IIoT World examines why most industrial AI initiatives stall at roughly 40% of their potential, and what manufacturers who push past that wall do differently with contextual intelligence on the shop floor.
Why Many Projects Stall and What Moves Them Forward
Endurance athletes talk about the “40% wall.” It is the point where the body feels spent, even though much more capacity remains. David Goggins uses it to describe the moment when discomfort is mistaken for a hard limit.
Industrial AI projects hit a similar wall.
Manufacturers deploy analytics, connect machines, and pilot AI models. Early results look promising. Then progress slows. Leaders question the return. Teams move on to the next initiative. The technology did not fail. The system simply reached the edge of what it could do without a deeper understanding.
At CES 2026, in a conversation with Del Costy, President and Managing Director, Americas, at Siemens Digital Industries, this pattern came up repeatedly. The issue is not ambition or tooling. It is context.
Where Industrial AI Actually Hits the Wall
Factories generate massive amounts of data. Speeds, temperatures, alarms, cycle times. Most AI initiatives start here. They assume that more data leads to better decisions.
What often goes missing is meaning.
AI struggles when it does not understand what the factory is trying to do at that moment. Which product is being built. Which configuration is running. What constraints apply to this line, today, right now. Without that context, insights stay abstract and hard to act on.
This challenge becomes sharper in brownfield environments. Existing plants carry years of equipment, process variation, and operational workarounds. Leaders cannot pause production to rebuild systems from scratch. Progress depends on layering intelligence into what already exists and tying it directly to daily production problems.
The 40% wall appears when data flows freely but understanding stops short.
The Second Wind: Context on the Shop Floor
Momentum returns when data becomes contextualized. When systems connect signals to product intent, process conditions, and operational goals, AI starts supporting real decisions on the line.
This work does not start with advanced models. It starts with knowing how the shop floor actually runs. What matters to the line manager during a shift. What causes quality to drift. Where throughput slows under real-world conditions.
Context turns AI from a reporting tool into a production partner.
What Changes After the Wall
The shift past 40% happens when intelligence moves closer to where work is performed.
Del Costy describes a model where production data is validated against a digital representation of the process as it runs. Simulations are no longer limited to design or planning. They operate alongside production, increasingly at the edge, where data is created. This allows conditions to be checked, adjusted, and understood within the cycle of work itself.
At the same time, machines begin to understand the product, not just the motion. Product variation, configuration changes, and real-world disruptions become part of how the system reasons. That is the point where automation gives way to adaptive behavior.
This is also why Siemens focuses on its own factories as test environments. The limiting factor is not algorithms or cloud capacity. It is the delay between data creation and operational insight. Reducing that lag is what allows AI to support decisions that matter during production, not after it.
The 40% wall in industrial AI is not a ceiling. It is a signal. Progress accelerates when systems understand the work, the product, and the conditions under which production happens.
Sponsored by Siemens
Frequently Asked Questions
1. Why do most industrial AI projects stall?
Most industrial AI projects hit a wall at roughly 40% of their intended scope. The primary reason is that AI models trained on decontextualized data produce outputs that operators do not trust enough to act on. Without understanding the operational context behind sensor readings, AI predictions remain too noisy for production use.
2. What is contextual intelligence in manufacturing AI?
Contextual intelligence adds operational meaning to raw data. Instead of feeding an AI model a temperature reading in isolation, contextual intelligence attaches the equipment state, product being manufactured, shift conditions, and historical maintenance records. This context reduces false positives and makes AI outputs actionable on the shop floor.
3. How do manufacturers move past the 40% wall in industrial AI?
Manufacturers that scale past 40% invest in three areas: data contextualization layers that add meaning to raw sensor data, feedback loops where operators validate and correct AI outputs, and incremental deployment strategies that build trust across specific production lines before scaling plant-wide.
