In this article, I want to share some observations and comments on the use of LLMs, which have now become an indispensable technology for people.
Have you ever looked at the messages you receive in your daily life, the documents or presentations sent to you, and thought, 'This looks like it was written by AI'? I experience this often. In fact, I no longer say 'looks like' and can clearly tell that I'm looking at a text generated by artificial intelligence.
Sometimes I look at WhatsApp groups. Someone asks a question, someone else answers. When I read the answer, I realize. That message wasn't written by a human. Sometimes even the question wasn't written by a human. Two AI models are talking behind human identities, and sometimes very absurd dialogues emerge. Maybe they don't notice, but I do. Most people now realize it.
Everyone is using AI and should use it. This is a fact, and there's nothing wrong with it. Some use it like a search engine to quickly access data; others have documents, presentations, visuals prepared and present them as if they were their own production. Of course, if we have access to a powerful service like AI, we should use it. The problem is not using it. The problem is taking the raw output as it is.
Let's recall how we used to prepare a document in the past. We would write every word one by one. Your focus would inevitably be on that task. You would get stuck and spend time, think as you write, write as you think. That process was laborious, but it made you master the content.
Now we have AI and think, 'I can handle it in 2 minutes.' Preparing something entirely from scratch with our own effort and ideas seems impossible when we have such power at our fingertips. But the real issue lies here. If we want to produce something valuable, we have to spend time on it. No matter how advanced AI models become, no matter how professional the product they produce, if your fingerprint is not on the result, someone else can do it for you. Even an AI model can do it without any human involvement. Because you're not making a difference and not adding your originality. The product you produce is not valuable. If someone else gave the same prompt, the same thing would come out. The person who makes a difference is not the one who takes the raw output, but the one who guides, filters, and makes the final decision.
So, what should we do if we really want to add value and embed our originality into the output? In this fast-paced world, being late is a much worse and unacceptable phenomenon than presenting raw AI output. Of course, we will use AI; it's now a necessity like electricity and water. What really needs to be done is to evaluate and transform AI outputs while maintaining the dedication and focus we had when doing it without AI. From the outside, this might seem easy to do. However, in my opinion, doing this is much harder and requires much more discipline than preparing without AI.
Now more focus is required. You need to read what AI has produced line by line, find out what's missing, what's wrong, what's not yours. Take the AI output as a template and make additions, updates; realize the direction of the process; check if the product is in line with what we want and repeat this process in a loop format. When this loop ends should not depend on your patience but on your satisfaction with the output.
Do you notice the power and difficulty of progressing in this workflow? Each time, you need to evaluate the product from start to finish with the same open mind, proceed with a solid foundation knowing your goal, and maintain this direction in every iteration. We also have to decide where to stop and when the refinement process will end. This requires incredible discipline and focus. In a world where everything is abstracted and our focus time is threatened to decrease every day, maintaining and improving this workflow might be the most important skill to acquire.
So, in essence, AI did not make the job easier. It increased the need for discipline.
As a software developer, this cycle works like this for me: First, I understand the task, plan in my mind what to change, how to change it, what to add. Then I leave the implementation to AI according to the specifications I have determined and examine the result in detail. Even if the system works, I need to dive into it to understand if it really works as I want. I always make the final decision. AI offers suggestions, but in a healthy product, the decision-making authority should not be with AI. It is a tool, not a magic box.
At Baksoft, we work with this mindset. We don't just deliver code. We understand, plan, and take ownership of the system. Ultimately, our main criterion is not whether the job is done. Did we really understand what you wanted? Could we produce a digital solution suitable for the problem? Is the product we offer valuable? We never finish the work without answering yes to these 3 questions. While writing this article, I used exactly the method I described and stopped at a certain point early. Could you guess which parts of the article were written by AI and which parts were written by me?
