Executive Action Plan for AI Network Evolution
It is daunting to be an executive today trying to guide technical team leadership in developing responses to the changes being driven by AI evolution. Below is a high-level analysis of the current
situation and key guiding principles that can make this job manageable.
We used to build and optimize networks based on what people wanted, needed, and how organizations sought to influence those things directly. We did that by collecting data and by combining the
data with algorithms to help us create optimal architectures. GenAI has changed things so dramatically that the data we used to collect may not be useful anymore.
If you are collecting the same data you were four years ago, you may be basing decisions on a wrong foundation. At the same time, differences between pure communication, processing, and storage
are becoming blurred. This leads us into a period of ongoing dramatic changes. Changes we have to adapt to. Because of ongoing AI evolution, we must adapt not just once, but rather in a stream of
adaptations.
As an executive, your most effective response is to help your team continuously:
- Carefully monitor what data they collect and how they analyze it;
- Track the drivers that shape the immediate environment;
- Make the architectural adaptations necessary to meet the immediate environment while also being able to easily evolve to the next adaptation.
Background
It is an old saying that “you get what you measure.” So, measuring the right things is critically important. Approximately 100 years ago, we started collecting more data than we could deal with manually. This led to the punch card sorting of the US Census data. Punch cards led to the early mainframe computers. Through the transitions from mainframes to minicomputers to PC’s, to phones - from circuit-switched twisted pairs to TCP/IP, fiber, wireless, etc. we perfected what data to gather. The means for storing and analyzing the data changed. But the underlying worldview centered on human users drove what data to collect.Now, with the advent of GenAI, this is changing. The data you were capturing four years ago may not be meaningful today. At the same time, how and where AI processing is being done is blurring the lines between communication, processing, and storage. What’s worse is that things will continue to be in such a state of flux that the corrections you make today may not be viable in another year or two.
Data Example
Four years ago it was reasonable to count the number of clicks to determine the effectiveness of an advertising or PR campaign. It was assumed that each click was a measure of human eyeballs reached and affected. To make sure that these numbers stayed reliable, organizations tried to prevent bots (built with conventional code) from being able to click. The intent was to keep the data ‘pure’.Many organizations are still counting the same way now and still analyzing the resulting data in the same way. However, GenAI has fundamentally changed things.
Now, leading-edge campaigns are seeking out bot access. They are making special provisions for Intelligent Agents to have access to content that is not visible to humans. The intent is to get the Intelligent Agents to put key messages into AI training data sets. Once in the set, the messages are trained into AI’s.
