Some guests make you pause halfway through the recording and think, “Okay… this one’s going to need a second listen.”
That was the case with Dan Jeavons, president of Applied Computing, formerly VP of Computational Science and Digital Innovation at Shell — and one of the people who has quite literally been shaping how data, AI, and physics come together in industry.
From ERP Reports to Foundation Models
He began, like so many, somewhere between spreadsheets and SAP.
“The biggest value of having an integrated system is the fact that you have an integrated data layer,” he recalls. “I didn’t like the systems much — but the data was really interesting.”
That curiosity led him from analytics experiments in R and MATLAB to building Shell’s first Advanced Analytics Center of Excellence — which, as he jokes, “was neither advanced nor excellent… but we got better quickly.”
Thirteen years later, he was leading teams across AI, data science, and advanced physics modeling — and wrestling with a problem that every industrial data leader knows too well:
“You either rely on physics and trade off flexibility, or you rely on statistics and trade off explainability.”
What AI Looks Like From the Plant Floor
Dan has worked across the energy value chain — from offshore wells to refineries — and says something that surprises many:
“From a data perspective, it all looks very similar.”
Distributed control systems, process historians… “whether you’re on a platform in the North Sea or in a petrochemicals plant, the data architecture doesn’t really change,” he says.
And that’s what makes the AI opportunity so big.
If every facility generates data in roughly the same way, then algorithms can be adapted and scaled — not rebuilt from scratch each time.
Why IT/OT Convergence Still Hasn’t Happened
At one point, we asked the question: Has IT/OT convergence really happened?
Dan didn’t hesitate:
“No. We’re only scratching the surface.”
He describes today’s operations as “a DCS at the heart of the operation, surrounded by siloed engineering processes — reliability, maintenance, safety — each with their own tool, using a fraction of the data.”
Adding AI layers on top of that, he argues, is helpful but incomplete:
“We’ve added a layer of intelligence on top of existing systems. But it hasn’t changed the work process yet.”
True convergence, he says, will come when AI doesn’t just analyze the work — it redefines it.
The Real Meaning of “Digital Twin”
Few topics create more buzz (or confusion) than digital twins. Dan gives one of the clearest definitions we’ve heard:
“A true digital twin must do three things: represent the physical world, be interrogable in real time, and run simulations that explain why and what next.”
That’s a high bar…
“The technology exists,” he says. “We just haven’t stitched it together yet.”
Change Management: The Hardest Part
Dan’s third “impossible problem” isn’t technical — it’s human.
“These facilities are extremely risky. They’ve run safely for 40 years. So when you say, ‘Let’s change everything,’ it’s a hard sell.”
He lays out the classic resistance:
* It works, don’t touch it.
* We can’t risk downtime.
* We’re here to deliver return on capital, not to experiment.
And yet, as he points out:
“Even with the way we run things today, we still have reliability problems, we still have safety exposure, and we’re losing expertise fast.”
His conclusion is blunt:
“Someone is going to figure this out — and when they do, they’ll be 50 % more efficient. If you’re not on that train when it happens… good luck.”
Rethinking the Cloud Debate
When the topic of cloud reliability came up (AWS outages, anyone?), Dan didn’t dodge.
“The idea that you’re safe because you’re air-gapped is a fallacy,” he said flatly. “Most OT environments are already virtualized — effectively private clouds. The question isn’t if you’re exposed, it’s how well you manage it.”
The challenge, he says, isn’t cyberthreats — it’s change management in the cloud era.
“Continuous deployment doesn’t work in operations. We need cloud architectures that respect industrial change control — and OT vendors who step up to modern security standards.”
From Use Cases to Foundation Models
Dan’s view of AI’s future is clear: we’re moving from narrow, use-case-specific algorithms to general-purpose foundation models that can reason across disciplines.
“Before 2023, companies built algorithms for individual problems: corrosion, valves, compressors. Now, the next generation of models will handle all of them because they understand physics, language, and time series together.”
He tells the story of Sam Tukra, his former colleague (now Applied Computing’s co-founder and Chief AI Officer alongside Callum Adamson) who figured out how to make those three domains “talk” to each other.
“He built an agentic system that cross-validated physics, language, and time series. I was equal parts proud, frustrated, and amazed. Suddenly, you realize — this is it.”
The result is Orbital, their platform that blends these layers — a system that can predict, explain, and reason across disciplines, from reliability to safety to economics.
Looking Ahead
Dan calls this convergence of physics and AI an “inflection point for industry.” He’s convinced that in the next decade, the companies who embrace it will operate differently — not because AI tells them what to do, but because it changes how they work.
So that means that we need to plan for another podcast in a year or so from now ;)
Thanks for listening!
Stay Tuned for More!
Subscribe to our podcast and blog to stay updated on the latest trends in Industrial Data, AI, and IT/OT convergence.
🚀 See you in the next episode!
Youtube: https://www.youtube.com/@TheITOTInsider Apple Podcasts:
Spotify Podcasts:
Disclaimer: The views and opinions expressed in this interview are those of the interviewee and do not necessarily reflect the official policy or position of The IT/OT Insider. This content is provided for informational purposes only and should not be seen as an endorsement by The IT/OT Insider of any products, services, or strategies discussed. We encourage our readers and listeners to consider the information presented and make their own informed decisions.
