How To Make Tribal Knowledge A Company Asset: AI Can Help, But Can’t Do It Alone
Aaron Rallo is Founder & CEO of Trovia, a knowledge layer for content that makes enterprise AI accurate and trustworthy.
gettyWhat’s the one thing that McDonald’s does exceptionally well?
It’s foundational to scaling globally, operating consistently and maintaining brand loyalty. It’s an activity humans have done since the dawn of time. It’s astonishingly difficult for most companies.
Although it sounds simple, sharing knowledge is challenging to do effectively. After reading an SOP, attending training or shadowing a colleague, no employee can perform a task exactly as specified. Effective knowledge sharing requires excellent training resources, sufficient time to learn and consistent performance feedback.
Much of McDonald’s success came from meticulously documenting their procedures and even teaching them at the Hamburger University. Not all organizations have the resources to encode their processes into a full-blown curriculum, but all organizations need to transfer knowledge if they wish to sustain and expand.
In the rapidly growing residential and commercial cooling industry, for example, thousands of new technicians are needed, and most of the existing workforce will be retiring soon. In this industry, like most, while some knowledge can be documented, the really crucial stuff lives in the heads of your top performers. According to organizational theorist Nonaka’s widely accepted model, knowledge transfer follows an infinite path from implicit to explicit. Your grandmother’s pinch of salt is implicit knowledge. Writing it down as ⅓ teaspoon of salt makes it explicit. Once the knowledge is recorded, operating procedures evolve and the cycle starts again.
In university, I worked in a warehouse for a fast-growing retailer. There were procedures for optimizing the packing of delivery trucks, but the senior team knew certain steps only applied to certain products.
This “tribal knowledge” exists in some form in every company and it creates at least two problems: you lose it when your knowledgeable people leave and your new hires burn time solving problems that have already been solved. The senior people around me did their best to share as much of this tribal knowledge as they could, but we didn’t have an effective way to capture what they knew.
This is an area where AI can make a real difference. AI’s sheer capacity for ingesting and synthesizing documents make it a game-changing tool for organizations to be able to codify, standardize and scale its operating procedures—those that are explicit, i.e., already written down, and those that are implicit, i.e., living in the heads of your best people.
Here is a practical approach to setting up a knowledge-sharing system for your business that enlists the help of AI. AI can’t do it all, but it can make much of this process a whole lot easier.
Pull every SOP, manual, checklist and training doc into one place. Right now, these are likely scattered across drives, inboxes and binders. Then, use AI tools to reconcile the content. Look for any conflicts in the data. Is there a policy from 2021 that says something different than the one from 2025? Corrections can be immediately captured. Next, address knowledge gaps by querying the content to see what it can’t answer. Doing this with AI is much less time-consuming than doing it manually.
To capture what’s missing and ensure your system’s knowledge is correct, you’ll need to interview your subject matter experts. The human element of knowledge sharing has its own challenges. Contrary to popular belief, senior team members do not withhold knowledge for fear of losing status. Szulanski’s acclaimed study suggests they don’t intentionally withhold information at all. The issue is structural, not psychological. The biggest barriers are 1) the recipient’s ability to absorb the information (they do not have sufficient prerequisite knowledge), 2) causal ambiguity (it’s not always evident which part of the skill is actually the most valuable to teach) and 3) an arduous relationship (there isn’t enough ease of communication to fully transfer the information).
• Instead of handing your experts a blank page and asking them to produce a training doc, use AI to turn your existing procedures into a first draft.
• Make it effortless for your experts to review content. It should be one tap or one message for a senior tech to say, “ignore step 3, here is what really works” and have it update the record. Use an AI system to generate whatever artifact—a doc, a quiz, a text—that will be the easiest for your expert to review.
• Storytelling is more effective than interrogation. “Walk me through the last job that didn’t go as planned” will elicit a richer share than, “What are the main things you check for to ensure a job goes well?” Feed those stories back into your AI knowledge sharing system so the judgment gets captured next to the procedure.
As Nonaka’s paradigm suggests, processes are codified and then implicit knowledge arises that needs to be documented and captured. Your knowledge base won’t ever be perfect. It requires consistent perfecting, i.e., ongoing maintenance, for it to be useful. What’s incredible about using AI for this process is that you can build a system that gets smarter the more you use it. Every query of the knowledge base improves its accuracy, relevance and comprehensiveness.
What’s also neat is that this dynamic knowledge base consists of hard-won operational know-how and your learnings over time. If sufficiently robust, it could even be a valuable asset that extends beyond knowledge transfer. That is the difference between knowledge that walks out the door and knowledge that compounds.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
