Why Every Industrial Company Needs to Become a Data Streaming Company

When HiveMQ CEO and Chairman Barry Libert joined HiveMQ, it wasn’t because data streaming was a familiar category. It was because he recognized something that most organizations still haven’t: the companies that build a reliable operational data foundation now will compound their AI advantage for years. The ones that wait will spend that time catching up.

Barry joined the AI & IoT Leaders podcast recently, where he shared what 48 years building and advising companies across every industry has taught him and why that’s more relevant than ever for industrial AI. Here are the takeaways: 

1. There Is 50 Times More Untapped Data in Devices Than in All Digital Content

AI training so far has been built on text, audio and video - written human knowledge. But by many estimates, physical devices and operational systems contain roughly 50 times more data than all that content combined. Most of it has never been captured, let alone connected to an AI model.

Barry has seen this gap from both sides. Before joining HiveMQ, he had worked with data streaming companies in the human communication space - Webex, AT&T - and watched them sit on enormous data pipes without extracting intelligence from them. AT&T and Webex were streaming vast amounts of data between people and had almost no understanding of what was passing through their pipes. That, he notes, is not a neutral outcome. Broadcast without intelligence is a commodity. It gets commoditized and then it loses. 

Data streaming between devices - machines, sensors, factory equipment - combined with the maturity of AI models to do something with that data. That is the convergence that is reshaping industrial operations.

"The amount of data that is going to accrue to large language model providers, edge providers, and cloud providers from operational systems is going to be exponential." - Barry Libert, Chairman and CEO, HiveMQ

2. What Does Operational Intelligence Actually Look Like in Practice?

The ‘enterprise brain’ describes what becomes possible when real-time operational data is connected, contextualized and made queryable.

He uses two examples:

The first is what HiveMQ has built internally. HiveMQ has constructed a real-time operational view that spans device activity, communications, financial flows, sales and marketing data and team collaboration - all in a unified environment. What might once have required a multi-million-dollar ERP implementation and a consulting team, now runs on a platform built with modern AI coding tools. It is a working demonstration of the model it sells.

The second example comes from a healthcare context shared on the podcast: a company that instrumented an entire hospital - patients wearing wristbands, nurses and physicians carrying sensors, every piece of equipment tagged - and built an ontology not by department but by process: a hip operation, a pharmacy dispensation, a triage pathway. Collectively, this created a real-time process dashboard that revealed, for the first time, where the actual bottlenecks were. Surprisingly, these were not in treatment, but in medication collection - something previously invisible to any existing system.

That is what operational intelligence looks like when the data foundation is right. Not a dashboard for its own sake. A tool that shows what was previously hidden and enables action in time to matter.

3. How Does Data Streaming Connect to AI - and Why Does the Order Matter?

You cannot shortcut to AI outcomes without getting the data layer right first. Streaming is not the interesting part but the prerequisite.

The platform journey that HiveMQ describes - connect operational data, contextualize it, analyze it, then act on it - mirrors the progression Barry has watched happen in every technology transition he has been part of. You stream the data. You build the ontology, the shared understanding of what that data means across systems and teams. You create the semantic graph that allows machines, operators, and AI models to reason over it. Then, and only then, do you get to autonomous action at the edge.

Machines will receive signals AND they will send signals upstream to the supply chain, to maintenance teams, to the systems that schedule parts before a breakdown occurs. This is the natural extension of what already happens with consumer devices: Barry’s Tesla tells him when it needs attention before anything fails. The industrial version of that, at scale, across a factory floor or a network of facilities, is what the platform journey is designed to enable.

"I believe the first foundational level is data streaming. We have to stream the data - not just between people, but between devices and devices in the cloud and devices on the edge - and then take the insights and allow the devices to operate more efficiently and autonomously." - Barry Libert

4. What Is the Role of Humans as AI Takes on More Cognitive Work?

Barry is, by his own description, an optimist. He has watched every wave of technology produce a version of this debate - “technology will take the jobs, humans will become obsolete” - and in 48 years, he has seen unemployment stay low in mature economies even as entire categories of work were automated away.

His view on what changes this time: AI forces humans to get precise about what they actually contribute. The knowledge workers who will thrive are not the ones who can execute tasks faster - AI will do that. They are the ones who can frame the right questions.

That framing insight comes from his McKinsey background. The consultant's value was never the answer; it was identifying what question to ask in the first place. It’s a skill that does not get automated easily. Judgment about what matters, what to measure and what to do when the ontology reveals something unexpected remains a human contribution.

What he acknowledges is the gap. New jobs will emerge from this transition because they always have. But there is a valley between the speed at which AI is replacing current roles and the speed at which new roles become clear. Organizations and governments that recognize that gap and act to close it, will produce better outcomes than those that treat disruption as someone else's problem.

5. What Should Operations Leaders Do Right Now?

Barry's summary for any industrial organization thinking about AI readiness is direct: start with data streaming. 

Data streaming is not the end goal - it is the gate. You cannot build the ontology, create operational intelligence, or get to autonomous action without it. 

The customers HiveMQ works with are already asking the right questions: Can you help us visualize all our devices, all our data flows, all our human operators in real time? Can you help us build an ontology that makes sense of it all? Can you help us begin to automate against that understanding?

Those questions, in that order, describe the platform journey. And the organizations that start asking them now, rather than waiting until the pressure to act becomes unavoidable, are the ones that will be in a position to act when the decisions that matter most arrive.

Data Streaming Is a Strategic Decision, Not a Technical One

Barry has helped build companies across 48 years and four decades of technology cycles. His conclusion about this moment is consistent with what he has seen at every inflection point: the organizations that move early to adopt the foundational layer of the new architecture compound their advantage. The ones that wait spend years trying to close a gap that keeps widening.

In this cycle, the foundational layer is real-time operational data: Streaming it reliably, contextualizing it into a shared understanding, and making it available to AI systems that can reason over it and act on it. That is the sequence. That starts with getting the data moving.