98% of Manufacturers Are Exploring AI. Only 20% Are Ready.

Key Points

• 75% of manufacturers expect AI to be a top-three contributor to operating margins by 2026, but only 21% say they're actually ready to make that happen.
• The readiness gap isn't about technology. It's about data. Siloed systems, legacy ERPs, and inconsistent formats are the real blockers.
• Manufacturers that start with data cleanup and one connected use case will outperform those chasing flashy AI projects without a foundation.

I talk to manufacturers every week who tell me some version of the same story. They know AI matters. Their board is asking about it. Their competitors are talking about it. They've maybe run a pilot or two. But when it comes to actually deploying AI in production, they hit a wall.

A new study from TCS and AWS just put numbers behind that wall. They surveyed 216 senior manufacturing leaders across North America and Europe, spanning automotive, aerospace, industrial machinery, and process industries. The headline: 98% are exploring AI. Only 21% say they're ready for it.

That's a gap you can drive a forklift through. And it's exactly the gap I see in the manufacturers we work with at Rotate.

What "Not Ready" Actually Means

When the study says manufacturers aren't AI-ready, it's not talking about a lack of interest or budget. Over half of manufacturing transformation spending in the next two years is allocated to AI and autonomous systems. The money is there. The ambition is there.

What's missing is the foundation. Specifically, data infrastructure.

Manufacturing data is scattered across systems that were never designed to talk to each other. You've got historian databases tracking production metrics. ERP platforms managing orders and inventory. SCADA systems monitoring equipment. Maintenance records in one system, quality data in another, shipping data in a third. Each one stores data in its own format, on its own schedule, with its own quirks.

The result: 73% of manufacturing data goes completely unused. That's according to industry research that puts the cost of data silos at $3.1 trillion annually across the sector.

AI needs clean, connected, real-time data to work. What most manufacturers have is fragmented, batch-processed, siloed data that's locked inside systems running software from the early 2000s. No amount of AI sophistication can overcome a data problem.

The Legacy System Trap

Here's where it gets painful for mid-sized manufacturers specifically. The systems running your operations are often decades old. They work. They're paid off. Your team knows how to use them. Ripping them out isn't realistic.

But those same systems are the biggest obstacle to AI integration. IBM's Institute for Business Value found that 53% of executives say difficulties integrating AI with legacy systems directly derailed their AI initiatives. Not slowed them down. Derailed them.

The issue isn't that old systems are bad. The issue is that they don't have modern APIs, they store data in proprietary formats, and they weren't built for the kind of continuous, real-time data access that AI requires. Your ERP might be excellent at what it does, but it probably can't feed a live data stream to an AI model that needs to make decisions every few seconds.

This is especially true for agentic AI, which is where the industry is heading. The TCS study found that 74% of manufacturers expect AI agents to manage 11-50% of routine production decisions by 2028. AI agents that can evaluate conditions and take action autonomously. But agents need continuous access to accurate, real-time data across systems. If your data is batch-processed overnight and trapped in silos, autonomous AI workflows are dead on arrival.

Where It's Working Already

The study isn't all bad news. Manufacturers who have invested in their data foundations are seeing results.

67% report improved real-time supply chain visibility from AI. Nearly 40% are already seeing measurable gains from embedding AI into quality control and production planning. Over 30% forecast meaningful productivity improvements from AI-led modernization.

In the work we do at Rotate, the pattern is consistent. The manufacturers getting value from AI aren't the ones who bought the fanciest AI platform. They're the ones who did the unglamorous work first: cleaning their data, connecting their systems, and starting with one use case that could prove value quickly.

I worked with a manufacturer last year doing roughly $8 million in revenue. They weren't trying to build a smart factory. They just wanted to stop spending 15 hours a week on production reporting. We connected their ERP data to an automated reporting pipeline. The reports now generate themselves every morning. Their production manager got 15 hours back. That opened the door to a second project, then a third.

That's how AI adoption actually works in manufacturing. Not a big bang transformation. A sequence of connected wins that build on each other.

How to Close the Readiness Gap

If you're in that 79% of manufacturers who aren't AI-ready, the path forward isn't buying an AI platform. It's fixing your data.

Start with an audit of where your critical data lives. Map your production data, quality data, inventory data, and order data across systems. Identify which systems can share data through APIs and which ones are locked. This exercise alone will show you exactly where the bottlenecks are.

Next, pick one use case with clear value. Production reporting, quality monitoring, inventory forecasting, or predictive maintenance on your most expensive equipment. Something where the data requirements are manageable and the ROI is obvious.

Then build the data pipeline first. Before you worry about AI models or machine learning, make sure the relevant data from two or three systems can flow into a single place in a clean, consistent format. This is the step most manufacturers skip, and it's why their AI pilots fail. You're not building AI. You're building the plumbing that makes AI possible.

Finally, prove value fast and expand. Get that first use case working, measure the results, and use those results to justify the next project. Each project should connect more data and enable more automation. Over 12-18 months, you'll go from siloed data and manual processes to a connected data layer that supports increasingly sophisticated AI.

The Competitive Clock Is Ticking

The TCS study makes one thing very clear: manufacturers who figure this out will have a structural advantage. When 75% of your peers expect AI to be a top margin driver but only 21% are ready, the gap between prepared and unprepared companies is going to widen quickly.

The manufacturers that invest in data infrastructure now will be deploying AI agents for production decisions while their competitors are still trying to get their ERP to talk to their quality system. That's not a theoretical advantage. That's faster response times, lower scrap rates, better inventory turns, and tighter margins.

The 79% who aren't ready today don't need to stay there. But closing the gap requires starting with the foundation, not the flashy stuff. Fix your data. Connect your systems. Prove value with one use case. Then build from there.