Introduced with Fanfare, But Ignored on the Frontlines?
With the promise of “dramatically streamlining daily tasks,” many companies are actively rolling out generative AI tools like Microsoft 365 Copilot. However, are you hearing these kinds of comments from your teams on the ground?
“In the end, it’s faster to just do it myself.” “No matter how many times I instruct it, I don’t get the results I want. It’s so frustrating.”
Actually, this is the “wall of generative AI” that almost every company faces in the early stages of adoption. In particular, highly capable business professionals who are already masters of Excel and PowerPoint tend to harbor the strongest frustrations.
Why does AI, which is supposed to be a powerful weapon for operational efficiency, become a “nuisance” in the workplace? In this article, we will unravel the root causes and share the mindset needed to turn AI into a true ally.
The Reality: Copilot is Unnecessary for “Operational Pros”
First and foremost, Copilot’s true strength lies in supporting “beginners and those forced into inefficient workflows.”
For an “Excel master” who knows exactly how to build complex functions, or an “email pro” who can pull up a template in one second from past records, the time spent typing text instructions (prompts) to tell the AI “do this, do that” is a complete waste of time.
If a task takes you one minute to finish intuitively using keyboard shortcuts, but takes three minutes to explain to an AI—plus the time to check the output for errors—it is completely natural to feel that “doing it manually is overwhelmingly faster.”
The Biggest Hurdle: The Curse of the “100% Perfect Score”
So, why do we feel so much stress toward AI outputs? The fundamental reason is that we are demanding a “100% perfect final product” from the AI.
Generative AI is a “probabilistic language predictor” that guesses the “next most likely word” from massive amounts of data. It cannot perfectly grasp context or unspoken rules. Therefore, when you try to force the AI to output the “100% correct answer” exactly as it exists in your head, the following tragedies occur:
1. The Endless “Swamp of Prompt Adjustments”
“Make the tone a bit more polite,” “Avoid this specific phrase,” “Change the logical flow here.” To get closer to your ideal result, you find yourself rewriting instructions (prompts) over and over again. Before you know it, you have spent more effort and time instructing the AI than you would have spent writing the document yourself from scratch.
2. Unintended Degradation Due to AI “Randomness”
What is even more troublesome is what happens when you tweak the instructions and regenerate. Because AI chooses words based on probability, it is common to see that “the part that was fine just a moment ago has suddenly changed into an unnatural expression.” You adjust it to reach an 80% quality mark, but the next move drops it back down to 60%. This feeling of “taking one step forward and two steps back” breeds the intense frustration of feeling like “this task will never be finished.”
The Real Meaning of “Doing It Manually is Faster”
Looking at it this way, the true meaning behind the words “it’s faster to do it myself” spoken by frontline professionals becomes clear.
It is not just about typing speed. It is the highly rational and harsh reality that “the mental stress and time spent on endless trial-and-error (prompt adjustments) to make the AI score 100 points overwhelmingly exceeds the cost of creating a 100-point product from scratch by yourself.”
If your expectation is a “highly capable secretary who works flawlessly,” you will suffer from this gap. In reality, AI is closer to a “new employee who works incredibly fast but cannot read the room.”
A Prescription to Make Generative AI a Truly “Usable” Tool
How can we break through this situation and maximize the benefits of generative AI like Microsoft 365 Copilot? The answer lies not in changing how we use the tool, but in changing “how we approach our work.”
1. Accept It as a “60% First Draft”
Stop demanding perfection from AI. The greatest value of AI is “lowering the psychological barrier when creating something from scratch.” Let the AI handle the first few minutes of staring at a blank Word document or PowerPoint slide, and let a human refine the resulting “60% draft” to 100%. This division of labor is the most stress-free way to boost productivity.
2. Delegate “Information Processing,” Not “Operations”
Handle “operations” like Excel formatting and PowerPoint shape adjustments manually, and delegate “extracting meaning from information” to the AI.
- Have it summarize only the final decisions and your specific action items from a long, recorded Teams meeting.
- Have it pick out only the necessary information from a massive volume of internal regulations or past project proposals.
This kind of “information processing that physically takes time for a human to read” is the area where AI completely outperforms manual work.
Conclusion: Update Your “Way of Working” to Match the AI
The feeling that “I installed Copilot, but manual work is faster” is not wrong at all. It is a crucial sensor for identifying what AI is good at and what it is bad at.
To make generative AI a true business partner, you need the clear-cut mindset that “tasks that are faster to do manually shouldn’t be forced onto AI,” and the expectation management of “not seeking perfection, but using it as a stepping stone.” Only when we can update our very way of working to match the tool will its true value be unlocked.