In the first two articles in this series, I explored some of the assumptions AI is challenging across knowledge work, and why AI adoption inevitably becomes a learning process. Now I want to turn to the knowledge itself.
Knowledge capture has always presented a challenge. Some of an organization’s most valuable knowledge comes from experience, institutional memory, and conversations, and much of it may never be formally recorded.
Clearly, AI can’t find what isn’t there. But as it becomes better at searching, connecting, and analyzing organizational information, that obvious limitation deserves another look. In this post, I explore what AI might help us uncover and where the real knowledge gaps lie, outline a useful starting framework, and discuss how we could use AI to make more of the knowledge we create.
What If AI Can Find What We Couldn’t?
Knowledge management has long distinguished between explicit knowledge that can be documented and tacit knowledge that is difficult to articulate. But there is another category that can easily be confused with tacit knowledge: knowledge that could have been recorded but wasn’t, or was recorded but became difficult to find.
Historically, that distinction made little practical difference. If nobody could find the information, it was effectively inaccessible either way. Finding something often required knowing which system to search, what the file might have been called, who was involved, or roughly when it happened.
AI offers a different way of interrogating records. It can help describe previously inaccessible items. Retrieval-augmented systems can draw on approved organizational sources and synthesize information from multiple documents. Recorded meetings and conversations can become searchable alongside more conventional knowledge assets.
That may allow us to surface organizational knowledge we couldn’t previously find.
When the Problem Is Us
That leads to a less comfortable question.
If an AI system has access to an organization’s records and still provides unsatisfactory answers, where does the failure lie?
Organizations have always created selective records of themselves, with inevitable gaps. Meeting minutes are useful and record what was considered important at the time. However, useful yet random pre-meeting chat is not included. Annual Reports are carefully edited and condensed. Successful projects are more likely to produce polished case studies than failed experiments.
Gaps can emerge anywhere. People leave, systems change, and information repositories disappear or are forgotten. As retrieval becomes more capable, these weaknesses become harder to treat simply as search problems. The challenge shifts towards what we choose to capture and preserve—and how comprehensive we need to be.
Could AI Remember the Organization Better Than We Do?
There’s an intriguing possibility on the other side of this problem. What if an organization has recorded far more of its history than we realize?
Suppose the accepted institutional wisdom is that a previous initiative failed because customers did not want it. An AI-assisted examination of the surviving record might reveal a more complicated story: budget constraints, technical problems, disagreement between departments, changes in leadership, and assumptions that no longer apply.
AI has not necessarily uncovered “the truth”, but there might be more to unpack—it may reveal a gap between what an organization remembers and what its records show. We may therefore be approaching an unusual point in knowledge management: AI could have access to more of an organization’s recorded history than any individual working within it.
I want to stress this carefully—if AI can increasingly reconstruct parts of that history, any resulting account must still be carefully evaluated for provenance, authority, context, and what the AI may have missed or misrepresented.
But What About Knowledge That Doesn’t Exist Yet?
There is another kind of knowledge AI can’t find—the knowledge we haven’t yet created.
Organizations do not develop knowledge solely by retrieving what they already know. New knowledge emerges when people encounter problems, exchange ideas, disagree, experiment, make mistakes, and reconsider what they thought they understood.
This helps explain the continuing value of bringing people together. Research into hybrid working points to the importance of informal communication, workplace visibility, and opportunistic encounters in encouraging knowledge sharing. Similar dynamics exist at an in-person conference, seminar, project meeting, or simply in conversation between colleagues.
Everyone enters a room with varying levels of expertise and experience, but under the right conditions, something else happens. One observation prompts a further question or request for clarification. Someone challenges an assumption; another person connects it with an apparently unrelated problem.
People leave with new ideas and want to take them back to their own situations within their area of expertise, or their organizations. How can they best do that?
From Knowledge Capture to Knowledge Amplification
Consider what happens after a conference—despite efforts to the contrary, much of its value can be ephemeral. People might receive copies of slides or write up their notes. They might post on LinkedIn about how the conference changed their thinking. Ultimately, though, much of what they learned may never be captured or shared when they get back to their desk.
There are tools available to capture conference information in real time. With appropriate consent and governance, sessions and discussions can be transcribed and analyzed, preserving themes, discussion points, and unanswered questions.
However, I am more interested in what happens to those ideas when the conference closes. AI could help people reflect on what they have learned. A dialogue with carefully chosen prompts could help someone interrogate and organize their thoughts.
For example, which assumption did you reconsider? What might apply to your own organization? What would you like to try? What are you doing already? How can something be adapted for your workplace?
This moves knowledge capture beyond recording what was said, towards capturing what happened as a result. Those insights could be connected with what the organization already knows—or potentially reveal something it does not. It becomes a knowledge asset in its own right.
The process could look something like this:
human interaction → emerging knowledge → AI-assisted reflection and analysis → new connections → human interpretation → further knowledge
There is an immediate benefit to this. Conferences, seminars, and professional development represent an investment of money and staff time. AI-assisted reflection could help more of that value feed back into the organization. AI is no longer simply helping organizations find what they know; it can help them make more of what their people will come to know.
AI-assisted Reflection Can Lead to New Ideas and Insights
AI can help us find information we had forgotten, overlooked, or simply struggled to retrieve. In doing so, it may expose weaknesses in our organizational memory and make us think harder about what we capture and preserve.
For me, however, the exciting part is thinking about the knowledge that doesn’t yet exist. This emerges through conversations, experience, disagreement, unexpected connections, and new ideas. AI could help us capture those insights, connect them with what we already know, and help fresh ideas spread through the organization. Our only problem then is how much we will be learning!
Next in this series, I’ll explore what happens when knowledge systems go one step further—and start acting on what they know.
0 Comments