On True North 🧭, I sat down with John Baier, VP Solution Strategy at AVEVA, and one of the architects behind the evolution of the PI System into today’s industrial data platforms.

John has spent decades at the core of industrial data, from Rockwell Automation to OSIsoft, and now AVEVA. He has seen the stack evolve from historians to cloud platforms to AI-driven systems. In this session, we go deep into what actually changed over the last 20 years, what didn’t, and why data reliability is now the limiting factor for scaling industrial AI.

We discuss the transition from PI to CONNECT, the reality of industrial data architectures, and why connecting and contextualizing data is only part of the story if the underlying signals cannot be trusted.

Key takeaways of this episode:

🧠 Industrial data problems didn’t disappear; they scaled

🧱 PI solved trust locally, CONNECT solves scale and collaboration

📉 Data availability ≠ data reliability

🧪 Bronze-layer data determines whether anything downstream works

🤝 Radical collaboration requires data to be graded and understood

🚫 Connecting bad data scales uncertainty

We also covered:

⚙️ Why the industrial data stack is layered, not centralized

🌐 CONNECT vs hyperscalers: complementary, not competitive

📊 Where the medallion architecture breaks in OT

🔍 Why data observability is not enough without validation

🤖 Agentic AI and the risk of scaling bad decisions faster

🧩 MCP and the shift of value from apps to data layers

⚡ AI economics, infrastructure constraints, and what happens next

Julian and I go back to a cold LinkedIn message in 2017.
He'd seen what data validation looked like before it could scale, and he wanted to build something different.

He joined TrendMiner in customer success, moved into product, and now runs product and engineering. Startup, Software AG acquisition, private equity buyout. 

A few things we got into:

⚙️ Belsim → TrendMiner: how validation went from a six-month project to part of an analytics workflow
📈 What actually changes after acquisition (startup → corporate → PE)
🧠 Why industrial AI projects still struggle to move from pilot to production
🤖 The shift from dashboards to copilots to agents
🔗 Where industrial platforms fit in the emerging AI stack

What stuck with me:

– Data availability is not the same as data trust
– "Good enough" worked, but not anymore. 
– Validation was always needed: teams just absorbed the cost quietly
– LLMs are non-deterministic; industry needs consistency and explainability
– Fixing data downstream hides the real operational issue
– The historian gap is real: context gets lost between sensor and model

When we built TrendMiner, trust in the data was a hidden cost. 
With AI in the loop, it's the bottleneck.

Worth a listen if you're building, buying, or running industrial AI.

🎙️ True North Podcast

On True North 🧭, I sat down with Raj Munusamy, Product Portfolio Manager for Data Platforms & Agentic AI at Saint-Gobain, a 350+ year-old industrial giant operating across 1,200+ manufacturing plants worldwide.

Raj is a rare bridge between worlds: automation & control, historians & OT systems, and modern cloud-scale data platforms and AI. In this session, we go deep into what really breaks when industrial data scales, and why data trust, not algorithms, is the limiting factor for AI in manufacturing.

We discuss Saint-Gobain’s Metriks manufacturing data platform, the reality of Bronze / Silver / Gold data layers, and the often-overlooked “historian gap” where data-quality context is lost as signals move from sensors to dashboards and models.

Timeseer.AI engagement 🤝
Timeseer.AI is deployed as the Trust Layer next to the historian, currently live across 70+ Saint-Gobain plants (and scaling) to continuously validate Bronze-layer time-series data, detect issues early, and restore confidence in the data that feeds dashboards and AI.

Key takeaways:
🧠 Data availability ≠ data trust
🧪 The Bronze layer is where trust is won or lost
📉 Missing, stale, drifting data silently kills dashboards and AI
🧩 Historians store data, not confidence
🚫 Auto-fixing data can hide root causes in continuous manufacturing

We also covered:
⚙️ The “historian gap” (Sensors → PLC → SCADA → Historian)
📊 Why teams still spend massive time validating data
🔍 Detect → Score → Resolve → Serve as a trust framework
🛠️ Why Saint-Gobain chose a trust layer instead of building ad-hoc fixes
📈 What it takes to scale data trust across plants, teams, and use cases

No hype. No theory.
 Just what needs to be true for industrial analytics, AI, and autonomy to scale.

🎙️ True North Podcast

On True North 🧭, I sat down with Robert Feldmann, Director of Manufacturing Industry at Microsoft — formerly Associate Partner at McKinsey and today also an industrial advisor to Timeseer.AI.  At Microsoft, Robert helps manufacturers bridge the gap between shop-floor reality and cloud-scale AI, shaping how industrial companies move from automation to autonomy.

