80% Taken by AI, the Last 20% Held Only by Humans

80% Taken by AI, the Last 20% Held Only by Humans

In an Era Where AI Handles 80% of Work, the "Last 20%" Humans Should Hone

Will AI Take Jobs or Make Human Work Easier?

This question is no longer just a topic for the tech industry. Writers, document creators, salespeople, legal professionals, engineers, marketers, and executives—anyone working in front of a screen has likely pondered this at least once.

"How long will my job be needed?"

An article from Fast Company, published in Inc., provides a very clear perspective on this anxiety. AI might handle the first 80% of the work, but the real value lies in the remaining 20%.

The 80% mentioned here refers to tasks like research, organization, summarization, drafting, classification, comparison, and creating outlines—tasks that are reproducible and easy to proceduralize. They take time but have clear inputs and outputs, areas where AI excels. AI can finish the groundwork that used to take humans hours in just minutes.

On the other hand, the last 20% is different. It involves reading context, making judgments based on experience, taking risks, building trust with clients, and making responsible decisions. Is it okay to apply AI's answers directly to reality? To whom, in what order, and with what words should it be communicated? Who takes responsibility if it fails? These questions cannot be resolved by mere information processing.


AI is Taking Over "Tasks," Not "Jobs"

One reason for the confusion in discussions about AI is that "jobs" and "tasks" are often spoken of as if they are the same.

Consider the work of a lawyer. Researching past cases, reading lengthy documents, organizing points, and drafting documents are important tasks, but they are not the entirety of the value clients seek from lawyers. What clients truly want is to find a winning strategy, create favorable negotiation conditions, assess risks, and ultimately have the expertise to say, "Let's proceed with this decision."

AI can significantly shorten these initial tasks. However, it is humans who decide which arguments to choose, how far to go, and where to draw the line within legal, social, and emotional contexts.

This is not limited to law. The same applies to engineers. AI can write code, speculate on error causes, and suggest fixes. But when a production issue arises, it is humans who determine which customers are affected, whether recovery or root cause analysis should be prioritized, and what information should be communicated to management or customers.

The same goes for marketing. AI can generate dozens of ad copies and summarize market research. But who are the customers that the brand has made promises to? Is there a risk of backlash? Is this a situation where long-term trust should be prioritized over short-term click rates? Such judgments involve experience and responsibility.

In other words, AI is taking over the "processing" included in jobs. However, it is not taking away the meaning or responsibility of the job itself.


The Value of Speed Decreases, and the Value of Judgment Increases

In many workplaces, those who could produce quickly have been valued. Quickly summarizing documents, writing emails, coding, or producing meeting minutes. Of course, speed will continue to be important.

However, as AI becomes widespread, it will be harder to differentiate based solely on simple work speed. This is because anyone will be able to produce a certain level of drafts or analyses quickly.

The value will lie in those who can decide "what should be created."

The way questions are posed will become important. If you throw a vague question at AI, you'll get a vague answer. If the premise is wrong, AI will produce a superficially polished mistake based on that premise. AI's danger lies in how even wrong answers can appear very natural and convincing.

Therefore, in future work, more than just "can you use AI," the questions will be "what should AI be tasked with," "how much of AI's output can be trusted," and "how to connect it to real-world problems."

To add value in the remaining 20% in an era where AI handles 80%, not only expertise but also problem discovery skills, editing skills, ethics, interpersonal understanding, and decision-making skills will be necessary.


Expectations and Cautions Mix on Social Media

Looking at the reactions on social media to this article, it seems that rather than a large-scale controversy or viral explosion, there is a strong impression of quiet empathy among practitioners, especially on LinkedIn.

In Inc. Magazine's LinkedIn post, comments were seen with the sentiment that "AI can automate many tasks, but empathy, judgment, and understanding remain at the heart of meaningful innovation." Instead of viewing AI simply as a threat, there are voices taking it as an opportunity to reaffirm human-like values.

Additionally, in response to the discussion between Casey Newton of Platformer and Box CEO Aaron Levie, there was a reaction on LinkedIn that "whether AI takes jobs or creates new ones, a major transition period is coming." The important point here is that it's not just optimism. It's an understanding that rather than making work easier with AI, the very shape of work is changing.

In another LinkedIn discussion, the sentiment that "AI quickly provides the initial 80%, but the most effort is required for the final 20% of correction, verification, and style adjustment" was shared. This is something many AI users can relate to. AI is quick to create drafts. However, to confirm whether the content is truly correct, suitable for the reader, or aligned with one's intentions, human concentration is ultimately needed.

Furthermore, in comments on another post, there was a sentiment that "AI can extract data, but it cannot live with the responsibility for the results and actual decision-making." This hits the essence of the AI era. AI can make suggestions. However, it is humans who persuade customers with those suggestions, move organizations, and stand in the line of fire when failures occur.

