The E² Pattern in AI and Why It Matters

When the era of modern computing began nearly 80 years ago, software code was naturally very simple, dominated by a number of IF THEN ELSE looking statements interwoven with algebraic operators.  Simple, but very powerful when these instructions were performed rapidly and recursively.  It wasn’t long before the new science of Software Engineering emerged where practitioners of code development sought to build structures that emphasized computing efficiency and ease of maintenance.  These structures were called patterns, and were largely responsible for the advancement of computing from technological curiosity to the engine of almost everything meaningful in our lives, from sending men to the moon to advanced manufacturing to drug discovery.

It has become clear to us that AI is poised to follow a similar path. Right now we are clumsy cave men (and women) typing vacation plans into our favorite AI platform.  But the next wave of advancement of AI is when we start to use common structure to amplify the effectiveness of AI beyond its “armchair” use to write that next email to your boss.  Think of it like this: it was a major advance in human progress to figure out how to melt iron ore with other materials to make steel.  It was another and even greater step to master the science of casting that steel in molds to make parts for machines.  I believe we are at the doorstep of that next advance in AI.

The reason I believe this is from our own direct experience in performing projects for the companies we work with.  A very common request is for us to take some task that a human or group of humans do and create a parallel, automated equivalent of that task.  A contrived example: A human production planner for a company that makes furniture.  That person gathers a variety of data, puts it into a spreadsheet, conducts analysis, and generates a production plan for the next month—how many products to make of what type to meet customer demand.  This reveals two primary components: the human who is the expert in planning, and the spreadsheet that houses a model that the human uses to create the plan.  We call this the Expert/Expertise pattern or E² (“E squared”).

From an AI standpoint, this is also two parts.  The underlying model is built from a diagram that expresses the logic in the spreadsheet.  When we work on projects like this, we create the diagram and ask AI to build the code in the model for us.  The thoroughly tested and validated model now runs as a software application.  But we don’t stop there.  We then ask AI to become an expert in using the model to accomplish goals—goals like maximizing the profitability of the furniture factory while making the most efficient use of the wood and metal materials on hand.  This then allows the human planner to interact with this newly born expert as if it were a very smart, hard-working peer in production planning, while relieving the human from the mechanical burden of operating the underlying model.  This in turn frees the human expert to ask her expert peer all sorts of strategic questions about production planning for the next month, the next year, across disparate factories, and even two years from now when that pending merger is expected to go through.  The result: a much more powerful, productive human resource.

Again and again we are seeing candidates for the E² pattern as we “tour” companies and witness their operations, no matter what industry they may belong to.  We are also seeing AI advance in ways that allow it to serve the “expert” role quite effectively, especially when we developers do the clever work of informing the AI on the knowledge it needs to become an expert (this, we have discovered, is a whole skill set unto itself).

So the next time someone in your company asks, “hey we know AI is powerful, we just don’t know the best place to apply it”, consider the E² pattern and where that exists around the firm: those use cases where skilled humans are leveraging a model of some form to accomplish a value-creating task.  Do this and you might just find yourself riding the tailwind of AI instead of watching all of your competitors from behind.  You might also find more time to take that vacation that your AI just planned for you.