AI adoption
that survives
contact with
reality.
AI is the new layer
of automation.
Demos dazzle. Production humbles.
The gap between “look what it can do” and “it quietly does this for us every day” is enormous. Most AI pilots die in that gap. Closing it is a discipline, not a demo.
Automation is a way of thinking
How do you hand work to a system you trust, keep humans in control, and survive 3 a.m. when no one's watching?
Industrial or not — same questions
Every organization adopting AI faces the exact problems I've spent a career solving on the plant floor.
The value is in the seam
Real operations and modern AI rarely share a person. I work where the two meet — and have to hold.
Two worlds that rarely
share a person.
Eighteen years building automation that wasn't allowed to fail — and two years deep in modern AI. The value lives in the seam between them.
Automation that can't fail
Control systems on offshore platforms, safety systems in pharma plants, machining lines at one of the world's largest truck manufacturers. PLCs, SCADA, commissioning from first wire to final sign-off.
Agentic systems & RAG
Two years building, not spectating: agentic systems, retrieval-augmented generation over messy real-world data, structured LLM integration, model evaluation, fine-tuning, and local inference.
Trustworthy by design
In safety-critical work, a system that breaks in production isn't a disappointment — it's a hazard. You learn how to make automation survive the messy, high-stakes conditions of the actual floor.
Where operations meets AI
Most people have one half — real operations or modern AI. The value is in the seam: getting these systems to do useful work reliably, in the place where they meet reality and have to hold.
Bring me a real problem.
I'll tell you what I'd do.
Reach out
You bring the real situation — as messy and specific as it actually is. No pitch, no package required.
Honest take
I tell you honestly whether it's something I can help with, and how I'd actually approach it.
A real plan
If it's a fit, we go further — what to automate first, where AI pays off, how to keep humans in control.
Clarity either way
If it isn't a fit, you still leave with a clearer picture than you came in with. That's the deal.
Connect any system.
Control any process.
From PLCs and SCADA to LLMs and agents. Control systems, IT/OT integration, RAG over operational data, and the practical craft of making it all hold.
From PLC to LLM
I connect operational data living in awkward places to agents that can actually use it.
retriever = RAG('scada_logs')
agent.use(retriever)
agent.run('why did line 3 trip?')
I don't just advise — I build, commission, and ship systems that run under real conditions.
Humans in control
of what matters.
The boring discipline that decides whether automation is trustworthy or dangerous — borrowed straight from safety-critical engineering.
Oversight by design
Humans stay in control of the decisions that matter
Full audit trails
Every action logged, every decision traceable
Safe failure modes
Adopt AI without taking the operation down
I build, not just advise.
Here's how the work goes.
Notes from the seam
between ops and AI.
Writing about what it actually takes to adopt AI that lasts — from someone who spent eighteen years making automation survive the real world. New piece most weeks.
Got a real problem?
Ask me.
No pitch, no package. If you're trying to adopt AI for real, reach out — or subscribe to the writing and reach out whenever you're ready.