Bridging the Skills Gap: How Industrial Leaders Are Leveraging Agentic AI to Capture Institutional Knowledge and Modernize Operations

Asset-intensive sectors—ranging from global manufacturing and civil aviation to heavy energy and critical utility networks—stand at a profound operational crossroads. For decades, these industries have functioned as the backbone of the modern economy, relying heavily on the tacit, undocumented knowledge of seasoned technicians and engineers to maintain complex machinery and critical infrastructure. Today, however, an accelerating demographic shift threatens to sever this vital continuity. As veteran workers approach retirement en masse, organizations face the imminent risk of losing decades of institutional wisdom that has never been formally codified into enterprise systems.

This looming expertise drain is no longer a theoretical concern for human resources departments; it has evolved into a frontline economic and operational crisis. According to a landmark study by the U.S. Census Bureau, nearly one-fourth of the entire manufacturing workforce was aged 55 or older, setting the stage for a massive wave of retirements. Simultaneously, pipeline reports from the Aviation Technician Education Council, drawing on projections from Oliver Wyman, anticipate severe talent deficits, including a projected shortfall of approximately 9,100 certified aviation mechanics in North America alone. The U.S. Department of Energy has repeatedly warned that this generational handover directly threatens productivity, safety protocols, and long-term operational continuity across critical national infrastructure.

The financial toll of these maintenance and workforce vulnerabilities is staggering. Research conducted by the National Institute of Standards and Technology (NIST) estimates that U.S. manufacturers face roughly $119.1 billion annually in preventable losses, with equipment downtime alone accounting for an estimated $18.1 billion of that total. When combined with a dwindling pool of experienced maintenance personnel, industrial leaders are under unprecedented pressure to modernize their operational frameworks without disrupting daily workflows.

To explore how advanced technology can mitigate these systemic risks, Eysan Yazici recently hosted Vaibhav Kumar, Global Chief Technology Officer at IFS, on the Emerj AI in Business Podcast. Their discussion, centered on the theme "From Operational Insight to Agentic Industrial AI," examined how modern artificial intelligence can preserve institutional knowledge, democratize data access, and fundamentally transform asset management. Drawing on his extensive enterprise architecture background—which includes serving as Chief Architect for Shell’s Renewables & Energy Solutions organization and founding the Shell Agile Hub—Kumar outlined four critical strategies for industrial executives seeking to deploy AI with measurable operational impact.

The Anatomy of the Industrial Knowledge Crisis

The modern industrial landscape is currently defined by a unique convergence of macroeconomic and demographic pressures. According to Vaibhav Kumar, asset-intensive industries are simultaneously battling an aging workforce, a scarcity of incoming skilled technical talent, and mounting operational complexity driven by automation and advanced machinery.

This workforce shortage has immediate, highly visible consequences. In the commercial aviation sector, major manufacturers like Boeing have previously highlighted that severe skills shortages risk keeping grounded aircraft out of service simply because certified technicians are not available to perform required maintenance checks. At the enterprise level, Kumar notes that the absence of skilled personnel can bring complex manufacturing or energy production facilities to a grinding halt, resulting in financial losses that easily run into the millions of dollars per incident.

Paradoxically, the root cause of lost institutional knowledge is rarely an unwillingness among veteran workers to share their insights. Instead, the primary barrier has historically been the design of enterprise software itself. Legacy systems and traditional enterprise resource planning (ERP) platforms are notoriously cumbersome, creating a high-friction user experience. For a respected, veteran technician whose primary value lies in physical asset maintenance, navigating complex database screens and text fields is an unwelcome distraction.

"The problem that you’ve had with software for the longest time, enterprise software, is it’s just super inconvenient to go in and get data into the system," Kumar explains. "It’s a point of friction. Using that software is a point of friction for your most experienced worker who’s well respected by everybody else. So they don’t."

Consequently, the valuable diagnostics, improvised troubleshooting techniques, and preventive maintenance nuances generated by frontline workers remain locked in paper logbooks, digital spreadsheets, or—most frequently—the memories of retiring employees. Even when data is entered into legacy systems, retrieving and interpreting it requires specialized database queries that non-technical operations managers or incoming junior workers struggle to execute.

To reverse this trend, Kumar advocates for the deployment of consumer-grade, frictionless interfaces that make recording knowledge as simple as speaking or typing a natural language note. When organizations streamline both knowledge capture and retrieval, they ignite a powerful operational flywheel: frictionless contribution populates the system with vital data, while highly relevant, AI-generated answers give workers a compelling reason to continue contributing. Industrial leaders are advised to conduct a rigorous two-part audit of their workflows to identify where knowledge stalls, targeting specific operational bottlenecks for immediate AI intervention.

Democratizing Operational Insight Through Conversational AI

Beyond capturing fading expertise, industrial enterprises must radically improve how frontline workers access existing operational data. Traditional enterprise software architectures were built for accountants and data entry specialists, not for field technicians standing on a factory floor or plant managers monitoring live energy grids.

Kumar emphasizes that frontline and executive personnel alike require purpose-built, role-specific experiences that eliminate operational friction. The most promising pathway to achieving this is the integration of conversational AI layers—similar to consumer tools like Claude or ChatGPT—directly over core enterprise business data. Because modern workers are already familiar with conversational interfaces in their personal lives, adopting a similarly intuitive experience in the workplace dramatically accelerates user adoption.

