The global corporate landscape is currently witnessing a massive divergence between the capital invested in artificial intelligence and the tangible returns those investments generate. While enterprises have spent billions of dollars acquiring Large Language Models (LLMs), implementing generative AI interfaces, and training workforces on prompt engineering, the anticipated revolution in productivity remains elusive for the vast majority of organizations. According to recent data from Gartner, approximately 88% of HR leaders reported in late 2024 and early 2025 that their organizations have yet to realize significant business value from AI tools. This sentiment is echoed by a comprehensive Boston Consulting Group (BCG) study of 1,250 global companies, which revealed that only 5% are capturing meaningful value at an enterprise scale, while a staggering 60% have seen no material return on investment whatsoever.
The root of this failure is rarely found in the technology itself. Instead, the bottleneck lies in a fundamental misunderstanding of how human expertise functions within a corporate environment. AI is designed to scale knowledge that has been made explicit—information that is written down, codified, and structured. However, the most valuable assets within a company—the decision-making frameworks and intuitive problem-solving skills of top performers—exist as tacit knowledge. This "hidden expertise" is almost never documented, meaning that when a company "feeds" its data to an AI, it is providing the model with a skeleton of formal processes rather than the muscular reality of how work actually gets done.
The Rise of the Enablement Illusion
As companies rush to modernize, many are falling victim to what Swagatam Basu, Senior Director Analyst in the Gartner HR practice, describes as the "enablement illusion." This phenomenon occurs when organizational leaders mistake tool access and basic adoption metrics for genuine digital transformation. In this scenario, a company might celebrate a 90% login rate for a new AI platform or high attendance at training seminars, only to find months later that employee behavior hasn’t changed and productivity remains stagnant.
The enablement illusion is often born from the assumption that understanding a new system is the same as applying it. Historical precedents in corporate management highlight this disconnect. For decades, executives have attempted to migrate workforces from fragmented systems—such as personal Excel spreadsheets—to centralized CRM or ERP platforms. Despite intensive training and executive mandates, employees frequently return to "the old way" because the official system fails to account for the nuances of their daily reality. When an employee says a tool is "just easier" or "better for my workflow," they are often signaling that the official process lacks the flexibility or the specific contextual knowledge required to handle real-world exceptions.
In the context of AI, the enablement illusion is even more dangerous. If an AI tool is trained only on "idealized" process manuals, it will generate recommendations that look correct on paper but are useless in practice. When employees realize the AI doesn’t understand the "unwritten rules" of their industry or company, they either spend excessive time manually correcting the output or abandon the tool entirely, leading to the massive ROI gap currently observed in the market.
The Tacit Knowledge Paradox and Polanyi’s Observation
To understand why AI is struggling to replicate top-tier performance, one must look at the scientific distinction between explicit and tacit knowledge. Explicit knowledge is information that can be easily articulated, recorded, and shared—think of a company’s employee handbook or a software manual. Tacit knowledge, a term popularized by scientist and philosopher Michael Polanyi, refers to the things we know but cannot easily explain. Polanyi famously noted, "We can know more than we can tell."
For a senior engineer, a master salesperson, or an experienced project manager, decades of practice have turned their expertise into an automatic response. They make critical decisions based on "gut feeling" or "intuition," which are actually high-speed cognitive processes utilizing patterns they have recognized over thousands of iterations. Because this knowledge is automatic, these experts often cannot break their decisions down into discrete steps when asked.
When organizations attempt to implement AI, they typically feed the models their explicit documentation. The AI then scales this mediocre, formalized version of the work. It cannot scale the "magic" of the top 5% of performers because that magic has never been captured in a format the AI can digest. Consequently, the AI becomes a tool that reinforces average performance rather than elevating the entire workforce to the level of its best people.
The Economic Cost of Knowledge Loss
The failure to capture tacit knowledge is not a new problem, but the "Great Retirement" and the rapid pace of AI development have made it an urgent financial crisis. Data from the American Productivity & Quality Center (APQC) indicates that 85% of senior executives are deeply concerned about the loss of institutional knowledge as experienced employees leave the workforce. Despite this concern, only 8% of organizations have a consistent, effective strategy for preserving that knowledge.
The daily economic impact of this "knowledge gap" is profound. APQC findings suggest that the average professional spends approximately eight hours per week—a full workday—searching for information or re-explaining concepts they have already mastered to others. This inefficiency is a direct result of knowledge not being available in a transferable, structured form.
Furthermore, the onboarding process remains a significant point of failure for most enterprises. Gallup reports that only 12% of employees strongly agree that their company does a good job of onboarding. In most complex roles, it takes a new hire up to a year to reach peak productivity. This lag is almost entirely due to the time it takes for a new employee to absorb the tacit knowledge of their peers through osmosis. If a company cannot effectively transfer this knowledge from one human to another, it is statistically impossible for them to transfer it to an AI model.
A New Framework for AI Implementation: Extraction Before Automation
To bridge the gap between AI investment and value, organizations must shift their focus from "tool procurement" to "knowledge extraction." This requires a methodology rooted in educational science and behavioral analysis. Instead of asking experts to "write down what they do," companies must engage in a more rigorous process of deconstruction.
- Identifying Decision Nodes: Companies should identify the specific moments in a process where a high performer’s path diverges from a novice’s path. These are the "decision nodes" where tacit knowledge is most active.
- Behavioral Mapping: Instead of relying on self-reporting, analysts must observe experts in real-time or review their work history to identify the criteria they use to discard options, the "red flags" they notice, and the specific shortcuts they take that aren’t in the official playbook.
- Structured Codification: Once these insights are extracted, they must be turned into "structured explicit knowledge." This is the only way to create a high-quality dataset that can actually improve an AI model’s performance.
Organizations that have historically invested in "Knowledge Management" (KM) are finding AI adoption significantly smoother. These companies already have a culture of documenting the "how" and "why" behind decisions, providing a rich, high-context library for AI orchestration layers to draw upon. For these firms, AI is not a replacement for human thought, but a high-speed delivery mechanism for the company’s best collective wisdom.
Broader Impact and the Future of Corporate Intelligence
The implications of this shift are significant for the future of the global economy. As the labor market continues to tighten and the "silver tsunami" of retiring experts accelerates, the ability to "download" corporate expertise into AI systems will become a primary competitive advantage. Companies that fail to do this will find themselves trapped in a cycle of "AI mediocrity," where they pay for expensive technology that only produces generic, low-value outputs.
Moreover, this shift changes the role of the executive. Leadership is moving away from managing processes and toward managing "knowledge assets." The goal is to ensure that knowledge is no longer the personal property of an individual employee but a liquid asset of the corporation. When knowledge is successfully extracted and integrated into an AI-driven orchestration layer, the results are transformative. Onboarding times can be slashed from months to weeks, and new divisions can be scaled rapidly because the "blueprints for success" are readily available and actionable.
In conclusion, the current "AI winter" regarding ROI is a self-inflicted wound caused by the neglect of human expertise. If an organization’s last AI tool failed to meet expectations, the solution is not to buy a more powerful model. The solution is to look inward and ask whether the company actually understands how its best people work. Until the "tacit" becomes "explicit," AI will remain an expensive mirror, reflecting the documented bureaucracy of a company rather than its actual intelligence. The invoice for AI tools will continue to come due; whether it is viewed as a cost or an investment depends entirely on the quality of the knowledge fed into the machine.
