Collaboration with AI
Why AI Output is a Function of One's Judgment
AI (GPTs) are one of the most powerful tools of our age, one that will soon be smarter than the intelligence of all of the human race combined. But most of us use these tools like we used to use Google. Ask a question and expect an answer. Turns out, Google was just finding the right link and information, but GPTs are more nuanced in that sense. Just imagine, a large population is seeking companionship with AI tools. As bizarre as it may sound, it’s a reality of our time.
How many times does it happen that we are not fully satisfied with the response we get? The answer is non-zero. There are reasons why we don’t get what we are seeking.
“Judgement” Angle
AI requires the user to bridge the context gap. The quality of the result depends on two variables:
The quality of context provided.
The quality of the prompt.
Both are driven by judgment. Judgement is not a setting; it is the culmination of a lifetime of experiences and unique environments that an individual has been through. Consequently, the quality of any AI output is strictly a function of human judgment.
Fair to say, human judgment is the limiting factor.
Laziness
There is a fundamental friction in AI usage: high-quality output demands exhaustive context, yet humans are naturally averse to writing thorough prompts. Just recall, how many times have you written a 2-3 paragraph to an AI? Not many times, right? We are used to using GPT like we would use Google. Old habits die hard.
This “laziness” often prevents the AI from reaching its full potential, leaving the user with a generic result that lacks the necessary depth.
2-Step AI Collaboration Workflow
To ensure context-aware output, work must be planned with two distinct time allocations:
The AI Collaboration Session: An iterative phase of chatting and generating raw material, usually in ChatGPT or Grok.
The Scrutiny Phase: Focused, “GPT-less” time to tailor and refine the output outside the chat interface.
Skipping the second step is a big mistake which most make. Without rigorous scrutiny, you risk presenting work that misses the taste and specificity of your unique context.
AI can help you work faster, but only your scrutiny can help you.
How do I review work presented to me?
When reviewing the work of others, the goal shouldn’t be to play “AI Detective.” Whether a tool was used is irrelevant; I often encourage it like a broken record. The focus should remain on the same standard used in a pre-AI world: How much was the deviation from the expectation?
The true metric of progress is the volume of revisions needed. Over time, an individual’s work should require fewer and fewer reviewer inputs as their own judgment improves. However, that number will never be zero. Even with the most advanced AI, the final layer of human scrutiny remains the indispensable barrier between an AI-draft and professional excellence.
While reviewing, always judge the result, not the tool used to get there.
Having said that, what I often see these days is people presenting AI outputs as is.