Most mid-sized manufacturers are stuck in AI pilot purgatory, Kaufman Rossin research finds

Not one mid-sized manufacturer surveyed by Kaufman Rossin has fully operationalized AI across its business. Every company studied remains somewhere in the experimentation spectrum, according to research conducted among senior decision-makers at mid-sized U.S. companies and cited by Industry Valley. For operations and IT leaders who have been approving AI pilot budgets for the past two years, that finding is a useful reality check.

The data gap nobody fixed

The central constraint is not the AI software itself. It is the data infrastructure underneath it. Kaufman Rossin's research found that only 27% of manufacturing companies surveyed maintain a data warehouse or data lake. Across the broader mid-market, that figure is 60%, meaning manufacturers are running at less than half the data-readiness rate of comparable non-industrial companies.

Roughly 45% of manufacturers still operate with siloed data, and none in the study use machine learning platforms. Even looking at the full mid-market sample, only 16% have reached a fully managed and integrated data state. These are the organizations actually capable of feeding AI systems consistently reliable inputs.

ERP is the wall

Every manufacturer in the Kaufman Rossin study uses an ERP system. That ubiquity is precisely the problem. Legacy ERP platforms were not designed to connect natively with modern AI tooling, and 55% of manufacturers in the survey name that integration gap as their single biggest AI obstacle. The mid-market average for the same concern sits at 41%, which means manufacturers feel this friction significantly more than their non-industrial peers.

The challenge compounds because many of these ERP environments have years of customizations layered on top of them. Retrofitting API connections or data pipelines onto those systems is expensive, slow, and often requires vendor cooperation that is not guaranteed.

73% stuck in testing, 0% fully deployed

The Kaufman Rossin data puts a hard number on what many plant-floor technology teams already sense: 73% of manufacturing companies remain in early or basic AI preparation stages. Across the full mid-market sample, the same 73% figure applies. Only 7% of mid-market companies overall are positioned for company-wide AI scaling.

The wins that do exist tend to be narrow: time savings on a specific task, automation of a single accounting function, or individual productivity gains inside a process that is still broken at either end. The research flags the specific risk here, that a successful pilot can be mistaken for a completed journey. Budgets get reallocated, internal champions declare victory, and the underlying data problems remain.

Intent is high, readiness is not

Despite the scale gap, appetite for AI investment is nearly universal. Every manufacturer surveyed agreed that AI saves time, and 91% plan to increase their AI spending. That combination, strong intent paired with weak infrastructure, is precisely what creates compounding risk. More investment in AI tools layered onto fragmented data produces more pilots, not more production deployments.

The cultural dimension compounds the technical one. Industrial companies have historically built their competitive edge on operational expertise and domain knowledge rather than data-driven decision systems. Shifting that posture is harder than installing any platform, and the Kaufman Rossin research identifies it as a distinct barrier alongside the technical ones.

What your team should act on now

• Audit your data estate before your next AI purchase decision: identify where your most operationally critical data lives, how clean it is, and whether your ERP can export it in a usable format.
• Prioritize use cases where data is already structured and accessible rather than forcing AI onto fragmented or inconsistent inputs. Wins from clean data build the internal case for broader infrastructure investment.
• Evaluate your ERP vendor's AI integration roadmap explicitly. With 55% of manufacturers naming legacy system integration as the top barrier, this should be a standard line item in any renewal or upgrade conversation.
• Frame the internal AI initiative as a data infrastructure program first. Leadership alignment on data as a strategic asset is what separates organizations that scale from those that accumulate pilots.