Putting Knowledge in Writing: Modern KM and AI Can Elevate How Documents Transmit Knowledge

Putting Knowledge in Writing: Modern KM and AI Can Elevate How Documents Transmit Knowledge

Capturing what an organization knows has never been as simple as writing it down. Yet the fantasy persists: that knowledge can be bottled, labeled, indexed, and made universally discoverable. Portals. Wikis. Shared drives. Enterprise search. Content stacks that promise intelligence, insight, and reuse, if only the tags are right, the permissions tuned, and the categories aligned.

But writing is not knowing. And documents, for all their utility, are only fragments. They freeze knowledge, reflection, insight and data in time. They imply context but often fail to convey it effectively. Documents make declarations, but cannot guarantee understanding because technology offers few feedback loops for readers. And unless someone’s paying attention, they age into irrelevance, gathering digital dust as workflows and products, roles and people, technologies and customers that either no longer exist or have evolved beyond the state where those documents are relevant.

Read more on KM and AI on our Substack

Generative AI doesn’t solve this problem. The rapid rush to encode content, sometimes any content, reveals how brittle and shallow much of our recorded “knowledge” really is.

The Half-Life of Codified Knowledge

The more we treat documents as containers of truth, the more we misunderstand how knowledge lives and moves. A slide deck from a strategy retreat may be preserved on the corporate intranet, but its key takeaways—if they existed—were spoken aloud, refined through disagreement, misunderstood by some, reinterpreted by others. The artifact is residue, perhaps a pointer, but it is not the substance.

The same applies to policy documents, onboarding manuals, decision memos, and even technical specifications. Each contains some useful representation of thinking, but none is complete. Worse, none of these documents comes with an expiration date. Organizations rarely have mechanisms to reassess whether the assumptions, processes, or terminology embedded in their artifacts still hold. Simply “putting knowledge in writing” isn’t enough.

Knowledge ages. Context shifts. Audiences evolve. The problem isn’t just that old documents are hard to find; it’s that they persist long after their usefulness, clinging to authority because no one marked them obsolete.

AI can’t yet tell the difference between wisdom and obsolescence. It doesn’t intuit when the world has changed around a document. That remains our responsibility.

Putting Knowledge in Writing: Content Without Context is Not Knowledge

Organizations have long overestimated what documents can do on their own. A well-written procedure might describe a process step by step, but if the audience lacks the required background or misunderstands the context, the document will confuse rather than clarify. Internal emails, for instance, tend to be terse, domain-specific, and often deeply ambiguous to anyone not directly involved in the exchange. Their meaning depends on timing, relationships, and shared history. In isolation, they raise more questions than they answer.

Knowledge is not merely about access. It’s about understanding. And understanding depends on context derived from communication experiences that involve audience, purpose, scope, assumptions, background, interpretation, and use. The totality of the experience creates the understanding, not its artifacts, regardless of their informational fidelity. Like a compressed JPEG image, something is always lost, and the observer cannot recover what they never knew existed.

In that light, organizations can’t afford to treat metadata as an afterthought. For knowledge to travel better than it does today in most organizations, documents must carry more than titles and timestamps. They need meaningful scaffolding:

• Who is the intended audience?
• What does the reader need to know before reading?
• What assumptions underpin the recommendations?
• Has this been validated, and by whom?
• Is this current or historical?
• What bias, if any, does the author bring?

Without such scaffolding, knowledge seekers are left to guess. And in complex environments, guessing leads to misalignment, inefficiency, or worse, avoidable mistakes.

Dialog, Not Delivery

A single document cannot convey all that its author knows. It captures a snapshot, a distillation, a choice of what to include and what to leave out. Real knowledge transfer happens in the space around the document, in the questions it provokes, the comments it invites, the decisions it supports, and the misunderstandings it corrects.

This is why the most valuable documents live in conversation. They’re subject to revision. They’re linked to updates. They circulate among teams. They accumulate annotations. They change with experience.The get retired when they are no longer useful, or reinvented to create continued value.

Knowledge is social. Documents are static. The gap between the two is filled by collaboration.

