I just read LinkedIn News Article “What sets AI ‘power users’ apart from casual fans” by Megan McDonough.
This topic is exactly what I write about so here are a few of my thoughts. The article concluded that “savvy AI power users” are doing the following things differently than the “casual ChatGPT fan”:
- Discipline, time, focus;
- Treating AI as “a second brain to multitask and problem-solve”
- Prompt refinement; and
- Giving feedback to the AI models
While I agree that use of AI can produce “more precise results” giving them (Power-Users) a “leg up at work”, there are a few additional points that would have been more helpful.
First, I’m not a fan of the “AI Power-User” label because it implies a hierarchy within AI users and can intimidate beginners. A better framing is that there are people more comfortable using AI, mostly because they have been using it longer. AI has the potential to be more of an democratizer than a discriminator. With such a staggering ability to quickly summarize vast amounts of information that used to only be the domain of Subject Matter Experts (SME) with decades of training, AI can teach you just about anything you want to know with the push of a few buttons.
Although, reading through the article comments it seems many people find it highly desirable to have the label of “AI Power User”. This is misguided. AI is a tool, nothing more. When using AI at work the goal that should be highly desirable is to be considered excellent at your job and be a valuable contributor, not be the office AI star.
I see this same misguided effort with Excel. Some folks do amazing things with Excel (coding, arrays, scripts, etc) and while it’s nice to have them as a resource, they are rarely company superstars. The goal was never to be great at Excel, it is to harness Excel (and other tools) to provide the highest impact to your work. Same with AI.
Second, this article offers very little actual help to people that are not already power users. Lucky for you, you found this website with all of the tools and advice freely available to help you become more of a power user!
Using words like: discipline; time, and focus sound meaningful but are of little help as you stare at the prompt window wondering what to do next. And, treating AI as a second brain is more vague framing without substance.
Better to think of AI as your personal assistant rather than another brain. AI will provide you with the information you need to think more deeply and make decisions faster. But AI should never replace your critical thinking. You are the brain in this relationship. The moment you start offloading your thinking to AI is the moment you start making yourself more easily replaced.
The most useful tip in the article is to give your AI more feedback. This is an area many can improve. Too often people treat AI like a person. They do not want to offend it, tell AI it is wrong or ask it to rerun a request for the fifth time with a small variation.
I am not saying to be mean to AI. I still use “please” and “thank you” in my prompts even though being polite no longer improves model performance. Unlike a human assistant that will fantasize about stabbing you in the parking lot after being asking to redo a presentation for the third time, your AI is happy to comply. The more critical input and correction you provide the better the output should be.
The last point on prompt engineering I find less relevant than it used to be. In the early commercial AI models prompt writing was very much an art and had a big effect on performance. But newer models are so much more advanced that how you write the prompt is far less important than it used to be.
A more important factor the article missed is model choice. Most mainstream AI models now allow you select between different models depending on the task. For example my current version of Copilot offers the following four models to choose from in the prompt window:
- Quick response – for every day conversation
- Think Deeper – for more complex topics
- Smart (GPT-5) – everyday or complex topics
- Deep Research – detailed reports
I find the choice of model to have far more impact on the results than how the prompt is written. For most work I use the Think Deeper model as the output is more thorough with better references than the faster models. Deep Research used to only be available with a higher payment tier but looks like it is now available with my basic Microsoft365 membership.
If you are just getting started with AI don’t get hung up on crafting the perfect prompt or the ideal model choice. Just provide clear instructions of what you want and your AI will likely figure it out. If the AI is off the mark, correct your prompt and ask again. If you find the output is not detailed enough or lacking depth then try a more advanced model. To help you get started below are some prompting tips and phrases.
SBAI Tip#6 Prompt Generation Help:
- Choose the appropriate model, if you do not know which model is best then try multiple models with a similar prompt get a feeling for the difference
- Don’t stress about writing the perfect prompt – just be as clear as you can
- Give your AI clear, critical feedback on the output
SBAI Tip#7 Some Prompt Phrases to Consider Using:
- “This output is good but X is not relevant. Eliminate X and rerun the previous analysis.”
- “Give me five different variations of Y”
- “Generate a bullet point list that can be copy/pasted into Word”
- “Are you sure Z is correct? Ensure you are taking A, B, C into account and rerun the analysis”
Leave a Reply