The Knowledge Problem That Isn't Going Away
Every organization above a certain size has the same problem: knowledge is everywhere, and it is nowhere. Documents are scattered across SharePoint sites, Teams channels, shared drives, personal OneDrive folders, and notebooks that only the author knows how to find. Progress reports sit in email threads. Critical institutional knowledge, including the context behind a decision and the lessons learned from a failed initiative, lives in someone's head and disappears when they leave.
Most knowledge management initiatives address the symptoms without touching the underlying structure. A taxonomy is imposed. A SharePoint site is reorganized. Teams are asked to document their work in a new format. And then, six months later, the same knowledge chaos has reasserted itself, because the incentives and workflows that created it have not changed.
What changes the outcome is not a new filing convention. It is a structural transformation of how knowledge is created, captured, and retrieved, paired with the technology to make that transformation sustainable without requiring ongoing heroic effort from people who are already busy doing their actual jobs.
Knowledge management done well is not a system. It is a managed capability that evolves with the organization and compounds in value over time.
What an AI-Led Knowledge Management Engagement Actually Entails
A recent engagement with a global philanthropic foundation illustrates what a comprehensive knowledge management engagement looks like when it is approached with the right scope.
The engagement started from a familiar place: disaggregated knowledge assets, inconsistent folder structures, documents named differently by different teams, and progress reporting that existed in some places and was absent in others. The instinct for many organizations in this situation is to clean up what exists. Our instinct was to understand what should exist and then build the infrastructure to sustain it.
The process involved four interconnected pillars:
- AI-led knowledge transformation: Applying AI to audit existing knowledge assets, identify gaps, standardize taxonomy and naming conventions, and migrate content into a coherent structure at a scale and speed that would be prohibitive to accomplish manually.
- Reporting integration: We identified that knowledge management and internal progress reporting were deeply connected for this organization. KPI tracking and program reporting were part of the knowledge infrastructure, not separate from it. Integrating both into a unified approach expanded the scope and the value of the engagement significantly.
- Change management: No AI system can change how people work. That requires people. Our domain experts led the change management process, working with teams directly, supporting adoption, and building the habits and practices that make a knowledge system self-sustaining.
- Agentic AI roadmap: Rather than deploying a generic AI tool, we conducted a comprehensive assessment of the organization's knowledge management use cases, mapping 21 distinct use cases across the client team's workflows, and developed a prioritized roadmap of where agentic AI would create the most value and how.
What 21 Use Cases Reveal
The 21 use cases mapped during the engagement were not a theoretical exercise. Each one was grounded in how the team actually worked: what knowledge they needed, where they were going to find it, what happened when they could not, and what the cost of that friction was.
What the assessment surfaced was a differentiated landscape of where AI could help, ranging from what was already supported by existing tools like Microsoft Copilot, to what could be addressed with Evalueserve accelerators, to what required purpose-built agentic solutions. The result was a roadmap, not a wish list: a sequenced plan that started with quick wins and built toward a more sophisticated agentic infrastructure.
The two agents prioritized as quick wins were selected because they addressed the highest-friction points in the knowledge workflow that no other tool, including Copilot, was able to handle:
- Inbox Knowledge Harvester: An agent that monitors and processes email threads and inboxes to surface and extract knowledge assets embedded in day-to-day communications. It identifies relevant documents, decisions, and insights shared in email, structures them against the organization's taxonomy, and routes them into the knowledge base automatically. The inbox is one of the largest repositories of institutional knowledge in any organization and one of the most consistently overlooked by conventional knowledge management approaches. Privacy and confidentiality are built into the design from the start: the agent is scoped to defined mailboxes and content types, excludes personal and sensitive communications, and operates within the organization's existing data governance policies.
- Nudging Agent: An agent that integrates into existing workflows to prompt knowledge capture at the moment knowledge is actually created, whether at the close of a meeting, the completion of a milestone, or the resolution of a key decision. Rather than asking people to document retrospectively when context has faded, the agent creates low-friction capture opportunities in the flow of work itself, where the information is fresh and the documentation is worth having.
The two agents we prioritized address exactly the knowledge gaps that neither existing tools nor internal teams had the expertise to solve. That is where the value is: not in replicating what already exists, but in extending what is possible.
From a Single Team to a Program
When knowledge management is done as a project, a one-time effort to clean up SharePoint, the results are temporary. Teams revert. New knowledge accumulates in new places. The problem reasserts itself.
When knowledge management is done as a managed service, with ongoing governance, a roadmap of agentic AI improvements, and a partner responsible for sustaining and evolving the capability, the results compound. Each new use case added to the roadmap builds on the infrastructure that already exists. Each team that adopts the program extends its reach and its value. This is exactly what we’re seeing with our client. The initial engagement is now generating interest from three to four additional teams within the same organization.
For a large organization managing hundreds of programs and dozens of internal teams, this is the difference between a knowledge initiative that fades and a knowledge infrastructure that becomes a genuine organizational asset.
The Broader Applicability
The knowledge management problem is not unique to philanthropic foundations. The same dynamic, disaggregated assets, inconsistent structures, and critical knowledge trapped in the wrong places, exists in every enterprise across every industry.
What makes the problem tractable now is the combination of AI agents that can work at scale and domain expertise that knows what good looks like. Agents can process and organize content at a volume that no team of human analysts could match: ingesting documents, applying taxonomy, routing knowledge assets, and triggering capture prompts across an entire organization simultaneously. But an agent is only as good as its configuration. Without someone who understands the domain, the organizational context, and the governance requirements, an agent optimized for the wrong thing produces noise at scale rather than knowledge at scale.
That is why the Human-in-the-Lead model is not a constraint on what agents can do. It is what makes agents actually useful. Our domain experts define what each agent looks for, set the quality thresholds, manage the lifecycle as the organization evolves, and step in when edge cases require judgment that no agent should be left to make on its own.
Those things require people who understand the domain, understand the organization, and are committed to the outcome rather than just the delivery. That is the Human-in-the-Lead model applied to knowledge management. And that is what makes the difference between a knowledge management initiative that works for a year and a knowledge management capability that works indefinitely.
