In previous posts I used the term “tribal knowledge” quite often. A few of you asked me some clarifying questions, so I decided to write a blog post about it.
The term “tribal knowledge” is used informally to highlight when knowledge is retained by a community of people or “tribe” – often by word of mouth or unspoken habit. Tribal knowledge in IT operations is any knowledge that is not properly documented in a Standard Operating Procedure (SOP).
This term does not have a negative connotation per se. But it is an impediment to fast IT adoption – perhaps the biggest one. In my experience, about 50% of IT support activities are documented only by tribal knowledge.
When crucial know-how walks out the door
The most basic problem with letting crucial knowledge reside only in people’s heads is that people inevitably move on. Nearly everyone has been in a workplace situation where “good ol’ Doug” retires or changes jobs… and then you realize he was the only one who knew how a database had to be structured in the antiquated payroll app you still use.
The tribal knowledge problem is even more acute for AI transformations. Perhaps this is because AI is not just a very fast-growing technology, which speeds up the timetable that IT operations must support; it also emulates human decision-making. However, AI needs a context to operate – i.e., a model, as I outlined earlier this year. And that model needs to be built on well-crafted SOPs. Tribal knowledge, which by definition is fragmentary, context-dependent, and not thoroughly documented, doesn’t provide the necessary support to build an AI model robust enough to run IT operations.
Knowledge is power… and that’s the problem
That brings me to the second major problem with tribal knowledge: Usually the “tribe” who retains knowledge is not prone to share it. As the old saying goes, knowledge is power… and humans are not very collaborative when they need to share power.
Many IT staff, especially System Integrator teams, see AI for automating IT as a direct competitor for their jobs. I remember when we were deploying ignio for a huge Digitate customer: Timelines were dragging because SI management was resisting the idea that a machine could perform – with better quality – the same task that their team performed for years. It took months of discussion to move forward.
This means AI transformations’ real dilemma is how to share knowledge beyond the “tribe.” There are two fundamental approaches:
- Record tribal knowledge in proper documentation – a dull and thankless but essential job.
- Leverage GenAI to build and extract the necessary knowledge to model an AI support.
Leveraging GenAI is very promising, considering the recent developments in this field. However, the costs (in time and human labor) to define and run a GenAI model that can understand how IT operations teams perform their tasks (either ticket-driven or ticketless) mean this approach is not practical yet.
Currently, Digitate is investigating GenAI capabilities where ignio can absorb an ARD (automation requirements document) and configure ignio use cases. But for now, there is no other path to transformation than hard work.
Win over the tribal “chiefs”
To extract knowledge from IT Operations, you must figure out how to convince the tribe to provide the necessary know-how. Any tribe (and the ITOps tribe is no different) has its leaders. I see it over and over at each customer.
Convincing those who perform the task to be automated is relatively easy. These knowledge workers are also the ones who are alienated by the repetitive nature of their work. When we were implementing ignio Event Management at a customer, their NOC (network operations center) team thanked us for eliminating the tedious tasks of monitoring screens 24×7 for anomalies.
The biggest resistance comes from IT Operations leadership. They have the most to lose; when their tribes “lose” knowledge, they lose their ability to exercise their leadership. And hence they lose their status.
It is critical to invest relevant effort not only to explain how AI transformations are done but also –especially – what value can be created by AI transformations. IT Operations leaders need to be convinced that they are creating a bigger value for their direct customer. That way, losing current control is not a total loss, because providing additional value will grant them to be perceived as better leaders. Their status will increase, not decrease.
The bottom line: Tribal knowledge in IT is not an impediment if the “tribe’s” leader can be convinced of the value of AI transformations.
My practical advice to anyone who is starting an AI transformational journey: Focus on how to gain IT operations leadership buy-in. As of today, AI works; adoption is the challenge.
In a future blog I will describe how to focus on middle management to expedite AI transformation and how GenAI can help.