Key takeaways:
 🧠 AI is a processing engine.
 ⛽ Data quality is the fuel.
 ❌ No trusted data, no reliable outcomes (+ why).

We also touched on:
 🧩 Fragmented industrial knowledge
 🔁 Broken OT–IT workflows
 ⚡ Energy & compute constraints
 🏭 Why factories that don’t adapt may disappear
 🚀 AI as a platform shift, not a bubble

No hype. Just what needs to be true for AI to work in industrial reality.

🎙️ True North Podcast

On True North 🧭, I sat down with Stijn for a sharp conversation about strategy, AI, and long-term advantage.
We go way back, school friends, and today he’s one of the builders behind a Belgian unicorn in data governance.

What made this conversation special is where our worlds meet: OT and IT data quality and governance. 
The point where operational reality meets enterprise decision-making, and where trust in data becomes non-negotiable.

𝗙𝘂𝗻 𝗳𝗮𝗰𝘁: Amazon reportedly uses Wardley Mapping to support strategic decisions, even M&A thinking.

We talked about:
 🧭 Wardley Mapping to see what’s truly strategic vs already commoditized
🏗️ Why solving today’s pain often creates tomorrow’s constraints
📊 How OT and IT data architectures converge in the age of AI
🤖 The road from today’s AI toward AGI, and what that means for builders
⚖️ Balancing execution speed with strategic clarity

Big thanks to Hans de Leenheer for educating us on Wardley Mapping, and to Simon Wardley for inventing this powerful strategic lens.
If you’re new to it, start here: https://learnwardleymapping.com

𝗦𝗼𝗺𝗲 𝗵𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀:
🌐 Where IT Data Quality meets OT Data Quality, and how trust has different meanings in each world
👥 Who truly owns DQ, and why the technical DQ ≠ business DQ
⚙️ PPT model (People, Process, Technology) — aligning teams and turning strategy into results
🧠 Newton’s Laws of Data Quality, a creative analogy for inertia, force, and acceleration in data programs
🧩 DIY vs set-and-forget tooling — finding the balance between control and scalability
🤖 AI needs DQ, and DQ needs AI, the two-way loop that defines the future
🔁 Closing the loop — using customer feedback to improve trust and reliability continuously
📈 Data Quality trends shaping the next 5–10 years

This episode bridges the digital divide between business systems and industrial data — a conversation for anyone building the foundation of trusted AI.

🎧 The Laws of Data Quality — A Fireside Chat with Gaurav Patole

💬 𝗦𝗵𝗮𝗿𝗲 𝘆𝗼𝘂𝗿 𝘀𝘁𝗼𝗿𝗶𝗲𝘀: Where did data trust (or the lack of it) impact your business outcomes?  The most insightful replies will receive a free copy of Gaurav’s book.

On True North 🧭, I sat down with Bernd Gross, co-founder and CEO of Cumulocity, for an honest talk about the evolution of IoT — from the early hype to today’s AI-powered reality.

Bernd built Cumulocity back in 2012, long before IoT was fashionable, with a bold but straightforward mission: Connect anything, analyze everything, and bring intelligence to the physical world.

Fast-forward to 2024 — the team repurchased the company in a management buy-out, and it now stands as a leader in Gartner’s Magic Quadrant for Industrial IoT platforms.

A few stories we touch:

💡 The real reason behind the TrendMiner acquisition

🏢 Lessons from the Software AG years and the MBO that followed

📈 What it takes to land in Gartner’s top-right quadrant

🕸️ How to partner with hyperscalers who are also competitors

🤖 And Bernd’s bold AIoT predictions for the decade ahead

On True North, I sat down with Richard Beeson, former CTO of OSIsoft, for an honest talk about the company — and the man — that changed both our lives: Pat Kennedy.

Richard spent 32 years at OSIsoft. Before AVEVA, OSIsoft quietly built the nervous system of industry with the PI System,  and it still runs the world’s plants today.

A few stories we touch, without giving it all away:

• 🏢 The dentist's office in San Leandro that set PI’s culture in motion
• 🧩 Why AF didn’t originally mean “Asset Framework”
• ☎️ The rule about answering the phone before the 3rd ring 
• 💾 How PI v1 was shipped… 
• 🕰️ The moment Pat realized people valued history more than the project itself
• 🔁 The quiet design decision that made PI 3 a runaway success
• 🔓 Pat’s licensing ethos

It’s a candid conversation and a small tribute to a founder who built for engineers first and built something that kept working.