Introduction

I’ve been tasked to start creating content for our new AI ecosystem. As part of any new challenge, a person should take a couple of moments and try to understand the lay of the land. Part of this process (for me, at least) is trying to understand the personas within the community that I’m about to engage with, which can be frustrating and confusing. Here are the three groups I’ve identified to help simplify messaging and audiences for AI upstream content.

As you read this, realize we are living in a gold rush of AI content, but I’ve noticed a ton of mismatched content out there, so hopefully, looking through these lenses, you’ll see the boundaries and build content more targeted to who you are trying to reach.

The three groups of people I have named are “Users,” “Builders,” and “Designers” of AI. Designers are a subset of the Builders group, and Builders is a subset of the Users group. Imagine a pyramid with Users at the bottom, then the Builders, and finally Designers at the top, or the pyramid’s capstone.

Users

To start with, “Users” are the people who assume AI is like ChatGPT and want to “talk to it.” They are the people who are learning to engage with AI for “daily tasks” and spend time with the “chat system,” asking questions and building prompts. These people are the new “Prompt Engineers” who learn to ask the correct questions and build the queries for the AI to crunch the data it has access to and bring out the designed answer or content. These people want to use AI as Google’s “next generation” for different targeted data sets or engage AI to parse the internet. However, their goal is to help in conversational knowledge gathering, leveraging AI as an expert to help answer their questions.

These engineers have to understand and build some solid Prompt Engineering skills, more so than the “zero prompt” back and forth of a chatbot; they also need some skills in basic API manipulation and some level of scripting with something like Python to build some automation around engaging with the AI.

Builders

The first subset of the “Users” group is the “Builders.” They are the first ones attempting to build upon the AI models and ecosystem to give “Users” a more delightful experience. This can range from taking an upstream model and creating something like a RAG (Retrieval Augmentation Generation), a dedicated “library” of information, to writing applications on top of Models to make data pipelines more useful. They also leverage their Prompt Engineering skills to crunch through business data and logic to help drive business decisions.

This is different from just simple Prompt Engineering because of the scope of the data these builders can access. Most of the content from an engineering standpoint at the writing of this article appears to target this user group. While from a high level, it makes sense (this should be the largest group of people), in practice, it is lacking. The core challenge with the group is that they encompass more than just the AI space. If you build content that teaches something like PyTorch for this audience, it will go on deaf ears or confuse the group, but at the same time, something as “standard” as Prompt Engineering would be too beginner for them. The correct narrative here is building a “trusted AI RAG” or a practical hands-on application they can build off of, not the abstract aspects of something like Prompt Engineering or building Models from dedicated datasets. This group of people are already experts in Prompt Engineering, software design, web technology development, and some multifunctional languages like Python.

Designers

The final subset of the “Builders” is the “Designers” group. They are the ones who have built upon everything below them to design the best Models and tooling to push AI to the next level. These are the smallest group of people, the “capstone” of this pyramid. They are the hardest to find, and they have spent significant time learning the skills required to be effective in this position. They are the ones that take clean datasets, build Models or tune Models off of these datasets, and test for the outputs. They are Data Scientists who learn and understand how the Models are created and why the pipelines to test these Models exist. They can comfortably program in any language required for AI work, from Python to C++, and leverage things like PyTorch to design the AI and AI ecosystem to complete their required work.

Conclusion

Trying to figure out how to talk to these different groups of people is critical to getting adoption for whatever you’re attempting to show. Like all other communities, some people fit in with each other, some cross the boundaries, and some try to figure out where they want to fit. This can be frustrating when building content, but identifying these people will give you a better conversation and, ideally, better content so everyone will enjoy your presentations more.