I recently had to navigate a scenario that makes every operational lead feel uneasy, the realisation that critical knowledge of a core system was concentrated in one place. We were approaching a point where we needed that knowledge to be accessible to the wider business, and I knew that a standard, unstructured “knowledge download” wouldn’t be enough to stabilise our understanding of the platform for the long term.
The Management Ceiling
As an operational lead, I’m responsible for several software stacks and the processes that support them, but I am not a developer myself. My goal was simple but high-stakes, I wanted our documentation to be at a point where any professional, whether an internal team member or an external contractor, could pick it up and immediately understand the state of the software, the database relationships, and the principles behind how it was built – ultimately supporting the business’ ongoing resiliance
In thinking about how I’d approach the problem, it occurred to me that the friction in this kind of knowledge transfer is often the manager’s own technical ceiling. If I had simply written down what I knew from my perspective, the outcome would have been superficial. I didn’t have the technical depth to anticipate the specific questions an expert would need to ask twelve months from now. Without a better approach, we’d be at risk of “stumbling” across critical gaps in our understanding only after the person with the answers was no longer available.
Moving Beyond Model Overwhelm
The traditional way to solve this is to look for a framework. There are hundreds of documentation models out there, like Arc42 or C4, but finding the right one can become a task in itself. For a non-technical manager, these models can feel disengaging or overly academic. It is easy to get stuck choosing a model rather than actually applying it.
I decided to use AI as a thinking partner to bridge this gap. Instead of asking it to “write documentation”, I used it to help me navigate the world of frameworks and find a version that matched our specific context.
We didn’t just pick a template and fill it in, we spent time bouncing the problem back and forth. I provided the organizational context and the AI surfaced the technical principles. When a model felt too rigid, we blended it with elements from another. The breakthrough came when I had the AI reposition these abstract frameworks as a targeted Q&A directed at my specific use case.
From Novice to Auditor
This process turned me from a relative novice into a rigorous auditor. Instead of approaching our technical lead and asking them to “write down what you do,” I was able to present a precise, high-quality framework of questions. These weren’t generic prompts, they were specific inquiries about our software’s architecture, support dependencies, and development principles that I wouldn’t have known to ask 24 hours earlier.
Because the AI helped me do the heavy lifting of the structure, I was able to feed back the right information and build out a set of documentation that was 90% complete before any deep technical intervention was even required.
This meant the final 10% of the our time hasn’t been wasted on explaining the basics. We’ve used that time to tease out the “tribal knowledge” and the nuanced edge cases that usually only get found by chance when something breaks. We’ve moved from a siloed knowledge base to a “ready-state” blueprint for our systems.
Augmenting the Perspective
There is a lot of noise about AI having “killer features” or a singular “real power,” but I think that misses the point. My experience with this documentation project showed that one of AI’s value lies in its ability to augment a manager’s perspective and amplify the process of capturing knowledge.
It is about taking the principles behind established models and applying them in a way that is grounded in the reality of your specific business. AI allows you to bypass the “model overwhelm” and move straight to the practical application. It drives you to a far higher level of detail and anticipation than you would ever reach by simply writing down what you know from your own limited viewpoint.
If you have critical systems knowledge locked in a single individual, you have an operational vulnerability. You don’t need to be a technical expert to build a resilient system, but you do need a process that bridges the gap between management and execution.
Don’t wait until you’re staring at a knowledge gap to start the download. Use AI to build the framework for your documentation today, and ensure your systems are built to outlast any single individual’s tenure.