Mid-2026 industrial AI pulse check: Is this the year of agentic AI?

In short

• IoT Analytics assesses that many enterprises are entering the agentic & autonomous operations phase of the IoT Value–Maturity Curve in 2026.
• AI was the #1 topic of CEOs as we entered 2026, with discussions of agentic AI on the rise, and Cisco’s CEO called 2026 a turning point for AI and the year of agentic applications.
• Below, we note AI developments based on IoT Analytics research this year, including: transition from advisory output to active orchestration, commercialization of agentic AI, development of domain-specific models, the rise of physical AI, and emergence of prescriptive digital environments.

Is 2026 the year of agentic AI?

Going into 2026, AI became the #1 topic for CEOs, based on our quarterly analysis of corporate earnings calls. Riding on AI’s rise in interest, discussions around agentic AI, AI agents, and physical AI continued to climb.

Likewise, during Cisco AI Summit 2026, Cisco’s CEO, Chuck Robbins, shared his positive outlook for AI in 2026:

“We all believe 2026 is going to be a turning point for AI. We believe that this will be the year of agentic applications, and we believe that all of us know that the impact on what we do every day is going to change significantly.”

Chuck Robbins, CEO, Cisco, February 2026 (source)

Being over halfway through 2026, we ask, “Is this year actually the year of agentic AI?” Buzzwords and hype around AI have been high since the public release of OpenAI’s ChatGPT in November 2022, and the wait for real returns on (record-breaking) investments has lingered on.

In our view, 2026 is shaping up to be the start of the agentic and autonomous operations era. In our State of Enterprise IoT 2026 – Connected Operations in the AI Era report (published January 2026), IoT Analytics CEO Knud Lasse Lueth noted that many enterprises have grown beyond early tech maturity phases like IoT connectivity, platforms, and scaling and were entering the “upper end of the IoT Value–Maturity Curve, as AI and stronger ecosystems unlock the next stage of connected operations,” with agentic AI as the next big leve starting in 2026, as shown below.

Below, we draw on our research from this year and note several developments leading into and within 2026 that reflect our view above.

Agentic workflows: AI moves from advisory outputs to active orchestration

At Hannover Messe 2026, IoT Analytics had 12 analysts on the ground to identify the latest industrial technology trends and gain insights from leading vendors in the field. Our findings were published in our Hannover Messe 2026 event report and a public article titled Top 12 Industrial Technology Trends—As Seen at Hannover Messe 2026. During the fair’s 5 days, the team identified numerous agentic AI offerings from leading industrial automation vendors like Germany-based Siemens, France-based Schneider Electric, Switzerland-based ABB, and US-based Rockwell Automation.

Industrial agentic AI matured sharply over 2025. The team’s takeaway in 2026 stands in stark contrast to Hannover Messe 2025. In our 2025 article listing the top 10 trends noted from the Hannover fair that year, the team specifically noted that while agentic AI was emerging as a theme, demonstrations were largely rooted in simple automation. In 2026, the team noted 29 industrial agentic AI solutions that were much more mature. What’s more, 78% were already model-agnostic, and maintenance & troubleshooting stood out as the leading agentic AI use case at the fair.

For example, US-based industrial technology company Tulip, in partnership with Norway-based industrial software company Cognite and US-based hyperscaler AWS, showcased multi-agent troubleshooting: a Cognite Atlas AI agent (backed by Cognite Data Fusion’s industrial knowledge graph) passes rich contextual data to a Tulip frontline operations agent. The Cognite agent identifies the issue using industrial context, while the Tulip agent translates it into a real-time, actionable workflow for the frontline worker. The IoT Analytics team ranked this as agentic AI level 3 due to its multi-agent coordination capability.

Agentic AI crossed into autonomous execution. The agentic AI solutions at Hannover (as well as Maintenance Dortmund 2026) also represented a functional leap in what AI can do on the plant floor: from advisory outputs to execution. Agentic workflows are now executing multi-step industrial tasks with minimal human intervention, bridging the gap between digital reasoning and operational action.

Based on the Smart Maintenance Trends Report 2026 (published May 2026), AI applications are moving up the operational stack, transitioning from passive failure alerts to agentic AI modules that autonomously support safety and resource-level decision-making. For example, at Maintenance Dortmund 2026, IFS Ultimo showcased an agentic module that independently identifies safety-related content within a technician’s routine report and automatically generates a separate, compliant safety incident record without human intervention.

