After extensive experience with various tools, one truth becomes clear: no artificial intelligence can replace the nuanced judgment of seasoned professionals who have learned through experience, encountered real mistakes, and can discern between superficial solutions and genuinely effective ones.
This insight stems from my daily interactions with multiple AIs, which I personally invest in to understand their capabilities, limitations, and biases. The more I engage with these systems, the more evident it is that while AI significantly accelerates processes, it also increases the need for critical oversight.
This past weekend, for instance, I dedicated several hours to building a website using various artificial intelligences. Initially, I was impressed by the speed, apparent reliability, prompt responses, and seemingly coherent code. However, as I dove deeper, a familiar phenomenon appeared for those of us accustomed to tackling intricate challenges: AI often presumes it possesses more knowledge than it does.
It tends to offer authoritative recommendations even when incorrect. It can resolve one issue only to create another. Though its solutions may seem coherent, they frequently lack a comprehensive understanding of the problem at hand. Ultimately, I found myself dismantling much of what had been generated and reverting to a more traditional, manual, albeit slower, and ultimately more dependable method.
This experience encapsulates the present state of AI.
Artificial intelligence does not replace experienced professionals; rather, it enhances their capabilities. Those with sound judgment know when to accept suggestions, when to dismiss them, and when to recognize that behind an appealing answer may hide an incomplete perspective. Hence, it’s apt to view AI not as a substitute but as an amplifier: it strengthens the skilled professional while still requiring human oversight.
Moreover, an often-overlooked aspect is that no AI is devoid of biases. Each model carries its own inherent biases, cultural leanings, and training methodologies that inform their responses. Selecting an AI tool goes beyond mere power; it involves understanding the worldview it represents, the risks involved, and the tasks for which it is best suited.
In a professional environment, the key distinction lies not in who utilizes AI the most, but in who understands when to trust it… and when not to.
Artificial intelligence does not eliminate judgment; instead, it puts it to the ultimate test.
Hopefully, this discussion has been enlightening. My goal is not just to provide names but to ensure that when someone asks, “Which one should I choose?” you have more substantive insights to share than simply the most popular option.
If Tomorrow You Sit Down at Your Computer and Must Choose One
We’ve come to the conclusion. The most valuable takeaway I can offer is also the most straightforward: stop searching for the best AI. Instead, focus on determining which one will best meet your needs tomorrow.
If your work primarily involves writing emails and you need quick assistance, Copilot will likely be sufficient if you use Office, or Gemini for Google Workspace. There’s no need for a more sophisticated tool in that case.
If your tasks involve detailed research and require answers supported by reliable sources, Perplexity will significantly enhance your experience.
For those working extensively with their own documents, whether reports, PDFs, contracts, or presentations, NotebookLM represents a powerful workflow enhancement available today.
If you need a versatile assistant for thinking, writing, coding, or analysis without the hassle of selecting between options, ChatGPT remains the most balanced choice.
If writing quality is your priority, Claude continues to be the preferred option among daily wordsmiths.
Should you wish to explore an AI that acts rather than merely responds, Genspark deserves far more attention than it currently garners.
How This Series of Articles Was Crafted
I believe transparency is crucial, especially in a discussion about artificial intelligence.
I did not compose this text solely on my own. Instead, I collaborated with several AIs extensively. I utilized ChatGPT for structuring ideas and exploring varied approaches. Gemini was employed to verify information and work with reference materials. Claude was invaluable for reviewing phrasing, refining tone, and enhancing coherence between sections. NotebookLM helped me analyze and cross-reference the sources and prior materials I had gathered. Lastly, I engaged Mistral at specific junctures when I desired a more neutral perspective or a second opinion on certain paragraphs.
This involved numerous iterations; I didn’t write this piece in just one sitting. It was a cyclical process of proposing, reviewing, discarding, rewriting, and cross-checking with other tools. At times I completely reoriented my approach after seeing how another tool had framed it. In other instances, I dismissed suggestions from an AI if they sounded right but didn’t resonate with my voice.
Over time, I’ve developed a method I call the Stainless Iteration Method—an approach that integrates a final self-assessment from any AI I use. Following the completion of a draft, I ask the AI to evaluate itself on a scale from 1 to 10 and to explain why it didn’t achieve a perfect score. This self-assessment encourages the model to identify its weaknesses: where arguments fall short, where nuances are absent, or which paragraphs feel generic. I then present the same draft, along with the assessment, to another AI for evaluation. The second tool does not start from scratch; rather, it builds upon the previous assessment to confirm, contest, or refine the insights.
This approach helps mitigate the blind spots of each model. No AI truly understands what it lacks. However, by cross-referencing evaluations, one’s blind spots often become apparent to another. Essentially, this applies the same logic a seasoned editor would use, requesting a second read-through before publication.
At every stage, I carefully reviewed each paragraph, determining what to keep and what to discard, thus taking accountability for the text’s message. The ideas, opinions, personal anecdotes—such as my weekend spent building the website—and the overall structure are fundamentally mine. The tools provided speed, rapid iteration, and a reflective surface to challenge my thought processes.
This encapsulates my main argument throughout the article: AI does not write for you. However, if used effectively, it can enhance your capabilities exponentially. The text you just read serves as a practical illustration of that potential.
