Most days, the AI and LLM world is moving so fast that it’s like trying to catch a cheetah with a Frog!¹ New models, startups, and tools are popping up constantly, and it takes much work to keep up. But amidst the chaos, clear trends are starting to emerge.

I’ve noticed a distinct shift based on my experience in the AI trenches — implementing, advising, and witnessing AI projects firsthand. AI isn’t just a futuristic dream; it’s a readily available game-changer that transforms processes and jobs.

But here’s the kicker: no “one size fits all” exists in AI. I’ve seen at least four distinct approaches to AI, and understanding these is key to navigating this exciting landscape.

Let’s break it down, shall we?

1. The Citizen Prompter

Many organizations tried to use data meshes and lakes to put data into the hands of the “citizen data scientist.” While some organizations got quite far, especially where low-code tools were readily available, successes on this front have been rare.

Chat interfaces and large language models are rapidly changing the picture. I see more and more project managers and sales directives using ChatGPT to learn enough Python to dig into the actual data and use the LLM tools to move from low-code to prototyping. These “citizen prompters” use AI tools to supercharge their productivity, automate those tedious tasks we all hate, and unleash their creative potential.

The key here is training and experimentation. Tools like CoPilot and Gemini are potent but not magic wands. You need to invest the time to learn how to use them effectively.

2. Modern work on steroids

Where we see advanced low-code and scripting capabilities, a discipline Microsoft has aptly dubbed “Modern Work [Consultancy]”² appears. AI is revolutionizing how we work within organizations. It’s not just about automating simple tasks anymore; it’s about using AI to create genuinely “smart” workflows and solutions.  

The real game-changer for modern work is connecting the copilots with your organization’s specific data and systems. Custom Copilots and Gems³ can be trained on your data, allowing them to perform complex tasks and build robust “low-code” solutions. This means organizations can develop AI-powered solutions faster and cheaper than ever, automating workflows, personalizing experiences, and enabling data-driven decision-making.  

As the tooling progresses, I imagine more and more of the work we now do as one-off AI projects to start moving over to this category⁴.

3. AI-Powered Solutions

As any Modern Work -consultant knows, the process automation tools only take you so far. At some point, we realize we need a full-fledged solution to solve the issue. AI solutions are similar to traditional solutions in this regard. Building custom solutions offers flexibility and control but requires ongoing maintenance to avoid “dead code.”  

The money pouring into anything, even hinting at AI, changes the picture⁵. Many AI startups provide innovative solutions to anything from legal analysis to correctly watering your plants. This puts us in a curious spot. When a more robust solution is needed, we can suddenly swap the risk of dead code for the risk of a dead vendor.

4. AIOps

While I used to give a keynote speech on the silliest acronyms created by altering the DevOps acronym, DataOps, and its more recent variation, AIOps, do make sense.

To truly harness the power of your data and the capabilities of generative AI, you need to have a platform team or three constantly honing your capabilities, monitoring performance, identifying areas for improvement, implementing changes, and ensuring ethical use.  

Building this kind of “native” AI capability will allow your organization to stay ahead of the curve - if you can pay the price.

The Journey from here?

The AI landscape is constantly evolving. Understanding the different approaches to AI and embracing the need for continuous learning and adaptation will help us harness the power of AI to shape a better future.

While these four categories are preliminary and early, they paint a picture of where AI is going—if not for the Big Tech building the capabilities, then at least for the rest of us.