It's 2:47 PM on a Tuesday. Somewhere in your Beverages division, the carbonated line just stopped. Your energy drinks production is backed up. Bottled water is running overtime to compensate, and you're about to blow through your sanitation window.
You won't know about any of this for another 45 minutes.
When you finally do, you'll ask the question every division leader asks: Why did the carbonated line stop, and what's it costing me right now?
The answer will require three phone calls, a walk to the floor, and someone pulling data from two different systems. By then, the ripple effects have already spread across your shift.
Who Gets to Ask Questions?
As Prove It 2026 approaches, I keep thinking about last year's presentations. The dashboards were beautiful. The AI announcements were plentiful.
But no one addressed the fundamental problem: Who gets to ask questions?
Every system I saw was built for the automation engineer. The person who already knows which tags to query, which historian to pull from, and how to write the syntax to get answers.
What about the division leader who just needs to know why production is hemorrhaging margin?
AI Without Wisdom Is Just Fast Guessing
Here's what most industrial AI demonstrations get wrong: they show you what AI can do without showing you how it knows.
AI doesn't have wisdom. It has pattern recognition and processing speed. Wisdom comes from the people who understand what a 3% OEE drop actually means for quarterly numbers. From the operators who know the carbonated line always acts up after a long sanitation cycle. From the quality leads who can tell you which downstream processes will be affected if you push that batch through anyway.
The question I'll be asking at Prove It: Which systems are designed to extract operational wisdom from people beyond just automation engineers?
This matters because the people with the most contextual knowledge about your operation, the ones who understand root causes and downstream impacts, are rarely the ones who know how to query a historian.
Two Kinds of Visibility
Consider what happens when a yield issue surfaces on the energy drinks line. The shift supervisor knows something is off but can't articulate it in terms the system understands. They escalate. Someone creates a ticket. An automation engineer gets pulled from another project. Hours pass.
Meanwhile, the division leader is looking at end-of-day numbers trying to understand why throughput dropped. Unaware that the root cause was identified and then lost in translation somewhere between the shop floor and the database.
This is actually two visibility problems masquerading as one:
The data problem: Production systems speak in tags, fault codes, and historian queries. The people who need answers don't speak that language.
The people problem: Organizational knowledge lives in people's heads. Finding who actually knows the answer, versus who will just relay your question down the chain, requires politics, tenure, and luck.
Solving one without the other leaves you halfway there.
Context and Guardrails
The AI systems that will actually transform manufacturing won't be the ones with the most sophisticated algorithms. They'll be the ones with the most sophisticated understanding of who is asking and what they actually need to know.
When a division leader asks "Why did the carbonated line stop?", they want to know:
- What stopped?
- Why?
- What's the production impact?
- What's the cost impact?
- Who's working on it?
- Who actually knows about this problem?
When an automation engineer asks the same question, they want tag values, fault codes, and historical patterns.
Same question. Different contexts. Different answers needed.
This is what I mean by extracting wisdom through guardrails. The system doesn't become enlightened on its own. It becomes useful when it bridges raw operational data to human decision-making. And when it bridges the right questions to the right people.
What I'll Be Looking For
At Prove It this year, I won't be impressed by dashboards.
I'll be looking for systems that answer the questions keeping division leaders up at night:
- Why did we miss target on Thursday?
- What's driving variance between my energy drinks and bottled water lines?
- If this downtime continues, what does it mean for my monthly numbers?
More importantly, I'll be watching who can ask those questions. If the answer is "only people with technical training," we haven't solved anything. We've just built faster tools for the same small group.
And I'll be watching whether those systems can help me find the people who have the answers. Without the three phone calls and the organizational guesswork.
The question isn't whether AI can transform manufacturing. It's whether we're building systems that make operational wisdom accessible. Both the wisdom locked in our data and the wisdom locked in our people.