When AI Hits the Data Wall: The Lesson Industry Can’t Keep Postponing

68% of manufacturers already use artificial intelligence. But the report confirming that number also explains why, without reliable data, that investment stays half-finished.

Last updated: August 2026

KPMG has just published its Global Tech Report 2026 on industrial manufacturing, based on responses from 258 technology leaders across 22 countries. The headline is encouraging: 68% of manufacturers already deploy AI somewhere in their operations, and 49% report concrete financial benefits from those investments. The old narrative of AI stuck in isolated pilots is finally fading.

But the report itself points to the catch. 83% of executives believe they’re building a solid data foundation to support AI. Yet 76% still cite unreliable data as one of the top risks to their AI implementation. That gap between confidence and actual capability is arguably the most important number in the whole study.

AI isn’t failing for lack of ambition — it’s failing for lack of foundations

The pattern repeats across plants worldwide: sensors that don’t speak the same language, legacy systems that don’t integrate, information scattered across ERP, MES, and parallel spreadsheets. AI is only as good as the data feeding it, and right now, in most factories, that data remains fragmented. 41% of companies surveyed by KPMG rank operational efficiency as their top priority — but chasing efficiency on unreliable data is, at best, optimizing blind.

Reliable data, measurable sustainability

This is where innovation and industry connect directly to sustainability. Digital twins — virtual replicas of a plant or process that let you simulate before you act — can deliver energy savings of up to 30%, according to recent studies on their industrial application. But a digital twin built on inconsistent data predicts nothing; it just moves the noise into a more expensive model. Industrial sustainability — cutting consumption, anticipating maintenance, matching production to real demand — isn’t achieved with more sensors, but with governed data: traceable, integrated, and consistent across plants.

It’s no coincidence that the International Energy Agency projects industrial electricity demand to keep growing strongly through 2030. Every percentage point of efficiency that AI and digital twins can claw back from that consumption has a real, measurable impact — in euros and in tonnes of CO2. But only if the data foundation behind it is genuinely solid, not just perceived to be.

Data governance: the next frontier of industrial engineering

The message for 2026 is clear: the conversation is no longer “should we adopt AI?” but “do we have the data foundations for that AI to generate real, sustainable value?” The companies seeing returns aren’t necessarily the ones that invested most in algorithms — they’re the ones that got their data house in order first. In industrial engineering, that means interoperable data architectures, consistent sensor networks, and a governance strategy as carefully designed as the physical infrastructure itself.

At Talat, we see this every day in industrial digitalization projects: even the most advanced technology underdelivers when it rests on unreliable data. Conversely, a well-designed data foundation multiplies the value of any investment in AI or digital twins.

Does your organization actually know what percentage of its industrial data is truly reliable? It’s probably the most profitable question you can ask yourselves this quarter.

Explore similar insights