Representative End User Clients
Representative Automation Clients
Representative Software Clients
The energy at AWS re:Invent 2025 in Las Vegas was undeniable. The narrative has shifted decisively from the passive "Industrial Copilots" of last year to active, autonomous Agentic AI. Everywhere you looked, the talk was of autonomous swarms, "Supervisor Agents," and the new Strands Agents SDK empowering developers to build digital workers that can reason and act.
However, as I noted in my recent Top 10 Takeaways for Industrial AI at AWS re:Invent 2025, there is a stark divergence between the "hyperscaler dream" and the "brownfield reality." For manufacturers dealing with 20-year-old assets, proprietary protocols, and disparate silos, building these agents often hits a brick wall—the "Factory Wall." An AI agent is only as intelligent as the context it is fed, and in most factories, that context is trapped in the machine layer.
To explore how the industry is breaking through this barrier, I sat down at re:Invent with Joe Rosing, Worldwide Head of Smart Manufacturing at Amazon Web Services (AWS), and Torey Penrod-Cambra, Co-Founder and Chief Communications Officer at HighByte. We discussed the critical role of Industrial DataOps in untangling the mess of industrial data to fuel the next generation of AI.
Listen on BuzzSprout:
Watch on YouTube:
Below are the key takeaways and insights from our discussion at AWS re:Invent 2025.
The "Last Mile" Data Problem: Context is King
In our discussion, we addressed the elephant in the room: the gap between the promise of Generative AI and the reality of the factory floor. We often see customers attempting to dump raw telemetry data into a data lake, hoping an LLM will magically find the patterns.
Joe Rosing framed this perfectly. He argued that we aren't necessarily reinventing the discipline of engineering, but we are fundamentally changing how we execute it.
"It is traditional systems engineering, but now you're doing systems engineering with new tools in the toolbox. Without adding context to the prompt, you just don't get an accurate answer."
— Joe Rosing, Worldwide Head of Smart Manufacturing, AWS
This aligns with ARC’s research on the Industrial Data Fabric (IDF). You cannot simply bypass the hard work of data modeling. If you feed an AI agent raw vibration data without telling it which asset produced it, what product was running, and who was operating the line, you don't get intelligence—you get hallucinations.
Industrial DataOps: Bridging the IT/OT Divide
This is where HighByte enters the equation. HighByte has been a pioneer in what ARC defines as the Industrial DataOps-Centric archetype of the Data Fabric.
During the podcast, Torey Penrod-Cambra clarified a crucial distinction. Traditional DataOps originated in IT to improve data quality for analytics. Industrial DataOps, however, must handle extreme data variety—telemetry, transactional, time-series, and files—and add context to that data while it is in motion at the edge.
"It's where the edge meets the cloud. It's where OT meets IT. It's preparing that industrial data so it's ready to be consumed in powerful, scalable AWS services. We're just helping to make that simple for customers. We are trying to provide sample reference architectures and solid case studies to help other manufacturers get started."
— Torey Penrod-Cambra, Co-Founder, HighByte
This approach aligns perfectly with AWS’s recent announcements regarding AWS IoT SiteWise Edge enhancements. By using HighByte to model and standardize data at the edge—creating a "Unified Namespace" (UNS) structure—manufacturers can feed clean, contextualized data directly into AWS services like SiteWise and TwinMaker. This prevents the cloud from becoming a "data swamp" where data scientists spend 80 percent of their time cleaning timestamps instead of building models.
Speed as a Leadership Decision: The Gousto Benchmark
One of the most compelling parts of our conversation was the focus on speed to value. In my Top 10 Takeaways, I highlighted the "Builder" ethos as a competitive advantage. It is not enough to buy AI; you must build proprietary value with it.
Torey shared a powerful example of this "Builder" mindset in action with their customer, Gousto.
"They started with a very simple use case - reducing the mean time to repair. Within a year, they had gone beyond reactive to preventative, to prescriptive, to putting together entire simulation models—moving their OEE from 70 percent to 95 percent in the first six months. It's an incredible use case, but I don't think it's unique. We have many customers that are doing the same thing."
— Torey Penrod-Cambra, Co-Founder, HighByte
This demonstrates the compounding value of a strong data foundation. Once the data fabric is established for one use case (maintenance), it accelerates the next use case (simulation/process optimization) because the data is already modeled and accessible.
The Strategic Imperative: Don't Wait for Perfect Data
A recurring theme in my conversations with industrial leaders is "paralysis analysis." They know their data is messy, so they hesitate to start AI projects until they have a perfect, monolithic data warehouse.
Joe Rosing pushed back on this, urging leaders to adopt a more agile approach.
"Speed is a leadership decision. I challenge a number of our customers with that same kind of sentiment. 'Should I wait till all my data is clean before I get started, or do I just jump in?' Really, it's a leadership decision around how fast do you want to transform? The tools are available. Let's pick one outcome. Let's work backwards from that outcome so we have purpose and context for the data that we're going to use."
— Joe Rosing, Worldwide Head of Smart Manufacturing, AWS
Key Takeaway: No Context, No Agency
As we move from the era of "Industrial Copilots" to "Agentic Factories," the role of the Industrial Data Fabric becomes paramount. You cannot have autonomous agents making decisions based on bad data. To leverage the Strands Agents SDK or build a "Maintenance Agent," that agent needs to know that a specific vibration reading came from "Pump 3 on Line 1" and correlates with a specific work order in SAP. HighByte automates the creation of this context.
My key takeaway from this conversation matches the guidance from our recent Industrial AI Pacesetter Research: Do not wait for a perfect, monolithic data system. Start building your Data Fabric now by assembling best-of-breed components. Whether you use HighByte to model data at the edge or AWS SiteWise to store it in the cloud, the goal is the same: providing the trusted, high-quality fuel required to power the Industrial AI revolution.
Related AWS Blogs, Podcasts, and White Paper
AWS re:Invent 2025: How the Industrial Cloud is Becoming Physical
Industrial AI SPARC: Software Defined Manufacturing in the Era of Agentic AI
Software Designed Manufacturing, Episode 2: From Code to Physical and Embodied AI
Industrial AI SPARCs with AWS: Episode 3, Digital Twins as the Engine for Physical Intelligence
Whitepaper: AWS Industrial Data Fabric (Industrial Data Fabric): A Blueprint for Success with Industrial AI
Stay Connected and Contribute to the Conversation
The dialogue on industrial AI and digital transformation is constantly evolving. To stay informed and hear more insights from industry leaders, we invite you to subscribe to the ARC Advisory Group's Digital Transformation podcast series.
We believe the best conversations include diverse perspectives. If you are an innovator in this space and would like to contribute to a future discussion, please reach out to Colin Masson at ARC Advisory Group.
Engage with ARC Advisory Group
For ARC Advisory Group recommendations for Navigating the AI Wars—including the Industrial Robot Wars—Closing the Digital Divide by Embracing Industrial AI, assembling your Industrial-Grade Data Fabric, and governing and guiding major decisions about enterprise, cloud, industrial edge, and AI software, please contact Colin Masson at [email protected].
Or set up a meeting with my fellow Analysts and I, at ARC Advisory Group to find out more about our Executive Insights Service for Industrial organizations, and Industrial AI Insights Service for vendors.