AI Doesn’t Make You Good At Your Job

So, I’m a guy who happens to have the paid versions of Claude, ChatGPT and Gemini at the same time for different use cases from image editing to brainstorming and coding. I happen to love the tech but dislike the way large corporations think they can replace people with what is essentially a very spacey employee that needs to be called out. With that said, I think I’m a little more qualified to write about how these tools can shape our work because I’ve used the major companies’ tools for work and large projects.

First thing’s first, lazy people make lazy output.

In 2023, a lawyer submitted a legal brief citing several court cases that didn’t exist. ChatGPT had fabricated them. He hadn’t read the brief before filing it. He faced sanctions and came close to losing his license.

For some it might be an example of how AI is making people dumber. I view it as an example of why I would never choose that lawyer for trying to offload his legal work to a tool that, at the time, was known for making mistakes and hallucinating. It was a terrible judgement.

I’m sorry but, that guy was lazy and an idiot. He was both of those things before he touched the AI.

On AI caused laziness

People who call AI users lazy are usually judging something that genuinely is lazy: some crap AI generated content that the person behind added nothing of their own quality checks, thought or input. Anyone who has spent time on LinkedIn over the last two years has likely seen it.

That’s a problem with how a specific person is using a tool, not with the tool itself. Getting something usable out of AI requires knowing and figuring out what you want, evaluating what you got, pushing back when it misses, and repeating that several times before anything is ready. The people doing that aren’t cutting corners and are iterating.

Hiring managers are already making this distinction. The difference between someone who uses AI to think faster and someone who uses it to avoid thinking tends to show up in the work itself.

Professional insight makes professional work.

When I started building my own web project using Claude code, I didn’t have a formal design background. What I had was my previous experience designing apps for personal projects. When something looked generic, I could say why and when a layout didn’t serve the actual user. I could name what was missing. That knowledge is what made the AI create EXACTLY what I wanted. Without it, “build me a nice website” produces exactly that: something that looks like a generic website, optimized for nothing in particular.

This is true in marketing, design, development, basically anything that is beginning to utilize AI. The amount of times this tool goes on a tangent and I have to say, “that’s not right” is regular and expected. Knowing when to adjust and having the experience you bring is what separates something usable from slop.

The results come from asking questions.

The most useful thing I’ve done with AI is not ask it to produce things. It’s ask it to help me think through things before I do.

When I was developing a product concept, I didn’t open with “build this.” I started with questions: what needs to exist before anything else can work, what’s the simplest possible version, which decision depends on another one being made first, what would make this too complicated to actually use. Sometimes the responses were useful. Sometimes they missed the point entirely and I had to explain why, which often clarified my own thinking more than the response did. That back and forth is where the actual work happened. The document at the end was just a record of conclusions I had already reached.

Most people skip that part. They go straight to asking for output, get something 60 percent right, and either use it anyway or feel stuck because it doesn’t quite fit. Going in earlier, while the thinking is still loose, is where AI actually earns its place in the process.

You still have to check it (duh)

The lawyer story from 2023 is an extreme case, but the pattern behind it is common. Someone asks for something, gets a response that looks confident and complete, and moves forward without checking it. In marketing that might mean a statistic that sounds plausible but is invented, a claim that contradicts something else in the same document, or a tone that’s completely wrong for the audience.

AI produces fluent output but fluency has nothing to do with quality information. You can say a lot of eloquent words that at the end of the day mean nothing. No amount of “make this more complete” or “you are an expert writer” prompting removes that.

Quantity of work is not quality work

I think this is the worst part, AI has encouraged mediocre workers to look at the output of their first request into a complex task and think, “Wow, I’ve done so much work!”. It could be riddled with errors, irrelevant information, the wrong context and they’ll walk away feeling like they did something grand when they just made some slop in large quantity. Great, you made some content; it’s garbage, but you made content I guess.

What quality actually looks like

Knowing your craft well enough to direct the process is what makes any of this work. AI will confidently generate whatever fits the pattern of its training, good direction or not. The ceiling is always the person running it.

That yes man quality is actually the biggest tell. These tools will agree with bad ideas, elaborate on flawed premises, and produce eloquent garbage on command without hesitation. The only check on any of that is the operator. So if you’re consistently getting something usable out of the process, it’s because you knew what usable looked like before you started.