To illustrate the profound efficiency gains of conversational enterprise AI, Kumar contrasts a traditional executive workflow with an AI-enabled one. In a conventional corporate setting, when a Chief Financial Officer (CFO) requires a specific operational metric or maintenance cost breakdown, the request typically triggers a cumbersome chain of intermediaries:

1. The CFO contacts a vice president of operations.
2. The VP tasks a regional plant manager with pulling the data.
3. The plant manager assigns a database administrator to run specialized SQL queries.
4. The resulting report travels back up the chain of command, a process that can take days.

In contrast, deploying a conversational AI layer directly over the same underlying enterprise databases allows the CFO to query the system directly in natural language and receive a comprehensive, verified answer within sixty seconds. Kumar identifies this democratization of insight as the primary productivity driver of enterprise AI.

Crucially, Kumar is quick to clarify that this technological leap does not eliminate the need for human expertise. Drawing from his own software development and presentation-building workflows, he notes that tasks that once required days of drafting, storyboarding, and review meetings can now be initiated through detailed prompts in a matter of minutes. However, human oversight remains absolute. The human defines the intent, directs the agent, and verifies the final output, while the AI compresses the temporal gap between initial concept and final execution.

This philosophy directly governs how industrial organizations should approach agent autonomy. Kumar champions a human-in-the-loop paradigm where background automation is strictly opt-in: "It’s the human that always drives the agent. It’s only by a human’s choice do you decide, well, actually, I want agents to just work in the background for me. And when you do decide that, you’re then given the option and the ability to say, okay, when do you bring me in the loop?"

Prioritizing Deployments Based on Data Gravity

As industrial organizations race to adopt artificial intelligence, a central strategic challenge is determining where to deploy these tools first to maximize return on investment. According to Kumar, enterprise AI’s success is fundamentally tethered to existing systems containing high-value operational data. Without a robust data foundation, artificial intelligence lacks the raw material required to retrieve meaningful insights or execute automated tasks.

However, leadership teams often stall digital transformation initiatives out of a misplaced obsession with data perfection. Many executives assume that operational data must be 100% clean, standardized, and validated before any AI model can be applied. Kumar explicitly challenges this assumption, noting that modern large language models are uniquely suited to identify duplicate records, flag anomalies, and assist in data cleansing. Deploying AI agents directly over legacy databases can actually accelerate data hygiene rather than depend on it as a prerequisite.

Instead of waiting for pristine datasets, Kumar advises industrial leaders to follow the principle of data gravity—prioritizing the enterprise systems that hold the organization’s most critical and frequently accessed operational records, regardless of whether that data is currently fragmented or incomplete.

"Incomplete or not, putting AI over the top of it is useful," Kumar states. "The way I think about it is: data has gravity. So where is the gravitational pull? Which systems have the most meaningful data in your company? Think meaningfully about how you can put AI over the top of that, and you’ll start to see good things happen."

This discernment is equally vital when evaluating software vendors. Kumar cautions against market hype, pointing out that some technology providers merely repackage existing integration Application Programming Interfaces (APIs) as advanced servers built on open connectivity standards like Model Context Protocol, subsequently marketing them as proprietary AI innovations. For industrial buyers, the litmus test should always be whether a solution delivers a seamless, consumer-grade operational experience rather than whether it carries trendy technological labels.

Cultivating Organization-Wide AI Literacy

The final and perhaps most challenging hurdle in industrial AI adoption is human readiness. Kumar stresses that building comprehensive AI literacy across the entire workforce must be an immediate priority for executive leadership.

Public discourse surrounding artificial intelligence often oscillates between exaggerated utopian claims and fatalistic skepticism, leading many observers to prematurely conclude that the technology is overhyped or incapable of handling industrial-grade workloads. Kumar argues that this skepticism stems from a fundamental misunderstanding of the technology stack. While much of the foundational AI model development occurred rapidly in recent years, the creation of tangible business value depends on the successful maturation of the application and enterprise software layer—a structural integration that is actively taking place right now.

"Just because you don’t see that happen today, or not enough of [it], doesn’t mean AI is not capable," Kumar asserts. "The best thing leaders can do right now is to prepare their people to be capable and literate in understanding AI. And that doesn’t just mean your engineers and your doers. That means everybody, top down."

Drawing parallels to the early days of enterprise cloud adoption, Kumar recalls that cloud computing was viewed as an ambiguous buzzword around 2010, yet by 2015 to 2018, corporate boards and technical teams spoke fluently about hyperscalers, multi-cloud architectures, and regional redundancy. Enterprise AI is currently navigating that exact transitional phase. Developing baseline literacy is the indispensable first step toward identifying strategic deployment vectors.

Furthermore, Kumar highlights a persistent cultural disconnect within many industrial enterprises. While corporate boards across manufacturing, energy, and utilities are aggressively pressing chief executive officers for comprehensive enterprise-wide AI strategies, middle managers frequently exhibit hesitancy or resistance. This reluctance often stems from job security concerns or deep-seated skepticism born from past failed technology rollouts.

To overcome this friction, leadership must actively dismantle the overstated narrative that AI will simply eliminate human jobs. Instead, organizations must champion a human-centric narrative: AI tools do not replace skilled workers; rather, they eliminate administrative friction and amplify human productivity to unprecedented levels.

Looking ahead, the cost of delay is fundamentally competitive. As embedded artificial intelligence rapidly transitions from an experimental differentiator to standard industry infrastructure, organizations that systematically capture institutional knowledge, democratize operational data, and foster workforce literacy will secure a decisive advantage. For industrial leaders navigating the retirement wave, the mandate is clear: treating AI literacy and data integration as an enterprise-wide imperative is no longer optional—it is the baseline requirement for operational survival in the twenty-first century.