Introduction: The Time That Disappears
"Now where did that document go?" — there's hardly a day in any organization where that question doesn't come up. Working on ZEROCK at TIMEWELL, I've had the chance to speak with information systems teams and corporate planning departments at many companies. The one thing I hear consistently: frustration at the enormous time spent just searching for information.
McKinsey research puts the figure at roughly 20% of a knowledge worker's time spent on information retrieval. Assuming 2,000 working hours per year, that's 400 hours — about 50 working days — consumed by searching. The number is striking. But for people in the field, it tends to feel about right.
This article examines why enterprise information retrieval is so inefficient, diagnosing the structural causes, and then considers what knowledge management should look like going forward.
Struggling with AI adoption?
We have prepared materials covering ZEROCK case studies and implementation methods.
The Three Structural Causes of Information Retrieval Failure
1. Scattered Information: The Maze That Organizational Silos Create
The most fundamental problem is that information lives across multiple systems and folders. Most organizations have different departments using different tools with different file management rules. Sales uses Salesforce for customer data. Engineering manages technical documentation in Confluence. HR stores policy documents in SharePoint. Each system stores its piece — and none of them talk to each other.
Cross-system search is difficult, which means every retrieval starts with a first-order judgment call: "which system should I check first?" At one manufacturing company we worked with, seven different document management systems coexisted within the organization. New employees averaged over 40 minutes to reach the information they needed — a startling reality.
2. Inconsistent Naming Conventions: The Fragility of Relying on Human Memory
The second problem is that file and folder naming conventions are not standardized. "Proposal_FinalVersion," "Proposal_FinalVersion2," "Proposal_ActuallyFinal" — most people have seen exactly this pattern. It sounds like a joke, but it causes serious problems.
When a file is created, the name feels perfectly identifiable in the moment. Three months or six months later, recalling the context of that name is hard. If you weren't the person who created it, finding it is even harder. As is pointed out in introductory knowledge management literature, "naming" — the most basic step in converting tacit knowledge to explicit knowledge — is frequently mishandled, which means organizational knowledge assets effectively become inaccessible.
3. Knowledge Silos: What Leaves When People Leave
The third problem is knowledge siloing in individuals. "You'd need to ask Tanaka about that" is a conversation that happens regularly in many organizations. The know-how for specific business processes or customer handling exists only inside one person's head, never captured in a form that can be shared.
This problem becomes acute when "Tanaka" leaves or transfers. Knowledge accumulated over years survives only as fragments in a handover document — the majority is lost to the organization permanently. At one IT company we supported, when a key engineer resigned, it took over three months to reconstruct the system configuration that only that engineer had known.
Why Knowledge Management Is Suddenly Urgent
DX and the Necessity of Knowledge Sharing
As the digital transformation wave sweeps every industry, many organizations are overhauling systems and rethinking business processes. But what tends to get overlooked in this process: "what do we do with the knowledge inside people's heads?"
You can deploy a new system, but if the know-how to use it effectively isn't shared, adoption stays low. And if you can't make implicit operational knowledge visible when redesigning processes, you can't design truly efficient workflows. METI's DX Report makes the same point: what determines DX success or failure isn't the technology deployment itself — it's how effectively an organization can leverage its knowledge.
Remote Work Accelerated the Problem
The COVID-era shift to remote work, which became permanent at many organizations, made knowledge management challenges more visible. Things that could be resolved by "asking the person next to me" became genuinely difficult in a remote environment.
Sending a chat message doesn't guarantee an immediate response. Information that used to be shared naturally in casual conversation now requires intentional broadcasting to reach anyone. The result, in many organizations: "information gaps" within the same company — significant variation in what different people know, even though they work for the same organization.
The Limits of Conventional Knowledge Management
The Hollowing-Out of Internal Wikis
Many organizations have tried internal wikis — Notion, Confluence, or a custom solution — with instructions to "consolidate information here." The results, more often than not: initial enthusiasm followed by gradual hollowing-out. Pages updated diligently at launch go stale as contributors change roles or get busy. What's left is outdated information — an "information graveyard." With no incentive to update and inadequate search, organizations fall into a cycle: "nobody reads it, so nobody updates it; nobody updates it, so nobody reads it."
The Labyrinthine File Server
Similarly, file server management has reached its limits in many organizations. Folder structures that have accumulated over years have grown far beyond their original design intent into something that nobody fully understands. At one trading company we spoke with, the file server folder hierarchy had reached 12 levels deep — reaching a target file could require more than 10 clicks. Unless you remember precisely where something lives, finding it is essentially impossible.
Knowledge Management for the AI Era
From "Searching" to "Finding"
Traditional knowledge management operated on a model of "store information correctly; retrieve it through the correct path." This approach depends on human memory and behavioral consistency — and becomes increasingly fragile as organizations grow.
Advances in AI — particularly large language models (LLMs) and RAG (Retrieval-Augmented Generation) — enable a fundamentally different approach: finding semantically relevant information regardless of where it's stored. Rather than depending on file names or folder structures, you ask in natural language — "how do we handle customer complaints?" — and get relevant documents, past emails, and chat history retrieved cross-system.
From Accumulation to Activation
The other important shift is the move from "how do we accumulate information?" to "how do we activate what we've accumulated?" The old model was: store as much as possible; rely on individuals to make use of it. With AI assistance, accumulated information can become the basis for generating new outputs.
Draft a new proposal using past proposals as reference material. Train a system on internal technical documentation so it can automatically answer questions from new employees. What was once a "nice to have" for knowledge utilization is becoming something that can sit at the core of business operations.
Conclusion: What's at Stake Is Knowledge as a Management Discipline
Knowledge management challenges are not merely a matter of convenience. They are management challenges that directly affect organizational competitiveness. Reduce time spent on information retrieval and that time becomes available for creative work. Convert experienced employees' knowledge into organizational knowledge and the organization becomes more resilient against attrition risk. Consolidate scattered information and data-driven decision-making becomes possible.
At TIMEWELL, this understanding drove the development of ZEROCK. Using GraphRAG technology, ZEROCK connects semantically related information scattered across an organization and delivers an environment where the right information can be found in under 10 seconds. If knowledge management is a challenge your organization is facing, we'd welcome a conversation.
The next article covers GraphRAG technology — what ZEROCK uses under the hood — explaining how it differs from conventional RAG and where it delivers the most practical value.