Everyone is talking about how AI is disrupting the jobs of knowledge workers, but much less is said about front line workers who represent 70-80% of the global workforce (Boston Consulting Group data). In fact, one of the most significant workplace shifts is unfolding on the factory floors, in the warehouses and logistics centers where 2.7 billion deskless workers keep critical industries like manufacturing, automotive and logistics running.
But as manufacturers are investing heavily in digital technologies like robotics, digital twins, AI and edge computing, the pool of highly skilled talent is shrinking. By 2030, over 60% of employers globally expect disruption due to skills gaps, with manufacturing, construction, and the energy sectors being among the most impacted (WEF Future of Jobs Report 2025).
Barriers to organizational transformation, 2025-2030
Evolution in the share of workers’ core skills expected to change and to remain the same within the next five years, 2016-2025
These skills shortages will be further exacerbated by the aging population, with 87% of organizations anticipating significant labor shortages as their workforce ages (Capgemini data). As these experienced engineers, mechanics and subject matter experts move to another company or retire, their key knowledge and insights, coming from years on the job, risk being lost. Keeping this know-how within the business is not only vital for preserving organizations’ intellectual property, but also for training the future workforce.
So how can manufacturers preserve this knowledge and close the skills gap?
What manufacturers are trying
As AI reshapes manufacturing, consumer grade Large Language Models (LLMs) are emerging as a conversational gateway between humans and machines, offering the promise of improved productivity and better knowledge sharing. In response, many manufacturers are experimenting with general purpose AI chatbots like Google Gemini, Microsoft Copilot or ChatGPT for technical support and upskilling front line workers. But as these implementations move from testing to production, they quickly discover the limitations of these mainstream GenAI tools when applied to manufacturing, field support and knowledge management. Here is why.
| Industrial Requirements | Generic AI Capabilities |
|---|---|
| Industrial support depends on finding the right document, diagram or photo quickly, on the job, with clear provenance. | Even with storage like Sharepoint, accuracy on a hundred documents severely drops when searching for product names, diagrams or in a different language than the source document. |
| Knowledge must be consistent, approved, and current so every site follows the same trusted guidance, shift after shift. | Mixes drafts, duplicates, versions, and public information such as YouTube videos, out dated blog, anonymous forums and competitors documentation, all together without any control. |
| Contextual knowledge from operations (client records, past interventions, part stocks, etc.) greatly increase answers quality. | Totally disconnected from daily operations, and likely to infer from online sources or fabricate information through the conversation in an attempt to help the worker. |
| Users need traceable sources and clear limits so they can verify before acting, especially for safety tasks. | Can provide confident text with fabricated facts, despite sources, pushing the burden of control and validation to the worker at the expense of the client. |
| Organizations need clear access rules, audit trails, and defensible handling of sensitive know-how and IP. | Some offer paid tiers with access to conversations, but visibility and auditability is incomplete, all the while data still resides on US soil and legislation. |
| A knowledge system must capture lessons learned and keep guidance updated as work orders and issues occur. | Cannot turn daily work or consolidate captured knowledge into maintained guidance; Structuring and updating the knowledge base remains a continuous manual effort. |
A better way to share specialist knowledge
The fundamental issue isn’t that consumer AI chatbots lack certain features, it’s that they’re solving a different problem entirely. They’re designed to make vast amounts of public information conversationally accessible. Manufacturing needs something else: a way to preserve, structure, and share the specialized expertise that exists within the organization itself.
Real knowledge management for industrial environments means treating organizational expertise as a living asset that requires continuous cultivation. It starts where your knowledge actually lives, not just in technical manuals, but in your operational systems, in how experienced engineers solve recurring problems, in the decisions captured in quality investigations and deviation reports. This knowledge must be continuously captured, validated by domain experts, and made accessible exactly where work happens, whether that’s at a machine station, in a maintenance workflow, or during a shift handover.
The difference between a general-purpose AI tool and an AI knowledge management system isn’t measured in the number of parameters, it’s measured in how deeply it integrates with your operations, how rigorously it maintains accuracy, and how it simplifies the process of capturing, structuring, sharing and improving your operational knowledge. When knowledge systems are purpose-built for industrial environments, they become part of your operational infrastructure, running within your security boundaries and evolving with your processes. They don’t just answer questions, they help identify where knowledge gaps exist, ensure critical expertise isn’t locked in one worker, and empower your entire workforce with expert-level guidance by simplifying the entire knowledge management process.
For manufacturers facing skills shortages and an aging workforce, the question isn’t which AI tool to adopt. It’s whether you’re treating knowledge management as a strategic capability that requires the same rigor you apply to your production systems, quality controls, and safety protocols.
