In industrial manufacturing, AI stalls between systems, not on ambition.
11-17-2026
A generative model that drafts a distributor campaign, a spec sheet, or a service renewal notice operates in engineering-heavy, documentation-sensitive, safety- and compliance-bound work. When an error in that work carries regulatory, reputational, or operational risk, slowing down is a reasonable choice, not a failure of nerve. The question worth asking is where reasonable caution hardens into a stall, and what the organizations that get past it do differently.
Two sources give industrial manufacturing a clear read. The first is the Adobe and Incisiv study of AI operating-model maturity, covering 5,633 executives across 12 industries and 11 markets, with 467 active respondents in this sector. The second is a set of first-person conversations with senior leaders across European industrial manufacturers, conducted through Adobe's EMEA listening tour. The data shows where progress is concentrated. The conversations explain why. Read together, they point to the same conclusion: the barrier is the operating model, not the technology.
Activity crowds into the middle.
58% of respondents sit in the piloting or deploying stages, a concentration the research calls the Frozen Middle. Within that group, 68% convert fewer than one in four pilots into production. AI runs inside individual functions. Account intelligence informs one sales team. A distributor content workflow serves one channel. An aftermarket model serves one service line. None of them become repeatable across the commercial model.
The reason is structural, not a lack of intent. Scaling a use case pulls channel, service, data, risk, and technology teams together across boundaries that were built for long capital cycles, not rapid iteration. Global operations and multi-tier channels add coordination demands that no single function can resolve. The activity is real, but the connective work that would make it repeatable has not been done.
Authority stops at every function's edge.
Only 13% of the active base reports a dedicated AI owner with a formal title and budget authority. Everywhere else, accountability is part-time, committee-led, assigned project by project, or absent. This matters because commercial AI in this sector rarely stays within one function. An account program may need installed-base telemetry, distributor CRM access, and technical content at the same time. When authority ends at each boundary, decisions about data rights, integration, and priority pass through several owners, and no one can make a binding call.
The cost shows up at production. Across the active base, a median of 13 per 100 proposed use cases reach production, below the global average of 15. The largest drop comes at the initial IT review, then at security, integration, and budget reviews. Each gate is legitimate, but a use case spanning CRM, distributor portals, and manufacturing execution systems often stalls on ownership before anyone can test whether it would work commercially.
The integration debt behind the handoff.
86% of the active base describes the business-IT relationship as siloed, transactional, or consultative. The business defines the request, then hands it to IT to build. That cadence fits ERP and control-system upgrades measured in years. It does not fit AI-driven commercial work, which needs faster iteration and shared accountability.
The conversations add texture to the number. Leaders describe marketing and sales running on systems built decades apart, sharing data asynchronously at best. Only 1% of manufacturers report fully integrated, enterprise-accessible customer data, yet 82% of marketing leaders are now expected to contribute directly to revenue. Revenue accountability is mandated. The data to earn it is structurally out of reach.
Content sits on the same fault line. 48% of marketing organizations are still built around product or service lines, and none are organized around the customer journey. Product-line structures produce product-line content, which is poorly suited to the multi-role buying groups these firms actually sell to. Layer AI generation on top of that, and you accelerate fragmentation rather than fix it.
What the Leaders built instead.
The research identifies 342 Leaders across the full population, scored on operating-model maturity and deployment stage. They are a cross-industry cohort, but their behavior maps directly onto this sector's problem. Five patterns stand out.
Real ownership: Leaders are 7.2 times more likely to have a formal AI owner with the title, mandate, and budget to work across functions.
Day-one measurement: 44% of Leaders measure return comprehensively, against 7% of the active industrial manufacturing base.
Genuine business-IT partnership: 81% operate at a collaborative or integrated level, compared with 14% in this sector.
Learning culture: When AI underperforms, 44% of the active base reports blame or repeated failure. Among Leaders, the figure is under 1%.
Empowered champions: 42% of Leaders name finding and backing internal practitioners as the single people-or-culture move with the greatest effect on scaling AI, ahead of executive sponsorship at 16% and formal training at 11%.
Read together, these are not five projects. They describe one owner who spans demand generation, channel, and aftermarket, sets the outcome and baseline before a deployment starts, and brings commercial, data, technology, security, and compliance teams into the brief while integration constraints can still be mapped.
Governance is the table stakes. Structure is the differentiator.
Here, the two sources create a useful tension. In the conversations, the most sophisticated operators are investing heavily in governance, building the frameworks, IP indemnification terms, approval workflows, and audit trails that safe deployment requires. They are not hesitating out of doubt. Many follow a deliberate sequence of process first, then automation, then AI, which a surface-level survey can misread as an absence of plans. In this study, 42% report no current GenAI plans in content operations. That figure most likely describes sequencing, not disengagement.
But the operating-model analysis shows governance is where Leaders separate least from the field. Structure and workflow redesign show the widest gaps. The organizations most likely to scale are the ones redesigning how cross-functional teams operate and how work flows, not only how it is controlled.
The implication is direct and governance is necessary, but on its own, it is not enough. An organization that invests in governance without restructuring builds a well-governed version of the same stall.
Where this leaves executives.
The most common strategic regret in this data is not naming a dedicated AI owner earlier. That is worth sitting with, because it is a structural decision, not a technology one, and it is the decision peers say they would take back.
The practical next step is narrow. Take one priority commercial use case. Name the owner. Define the outcome and the baseline. Map the data, architecture, and governance dependencies before deployment, not after. That discipline will not remove regulatory exposure or consolidate a fragmented estate. However, it gives the organization a clearer basis for deciding what to scale, what to change, and where further investment is justified.
The tools are already in the building. What decides the outcome is whether the operating model can put them to work.
Explore the report to learn what the organizations that scale AI do differently.
Recommended for you
https://business.adobe.com/fragments/resources/cards/thank-you-collections/manufacturing