Industrial AI
The Roadmap from Chat to Virtual Operator
Forget the autonomous-control dream you’ve been sold. Industrial AI’s first real win will be autonomous investigation: agents that work through your data, documents and history so your engineers get started a lot faster.
The untapped half
Everything that doesn’t come from a sensor
We’ve spent years turning signals into data: building historians, contextualising tags, optimising how we handle time series. What we mostly left untouched is everything else. Before generative AI, none of that was something you could automate. You needed a person to hold it all in their head.
- Manuals
- Procedures
- Shift reports
- Maintenance notes
- Incident reports
- P&IDs
- SAP work orders
- Emails
- Operator know-how
Think of the things an experienced operator just knows: low pressure plus high temperature plus high rotation speed means watch out for cavitation.
Autonomous investigation
Why has Pump P101 been losing performance for three weeks?
Building software to answer that used to be a Herculean integration project. Now, an agent can (in theory) do the legwork.
Traditionally
The maintenance engineer
- Find the pump
- Pull up the P&ID
- Query the historian for anomalies in process conditions over the past weeks
- Check OEE and downtime events
- Dig through SAP maintenance history
- Find the manufacturer’s manual
- Look at the P&ID again
- Read the shift reports
- Mentally correlate everything
Now
The agent
- Read the shift reports
- Query the OEE database
- Retrieve the SAP records
- Look at the historian trends
- Find the pump manual
- Understand the P&ID
- Combine the evidence into a first-pass analysis, or help the engineer brainstorm
Will it replace the engineer? Almost certainly not, at least not beyond the obvious cases; even with all the data in hand, a proper investigation can still take a dedicated team days or weeks. But it can get them started a lot faster.
And that matters more every year. Many “super experts” are retiring, and it’s harder and harder to find people who stay within one plant their entire career (you know, the ones who can smell what the problem is). That’s why the toolsets popping up now become extremely valuable, especially for the younger generation and the everything-engineers.
If an agent can do all of that on request, what’s stopping it from doing it continuously, on its own? Watching your data around the clock, surfacing an anomaly, drafting a maintenance notification, summarising a shift, finding similar past incidents, recommending a procedure, preparing a work order. That’s a far easier problem than closing the loop with autonomous control, and it doesn’t require redesigning your brownfield plant to get there.
Watch
Can AI run your plant?
How should you start?
The Industrial AI Maturity Model
From people chatting with their own AI tools at home to a virtual engineer: four levels, moving from human centric to system centric. Between every two levels sits a stepping stone, the things you need in place to move up.
Click a level, or a ◆ stepping stone between two levels, to explore the model. You can also use the arrow keys.
Level
L0 · Shadow AI
What
People using their own personal AI toolset (the ones they know from home).
Risk
- Governance / legal: data exposure to third parties!
- Scaling: does not scale.
Stepping stone
From L0 to L1
Stepping stone
People use the company-provided AI toolset to create reports, small applications, etc. through chat.
Risk
A very simple step to take, but we should distinguish between what an individual can achieve and what the organization can learn.
Level
L1 · Chat · Individual
What
“Excel on steroids”
Risk
When these AI initiatives stay with the individual, this is the final station of your AI maturity.
Stepping stone
From L1 to L1+
Stepping stone
Community / excellence centers are set up to work on standards, best practices and a common language, and to build expertise.
Level
L1+ · Chat · Organizational
Examples L1+
- RCA bot
- 5-Why bot
- HAZOP support bot
- Training instructor bot
- …
Key message
This is your ultimate prototyping and learning level!
Stepping stone
From L1+ to L2
Stepping stone
- Centrally governed data and AI platforms (→ see our Capability Maps)
- Strong strategic collaboration with IT (→ see our Cooperation Models)
Level
L2 · Reason
This is where autonomy enters the picture!
Examples L2
- An anomaly is detected automatically
- The system checks all knowledge
- It lists potential issues
- It sends an operator to collect more data (you will never have enough data)
- It creates and assigns the right work order with the right priority
Stepping stone
From L2 to L3
Stepping stone
- Crystal ball: don’t try to predict what will happen (nobody knows).
- But… the number one requirement: strong cooperation between IT, Data/AI and Maintenance.
Level
L3 · Virtual Engineer
The future
Deliberately left open. Nobody knows yet what a virtual engineer will look like — but the plants that get there will be the ones that built the stepping stones first.
★ What it takes
This isn’t something IT or OT can solve alone
On the technology side: contextualisation, asset models, naming conventions, DataOps discipline, and increasingly the ontologies and knowledge graphs that let tacit engineering knowledge sit alongside your data rather than only in someone’s head. Our Industrial Data Platform Capability Map is the place to start.
On the organisational side, it’s simpler to state and harder to do: if both sides keep working in silos, none of this happens. We’ve written about that at length in our “What is IT/OT convergence?” explainer.
Starting today is always the better option. Picture two factories, ten years from now. One still has its historian, SAP, manuals, shift reports and engineering documents sitting in their own silos. The other has agents continuously working through those same sources, surfacing anomalies, opportunities and maintenance insights as a matter of course. You probably don’t want to be the first factory.
Go deeper
The full article series
Part 1
Can AI Run Your Plant?
Yes, if you believe the marketing. In this new video, we talk about what is actually possible today, and why “autonomous investigation” beats “autonomous control”.
Part 2
Two Problems Standing Between You and Industrial AI at Scale
Two fundamental challenges keep the industry stuck in pilot mode — and neither of them is a lack of algorithms.
Part 3
The Three Faces of Industrial AI
Numbers (Predictive AI), words (Generative AI) and actions (Prescriptive AI): a simple way to frame Industrial AI so engineers, operators, and managers can all speak the same language.