The Real Barrier to Scaling AI Is Data
/
From Manage Engine
Two new reports offer a detailed look at how organizations are adopting AI and where the economic impact is showing up first.
Building the foundations for agentic AI at scale from McKinsey
What 81,000 people told us about the economics of AI by Anthropic (makers of Claude)
Much of the data reflects knowledge work: software development, marketing, and office-based roles. But beneath the surface, the findings carry implications for industrial organizations managing plants, assets, and infrastructure.
Want to join the 10% club that successfully scales Industrial AI? Then be sure to grab your seat to be in the room with the hundreds of Industrial Intelligence leaders this June at Innovate Energy Week in Houston.
The Real Barrier to Scaling AI Is Data
One of the clearest findings from the McKinsey research is that most organizations are already experimenting with AI, but very few have successfully scaled it.
Nearly two-thirds of enterprises have tested agent-based AI systems, yet fewer than 10 percent have deployed them in ways that deliver tangible operational value. Why so few? Eight in ten companies reported that data limitations are preventing them from scaling AI initiatives.
For industrial organizations, this finding might feel familiar. Many already operate in environments with fragmented systems—maintenance platforms, SCADA systems, asset databases, and production logs that don’t always share context.
Agentic AI systems depend on:
Reliable, interoperable data
Clear governance and access controls
Continuous data quality monitoring
Without a solid foundation, automation attempts break down quickly. A single agent acting on incomplete or inconsistent data can create operational risk instead of efficiency.
For industrial leaders, the takeaway is straightforward: AI readiness is increasingly a data architecture problem.
Shoutout to Pieter van Schalkwyk who wrote on LinkedIn about how the McKinsey report underscores the importance of a strong data foundation ahead of AI adoption.
Start Small: Focus on High-Impact Workflows First
The McKinsey report posited that building value from agentic AI typically starts by identifying a small number of high-value workflows and automating them deliberately. We all know and love pilot programs because they’re common in industrial applications thanks to the way they help validate impact before scaling across an enterprise.
In industrial environments, workflows that could be prioritized for automation include:
Maintenance planning and work order management
Spare parts and inventory coordination
Inspection and compliance documentation
Production scheduling and optimization
Targeted pilots in high-impact areas allow organizations to build confidence while strengthening the data and governance systems required for broader deployment.
Governance Is Second to None
Both reports converge on a single theme that will shape the next phase of AI adoption: Governance is becoming as important as technology.
As AI systems gain autonomy, organizations must define:
What actions agents can take
What data they can access
When human approval is required
Scaling AI safely requires clear policies, automated oversight, and traceability across workflows.
In industrial environments, where safety and compliance are non-negotiable, this requirement becomes even more critical. Agent-based systems must operate within established operational controls, just like any other equipment or software system.
The Industrial Adoption Curve May Be Slower (But Bigger)
These two reports mostly center on office workers, not industrial techs out in the field. The Anthropic survey, for example, primarily reflects users working on tasks like coding, writing, and analysis. AI adoption will look different in offices vs. industrial environments.
In an office, an individual worker can open Claude and use it to analyze a report, generate marketing ideas, or automate follow-up emails.
Industrial operations, on the other hand, rely on proprietary processes, specialized equipment, and domain-specific data, which naturally slows adoption. A maintenance tech generally doesn’t have the same opportunity (or need) to outsource work to AI agents. What they do is tied to a larger data system, and AI automation likely has to happen at the team level, rather than the individual worker level.
But slower individual adoption doesn’t mean smaller impact. When AI reaches industrial workflows at scale, the results can be more transformative than anything an office worker could ever implement. We’ve been talking for years about AI’s role in industrial automation, inspection, and maintenance, and these reports confirm what we’ve seen in the field.
Workforce Concerns Are Real
The Anthropic survey highlights an important workforce dynamic that industrial leaders should not ignore, even though the report reflects personal Claude accounts.
The stat: Roughly one in five respondents expressed concern about job displacement due to AI.
This perception of risk is not evenly distributed. Early-career workers are significantly more likely to worry about displacement than senior professionals, which has direct implications for industrial organizations already facing skills shortages.
Younger engineers and technicians often:
Work more directly with digital tools
Have less institutional knowledge
Are more exposed to automation-driven workflows
Meanwhile, experienced operators and domain experts remain critical because, as Jonathan Alexander pointed out on LinkedIn, “Your 30-year operators will carry keys that public models do not have.”
Leadership teams can use this insight to shift workforce strategy. Avoid “replace people with AI” initiatives. Instead, focus on combining domain expertise with data and automation.
Jonathan Alexander will be a keynote speaker at the Industrial AI Nexus Summit, part of InnovateEnergy Week in Houston in June.
AI Is Expanding What Workers Can Do
The Anthropic study provides a useful lens into how workers actually experience AI in practice. One finding was that productivity gains are more likely to come from new capabilities than from simple speed improvements.
According to the report, 48% of users described productivity gains in terms of expanded scope—doing tasks they previously could not do—while 40% emphasized faster completion of existing tasks.
That distinction between scope and speed matters in industrial environments.
Historically, automation has focused on efficiency: completing the same work faster or with fewer errors. But AI introduces a different kind of value by enabling work that was previously impractical.
Examples in industrial settings include:
Generating predictive maintenance insights from sensor data
Automating root cause analysis after equipment failures
Creating inspection summaries from video or image data
Coordinating multi-system workflows without manual intervention
In other words, the biggest gains may not come from doing the same work faster. They may come from doing entirely new work that isn’t feasible with human brain power alone.
What Industrial Technology Leaders Should Do Now
The combined lessons from these reports point to a practical roadmap for industrial organizations preparing for the next phase of AI adoption.
Focus on foundations before scaling - Data quality, integration, and governance determine whether AI initiatives succeed.
Start with a small number of high-value workflows - Pilot projects create measurable outcomes and reduce risk.
Invest in workforce alignment - Experienced operators and engineers remain essential to successful deployment.
Build governance early - Operational trust depends on clear accountability and oversight.
These steps are necessary to build the operational infrastructure required to make AI reliable, safe, and valuable in real-world environments.
The Bottom Line
AI adoption is accelerating across industries, but the organizations generating real value are not the ones deploying the most models. They are the ones building the strongest foundations.
For industrial technology leaders, the path forward is:
Start with data
Focus on workflows
Strengthen governance
Scale deliberately
And when the technology is ready, the organization will be too.
Read the full AI adoption reports:
Building the foundations for agentic AI at scale from McKinsey
What 81,000 people told us about the economics of AI by Anthropic (makers of Claude)
Want to join the 10% club that successfully scales Industrial AI? Then be sure to grab your seat to be in the room with the hundreds of Industrial Intelligence leaders this June at Innovate Energy Week in Houston.