For years, knowledge management efforts have conflated access with utility. Portals proliferated under the assumption that if people could just find the right content, they’d know what to do. But much of that content was unqualified. Undated. Untested. Written for one audience and repurposed for another. At best, these repositories provided fragments. At worst, they undermined trust in the entire knowledge system.

Apply the old patterns, and the application of AI repeats those mistakes at scale, but it can also create new paths for collaboration.

What AI Reveals (and What It Doesn’t)

Generative AI introduces new possibilities. It can draft documents, summarize meetings, extract action items, and even suggest next steps. It can read across a corpus and find thematic links or highlight inconsistencies. But it doesn’t know which version of a document matters. It doesn’t understand organizational nuance or interpersonal dynamics. It doesn’t distinguish between a bold proposal and a sarcastic remark.

In this sense, AI amplifies the content problem. If the underlying source material lacks clarity, coherence, or context, the output will reflect those limitations.

This raises the stakes for how content is authored, reviewed, and maintained. It demands a shift from documents as products to documents as part of a continuous process (which they always should have been). AI can help drive engagement, not by generating content, but by sensing change and making stakeholders aware. It can pay attention in ways that humans cannot. While AI may never understand the entirety of context, it can reveal and elevate patterns in ways people cannot, and those patterns can create new conversations about relationships and practice, processes and facts, and guidelines and learning opportunities that become the basis for decisions and human input.

Of course, organizations that don’t value knowledge won’t make any more effective use of AI than they did of more traditional technology, or non-technology-based knowledge practices. AI cannot transform apathy any more than it can turn poorly captured knowledge into a meaningful enterprise asset.

The organizations most successful in applying AI to knowledge work are not those with the biggest content archives, but those with active, maintained, and contextualized knowledge systems. These systems rely not on humans to write everything down, but on people who actively engage in refining, interpreting, and validating what they know. Putting knowledge in writing is just the first step of an ongoing process.

Designing for Knowledge Use, Not Just Storage

The future of knowledge work is less about what we record and more about how we engage. Content should not be designed to sit on a server. It should be designed to live inside decision-making cycles. That means:

• Writing for the future reader who doesn’t know the inside jokes, the acronyms, or the circumstances that shaped the original content.
• Layering content for multiple audiences: executives, implementers, newcomers, skeptics.
• Describing confidence levels, assumptions, known gaps, and open questions.
• Assigning stewardship so documents don’t drift unexamined into the archive.
• Marking intent: is this the answer, a suggestion, or a provocation?

Documents, when treated this way, become more than containers. They become signals. Prompts. Artifacts of trust.

The Rise of the Knowledge Sculptor

As AI systems take on more of the initial drafting, new roles emerge. The knowledge sculptor isn’t just an editor. They’re a domain-aware curator. They shape AI-generated output, ensure alignment with intent, challenge assumptions, and prepare content for real-world use. They connect artifacts to context and ensure that what emerges from the machine has meaning to those who read it. Many roles will likely evolve to include this activity as part of their knowledge description, especially as personal agents take over task work.

While some may complain that the effort at putting knowledge in writing well is overwhelming or slows everything down, the focus should be on improving the quality of what circulates. Bad knowledge at scale is worse, and more dangerous, than no knowledge at all.

Documents Are Only the Beginning

Knowledge management was never about search engines or metadata fields. Putting knowledge in writing was, and remains, about trust, clarity, context, and shared meaning.

AI will accelerate how we create and consume content. But unless that content is built for understanding—unless it’s validated, explained, and maintained it will only accelerate confusion.

Documents will never represent the totality of what people know, but human history has used them effectively to help transfer knowledge, as long as teachers and mentors, managers and coaches, employ them to create knowledge environments where learners (and workers as learners) sense, adap,t and share, continously creating value through interactions with each other, with customers, and with the legacy and values of their organization’s in ways that make finding, applying and updating knowledge a critical activity.

For more serious insights on AI, click here.

For more serious insights on management, click here.

Did you enjoy Putting Knowledge in Writing? If so, like, share or comment. Thank you!