Since 1999, Bill Rokos has spearheaded the development of Parsec’s manufacturing operations management (MOM) platform, TrakSYS.
AI-based automation is becoming an expectation across industries. Even manufacturing—which has traditionally been slower to adopt emerging technologies—has been relatively quick to invest in and explore the tool’s potential. A study by my company found that AI adoption in the sector has skyrocketed over the past two years.
Nearly three-quarters (72%) of industry leaders now say their organizations use the technology in their operations, a significant boost from the 53% who said the same in 2024. This is undeniable progress. It’s the mark of an industry ready to move into a new, more precise and data-driven era of operations.
There’s a catch, though. Few respondents have fully extended AI tools across departments, and only 10% are using AI-/ML-enabled automation at scale. So, what’s holding up the rest?
The Obstacle Of Ownership
The answer is as psychological as it is operational, and we’ve been largely ignoring the former with attention so focused on the latter. As AI has worked its way into workplaces and production lines, much of the discussion has centered on infrastructure, data, tooling and computational potential. All fair, but it’s left an equally critical factor out of the conversation: the people, of course.
I hear you, and I’m aware—plenty of leaders have opined on the impacts of AI on those who interact with it. We’ve discussed the labor market implications, effects on work quality and worker skills, and new models emerging in the age of augmented productivity. I’m not talking about any of that.
The issue underlying this gap at this moment is both more complex and more basic than the traditional sticking points outlined above. When autonomous AIs are in the mix, we all struggle to answer what have always been relatively simple questions: What if it makes a mistake? Who gets the blame? Who handles the fallout?
The same questions apply to human workers, who themselves can make mistakes. But when AI enters the mix, things get more complicated. Blaming the system or agent only gets you so far, and, to many, it doesn’t feel like enough. Leaders must consider whether that blame extends to the operator who approved the output, the vendor, the CTO who signed off—the list goes on.
Most leaders (and tech vendors, for that matter) haven’t figured it out yet. Even those who have decided where the fault lies aren’t done. What happens next? How critical must a failure be before you consider a full shutdown? A changed process or new platform? Who pays for the damages?
Knowing the answers to these questions is especially pressing in manufacturing, where mistakes aren’t confined to dashboards and analytics. In factories and warehouses, the stakes are high, immediate and physical. Even small mistakes can affect worker safety, shut down lines or ruin materials. The lingering anxiety around it all is getting in the way of full-scale, enthusiastic adoption.
Under The Surface
Though few respondents in our survey name the phenomenon directly, anxiety about the ambiguity of accountability permeates the findings. Risk perception spreads near evenly between being too hesitant (60%) and too aggressive (40%). The leader-worker enthusiasm dichotomy that marked early adoption has flipped, with leaders now less enthusiastic about AI than their staff.
Additionally, analytics applications remain more common than automation, with deployments touching production and operational control missing from top use cases. Governance, cost and integration have replaced infrastructural limitations as leaders’ top barriers.
These findings paint a picture of leaders who, unsure where post-AI accountability sits, have backed away from transformation in areas with arguably the most potential. It makes perfect sense when considered in the context of today’s most available and visible tools.
Generative AI (GenAI)—the most commonly deployed technology in the industry—is inherently probabilistic, which can make it a poor fit for tasks that require strictly deterministic behavior. The most commonly available GenAI tools are favored because of this variability, which is what makes them so valuable in some contexts and so risky in others.
The challenge becomes greater when AI systems are deployed in ways that make their recommendations difficult to interrogate or explain. Operators may then be asked to approve outputs without fully understanding how the system arrived at them, creating distance between the decision, the person responsible for authorizing it and the eventual outcome.
This complicates things even more because existing governance and accountability frameworks break down when ownership cannot be easily attributed. A daunting diagnosis of a complex issue, to be sure—but naming it is the first step toward unraveling and overcoming it.
Supporting Scale
The key to unlocking automation at scale is reviewing and revising governance to cover three pillars that apply to every technology investment.
1. Explainability As The Standard
Incorporate a mandate that frames explanation as a requirement for all technology investments. Organizations that treat understanding as a universal standard can eliminate this enduring trust gap.
2. Chain Of Ownership
Decide who is ultimately responsible for the decisions entrusted to automated systems ahead of time, just as you have for purely human decisions in the past. Identify the levels of potential failure and how each person in the loop connects to them. Critically, leaders must communicate these chains directly and openly, so everyone feels comfortable about their role in AI-augmented workflows.
3. Contingency Planning
This is related to ownership, but distinct in its purpose. Once you know what could go wrong and who (or what) owns that misstep, you can plan your responses—both disciplinary and operational. This allows leaders to move with genuine confidence rather than suppressed anxiety.
Conclusion
This work might feel like stalling, especially when compared with the investments in data pipelines, infrastructure and legacy-system migrations that have defined much of the AI journey so far.
But this is maturity work, not a detour. Organizations cannot scale automation confidently if they have not decided who owns the decisions it makes, how those decisions can be challenged and what happens when they go wrong. The 72% adoption figure shows that the industry is ready to move forward. Building the accountability framework to match that adoption is what will determine how far it can go.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