Summarizing the reactions on social media, the points can be divided into three main categories.

First, the expectation that AI will reduce repetitive tasks and direct humans towards more creative work.

Second, the caution that over-reliance on AI's 80% output can lead to insufficient verification and unclear responsibility.

Third, the realistic view that in the AI era, what is needed is not just "how to use" AI, but also human judgment and relearning.

In other words, the discussions on social media align with the article's claims. AI is not making humans unnecessary but changing the place where human value resides.


The "Last 20%" Is Not Left for Everyone

However, we should not be too complacent here.

Hearing "If AI does 80%, humans only need to handle the last 20%" sounds optimistic. However, the gap between those who can handle that 20% and those who cannot may widen significantly in the future.

This is because experience is necessary for the last 20%.

Those who can judge where AI's output is shallow are people who have read, written, and revised many documents in the past. Those who can sense where AI's code is dangerous are people who have experienced dealing with failures and design mistakes in the past. Those who can notice when AI's market analysis is out of touch with reality are people who have actually faced customers.

In other words, the last 20% is not something anyone can easily do from the start. Rather, the experience cultivated through the steady 80% of work becomes the foundation for that judgment.

Herein lies the difficulty of the AI era. As AI takes over beginner tasks, there is a possibility that opportunities for young people to gain experience will decrease. Tasks like preliminary research, summarization, drafting, and revision, which seem mundane, were not just chores but also training for developing judgment.

Companies should not take this lightly. If they take away all foundational tasks from young employees just because they can be made efficient with AI, they will not develop personnel capable of handling the "last 20%" in the future. AI utilization and human resource development need to be designed together.


What Companies Need Is Not "AI Implementation" but "Responsibility Design"

Many companies believe that implementing AI will increase productivity. Indeed, in areas like document creation, summarization, and analysis, AI can significantly reduce work time. Studies also report examples where generative AI has led to time savings and quality improvements in writing tasks.

However, the true success of AI implementation is not determined by whether the tool was introduced. Who will verify AI's output? In which tasks is human approval mandatory? What verification will be conducted before presenting to clients? Who will take responsibility if errors occur?

Using AI without clarifying these points may increase the overall risk of the organization, even if surface-level productivity increases.

McKinsey's research also shows that organizations that succeed with AI emphasize how to incorporate human verification into model outputs. This is precisely the institutionalization of the "last 20%."

Rather than relying solely on individual efforts, design points where human judgment is incorporated as an organization. This becomes management in the AI era.


Five Skills Individuals Should Hone

So, what should individuals hone?

First, the ability to pose questions. AI is good at answering questions, but it is humans who decide what to ask in the first place. If the problem setting is shallow, the output will also be shallow.

Second, the ability to verify. It is necessary to not take AI's answers at face value, to check the basis, look for contradictions, and compare them against real-world constraints.

Third, the ability to read the context of a specialized field. The meaning of the same data changes depending on the industry, corporate culture, and customer circumstances. AI can provide generalities, but it cannot fully read unique contexts.

Fourth, the ability to build relationships. Customers and colleagues seek more than just correct information. Reassurance, trust, approachability, and a sense of responsibility are also important values.

Fifth, the ability to make decisions. Even when information is incomplete, there are situations where decisions must be made by a deadline. AI can list options, but it is humans who make the final choice.

These are not skills that can be acquired in a short period. They are built up through daily work, by failing, thinking, and correcting.


The Winners in the AI Era Are Not Those Faster Than AI

The idea of competing with AI has its limits. Summarizing faster than AI. Creating documents faster than AI. Writing code faster than AI. Such competition is disadvantageous for humans.

The place to compete is not there.

How to use what AI produces. What to add, what to remove, and what decisions to make. To whom, in what order, and with what words to deliver. That is where human value lies.

AI lowers the cost of tasks. This makes it possible to undertake a large number of projects, analyses, and prototypes that could not have been attempted before. However, the more possibilities increase, the more important it becomes to choose what to pursue. In a world with more options, the value of those who can make judgments rises.

Therefore, in the AI era, those who are truly at risk are not just those who cannot use AI. Those who use AI's answers without their own judgment are also at risk.

Conversely, those who can use AI as an excellent assistant while adding their own experience, perspective, and responsibility can demonstrate even greater value than before.

The era where AI handles 80% of work is not a time when 80% of human value is lost. It is a time when human value becomes more visible and more rigorously questioned.

The last 20% is not just finishing touches.

There lies the credibility of an expert.
There lies the experience accumulated on the ground.
There lies the responsibility that only humans can bear.

No matter how advanced AI becomes, humans who say "let's proceed with this decision" are necessary.
And those who can give weight to that one word will become irreplaceable in the AI era.


Source URL

Related Articles