“One case we have available in the software is based on safety incidents. When a technician files a report, the [AI agent] identifies if there is a potential safety incident as part of that registration. It then raises that incident automatically instead of needing to rely on the technician to do so. Another feature being released shortly is for the maintenance manager, who will receive a daily maintenance briefing directly on his phone.”

Representative from IFS Ultimo

While these agents are emerging to handle unambiguous tasks like ticket qualification and incident reporting, the Smart Maintenance Trends Report notes their current scope remains strictly constrained to ensure reliability in sensitive industrial environments, showing there is still room for advancement with agentic AI in this space.

Commercialization: Agentic AI transitions from PoC to on-the-market

Agentic AI hits the shelves as the pilot era ends. Of the 29 industrial agentic AI solutions observed at Hannover Messe, mentioned above, 72% were already commercially available, not just proof-of-concept or pilot solutions. This commercial availability reflects another trend noted by the team at Hannover Messe 2026: vendors are now commercializing their agents. Over the last 2 years, industrial software providers have added dozens of copilots/assistants to their solutions, often as free pilots. However, vendors are now concluding their initial free pilot phases for these solutions and are actively implementing commercial monetization structures for not only their generative AI but, now, also agentic AI capabilities. For example, citing customer demand for predictable costs, Siemens changed to a €2,100/user/year fixed-subscription model for its Eigen Engineering Agent.

Agentic AI impacting pricing models. One of several themes from the Q2 2026 earnings calls regarding agentic AI is that, for several companies, AI agents are upending software pricing. For example, US-based software company and hyperscaler Microsoft is shifting business apps from per-seat to “seats plus consumption”:

“We are seeing a new pattern emerge as customers shift from traditional seat model to seats plus consumption. The customer service category is at the forefront of this transformation as nearly 60% of our service customers are already purchasing usage-based credits.”

Satya Nadella, Chairman & CEO, Microsoft, April 2026

Meanwhile, US-based software company and hyperscaler Oracle is letting customers buy “bundles of tokens” and pay by outcome:

“We are simplifying how customers consume and pay for agentic capabilities […] customers can also purchase additional agentic capacity […] by purchasing bundles of tokens […]. We’re also introducing outcome-based commercial models that align pricing directly to the value derived.”

Mike Sicilia, CEO & Director, Oracle, Jun 2026

Proprietary industrial foundation models: Vendors bring domain expertise to industrial AI

Industrial AI vendors building proprietary models. Adopters piloting frontier models in industrial settings are hitting a wall: outputs lack the determinism, physics-based reasoning, and domain grounding that industrial use cases require. In response, industrial software leaders are training proprietary industrial foundation models.

US-based enterprise vertical AI company SymphonyAI’s Iris Foundry platform provides a model-agnostic architecture that enables customers to choose the verticalized AI models best suited to their industrial workflows (co-built with Microsoft). In April 2026, the company expanded Iris Foundry to include 8 industrial applications that it says were specifically built for energy operators using industry-specific failure nodes, process dynamics, and regulatory obligations. The model-agnostic nature of Iris Foundry enables the creation of multi-agent systems, where different agents can utilize different underlying models for tasks like predictive maintenance, anomaly detection, etc.

Siemens has also been investing heavily in this space, beginning with the 2025 release of its Industrial Foundation Model (IFM) made in partnership with Microsoft. While Siemens did not provide any major updates to IFM at Hannover Messe or other similar fairs this year that we have seen, it publicly committed to invest over €1B into industrial AI over the next few years in late 2025.

Physical AI: AI learns to act autonomously in the real world

Physical AI bringing agentic reasoning into real-world machine control. Physical AI is worth watching in 2026, as it is as big a buzzword in industry as agentic AI. While agentic AI orchestrates decisions and actions, physical AI pushes that same agentic reasoning into the physical world, embodied in robots and edge controllers that sense, decide, and act directly on their surroundings.

At Hannover Messe 2026, Germany-based industrial automation company Beckhoff Automation demonstrated its TwinCAT CoAgent as a physical AI solution that uses LLMs connected via the Model Context Protocol (MCP) to directly control real machine motion sequences within the TwinCAT automation platform. Engineers type a command, then CoAgent translates it into machine commands, orchestrates path planning, generates function blocks, and performs error diagnostics.

