thus avoiding the usual conversation around commoditization.

Hardware very much remains a critical part of supply chain visibility, with ownership of the data collection layer and the ability to build intelligence based on proprietary data acting as a critical differentiator for adopters, and a key foundation for Physical AI. Software platforms that purely aggregate data from third-party devices or systems tend to suffer by contrast, because of the lower defensibility of SaaS platforms.

A key enabler of intelligence at the local layer is a flexible Operating System (OS) that can run on both devices and infrastructure, orchestrate data, and configure profiles based on device and read-point roles. This is one of the key areas in which smart labels based on Bluetooth® Low Energy (LE), cellular, or other active technologies are able to stand out compared to Ultra-High Frequency (UHF) Radio Frequency Identification (RFID). While UHF RFID labels can send an ID to readers at various choke points, the ability to run an OS on a higher memory device and modify device profiles based on their evolving contextualized position within supply chains is a key standout feature for enabling low latency, contextualized, and AI-driven decision-making—allowing the concept of Physical AI to become a reality. Smart labels will be a key enabler of the Physical AI concept for this reason. 

As the visibility ecosystem matures and point solutions start giving way to more integrated offerings, many companies from across the supply chain visibility ecosystem are converging on the concept of physical execution instead of a platform-first approach. Enterprise customers are increasingly requesting data to be fed into data lakes, to enable them to be used across a wider variety of workflows, and to orchestrate different actions in the cloud and edge. Separate platforms tend to lead to buyer resistance, and value will increasingly be generated by headless architectures that manage the complexity of data analysis and sensor fusion, both at local and cloud levels, without focusing on frontends and dashboards.

Physical AI, in this sense, relates to a change in how adopters will consume data: less through pre-made dashboards, and more through local machine decisions orchestrated by intelligent gateways and infrastructure, and supervised and controlled by AI-native systems in the cloud. As vendors converge on execution instead of platforms, there will be consolidation and new product launches across the supply chain visibility market, spanning UHF RFID, IoT solutions, visibility control towers and aggregation software, and, ultimately, the fleet tracking market. While SaaS models are not disappearing entirely, suppliers should be aware of the changing and accelerating competitive landscape, and look to position themselves accordingly.