With solutions like these, physical AI can reduce skill gap issues by allowing operators to set goals (e.g., “clear this field” or “optimize this harvest”) rather than programming paths. The machine takes on the cognitive load of navigating rugged, unpredictable environments, enabling less-experienced workers to operate with veteran-level proficiency.

Agentic AI in digital twins: Prescriptive digital environments

Digital twins becoming engines for real-time, autonomous industrial control. Companies continue to advance their digital twins of physical assets. The new paradigm in an AI-led world: digital twins that form a part of an active, real-time computational engine. By fusing physics-based simulation with AI, industrial vendors are creating executable environments that can not only predict operational outcomes but even autonomously execute closed-loop control.

At Hannover Messe 2026, US-based technology company Dell showcased an XMPro-NVIDIA Omniverse demo in which live PLC data from a brewery centrifuge digital twin fed into an LLM to detect boundary violations and trigger small SCADA-level adjustments within human-defined limits. As shown in the image below, this is not a PoC. This is a real-world example with a real-world brewery: New Belgium Brewing Co. in Fort Collins, Colorado.

“With agentic AI, once you have identified something that is exceeding a boundary, you may allow AI to make very small changes to your production environment. For example, you may allow Omniverse to adjust [a parameter] by only 10%. Anything more than 10% would require a human to be involved. Omniverse can then connect back to the SCADA system to make those small changes—nothing big, only small changes—and notify an operator that it made the change on their behalf.”

Dell Technologies representative at Hannover Messe 2026

Our latest Digital Twin & Simulation Software Market Report 2026–2030 (published in July 2026) shares several examples of agentic AI being used to support digital twins. For example, US-based beverage company PepsiCo uses AI agents to simulate and refine system changes before physical implementation. Through Siemens‘ Digital Twin Composer, PepsiCo can create digital twins of its US manufacturing and warehouse facilities (including every machine, conveyor, pallet route, and operator path) with physics-level accuracy. Within this photorealistic 3D environment, AI agents continuously monitor live shop-floor data, flag anomalies, trace the source of those anomalies to specific physical assets, and recommend corrective actions directly to operators. This agent-driven simulation reportedly allows PepsiCo to rigorously test and optimize layout and operational modifications, successfully identifying up to 90% of potential issues before committing to any physical capital expenditures or changes.

2026 appears to be the start of the agentic AI era

Agentic AI era arrives while industrial constraints slow autonomous operations at scale. 2026 is the year agentic AI has gone commercial, shifting the conversation from “Can we do this?” to “How do we scale, govern, and charge for it?” Free pilots have given way to consumption- and outcome-based pricing, agents have moved to multi-step orchestration, and industrial foundation models, physical AI, and closed-loop digital twins have matured enough to make autonomous operation credible on the plant floor.

Yet, agentic AI has plenty of room to grow, with several noted limitations across our research:

• Data foundation deficits represent a significant bottleneck to scaling agentic AI, according to the Industry 4.0 & Smart Manufacturing Market Report 2026–2030. The Digital Twin & Simulation Software Market Report 2026–2030 notes that this serves as a growth driver for the digital twin market.
• Non-deterministic LLM behavior remains a challenge to customer adoption, especially in high-precision and highly regulated industries with strict validation standards, according to our AI Adoption in Machine Building report.
• Human-in-the-loop remains a core industrial AI requirement, as a bad decision in industry can stop production, damage equipment, and create audit issues, according to the Hannover Messe 2026 event report. For now, agentic AI remains strongest in recommendations, exception handling, and supervised execution.
• While agentic AI is becoming common in industrial software, capable of completing multi-step tasks, multi-agent orchestration remains rare, with most agents working inside one defined workflow, based on observations and discussions at Hannover Messe 2026.

There is plenty of motivation for vendors to tackle these issues. Based on findings in the Industry 4.0 & Smart Manufacturing Market Report 2026–2030, IoT Analytics believes that, aside from the robotics market, agentic AI platforms represent the largest “new” market opportunity in the smart manufacturing market for the remainder of 2026 and beyond.

Further research

Below, in our Insights+ section, we look at a few agentic AI-related insights from the AI-related insights section of our Hannover Messe 2026 event report, including:

• Vendors building foundation models pre-trained on industrial data
• MCP becoming the standard interface between AI agents and industrial systems
• Human-in-the-loop becoming a core requirement of industrial AI