● AI Workflow Crisis

Conditions of a Leader Who Uses AI Well: In the Generative AI Era, If Leaders Do Not Change, the Organization’s Workflow Falls Apart

The core point of this piece is not “which AI tool to use.”

What really matters is whether leaders use AI like a magic wand, or whether they keep pounding away like a washboard to extract insights.

Based on the story of Kim Ji-hyun, Vice President of SK’s AI Committee, the decisive battleground for AI transformation is not automating individual work, but redesigning the organization’s entire workflow.

In particular, if leaders use generative AI incorrectly, team members lose opportunities to grow, and while the speed of work increases, the reporting system stays the same, which can intensify organizational burnout.

On the other hand, when used properly, it becomes a powerful productivity-boosting tool that lifts corporate competitiveness through market analysis, report writing, customer response analysis, and strategic decision-making.

1. Core News: A leader who uses AI well is not “someone who tells people to use AI,” but “someone who actually tries it and changes the organizational structure”

The leader types most commonly seen in business settings these days can be broadly divided into two categories.

• The first is a leader who barely uses AI personally but keeps shouting, “We need AX too.”
• The second is a leader who uses it a little and then pressures the organization by saying, “AI can do all this, so why can’t you?”

The first type is closer to a slogan without execution.

The second type is even more dangerous.

That is because they fail to distinguish between what AI can do and what people must do.

Such leaders do not use AI as a productivity-enhancing tool; they use it as a way to dump unreasonable homework on their team members.

The point Vice President Kim emphasized is clear.

AI can accelerate a company’s digital transformation, but if leaders do not use it themselves, they cannot discover the organization’s bottlenecks.

In other words, the reason leaders must use AI well is not simply because their own work becomes faster.

It is so they can judge where the company’s reporting system, meeting format, approval process, and collaboration structure need to change first.

2. What to assign to AI and what leaders must do themselves are different

The most important criterion for AI use is to “think of AI like a person.”

When assigning work to employees in a company, there are tasks you can fully delegate, tasks that need interim checks, and tasks where the leader must make the final judgment.

AI is the same.

Tasks leaders can delegate to AI

• Drafting meeting notes.
• Summarizing competitor reports.
• Collecting market trend data.
• Scraping global news and research materials.
• Suggesting a report draft structure.
• Creating a list of expected questions for presentations.
• Checking for logical gaps in documents.

These tasks are highly repetitive and require a lot of time for searching and organizing information.

With generative AI, leaders can reduce time spent on simple data collection and focus on more important judgments.

Tasks leaders must do themselves

• Final decision-making on business direction.
• Deciding the core message of a report.
• Organizing strategic implications for the CEO, chairman, and board.
• Judging organizational risks.
• Forming one’s own interpretation of market changes.
• Deciding which AI-generated suggestions to discard and which to adopt.

If you expect AI to create the core insight for you, the leader’s role disappears.

AI should not be something that gives the answer in your place; it should be a tool that strengthens the leader’s judgment.

3. You should not make employees use AI unconditionally

AI usage that is useful for leaders is not necessarily good for new hires or junior employees in the same way.

This point is more important than it may seem.

For example, drafting meeting notes may be a task that leaders can delegate to AI.

But for employees who are still learning the job, drafting meeting notes itself is important training.

That is because the process of recalling the discussion, organizing who said what and why, and understanding the context of decisions builds work muscles.

The same goes for market research.

It is convenient if AI summarizes competitor trends.

But if employees rely on AI summaries from the start, they lose the sense they gain from reviewing various articles, blogs, analyst reports, YouTube content, and communities.

Vice President Kim’s perspective can be summarized like this.

• You can delegate what you already do well to AI.
• For what you still cannot do, you should do it yourself and build the basics.
• Repetitive tasks can be reduced with AI.
• Tasks requiring judgment and contextual understanding should not be handed off too easily.

If AI is used incorrectly, employees’ productivity may seem to rise, but in the long run, the organization’s learning ability can weaken.

4. You should not use it like a magic wand, but like a washboard

Many people use AI in a “abracadabra” manner.

They ask one question, and if the answer is unsatisfying, they simply conclude, “AI is not that great.”

Vice President Kim sees this as a magic-wand style of AI use.

In contrast, a truly good leader uses AI like a washboard.

Rather than asking once and stopping, they keep pounding away, asking again, requesting opposing views, checking evidence, and reinterpreting from different angles.

The important thing is not just that the final output gets better.

It is that the leader’s thinking changes in the process.

Insights such as “it could be like this,” “there was a perspective I missed,” and “this market can be interpreted that way” begin to emerge.

A leader who uses AI well does not stop after asking two or three questions.

They keep debating within a single chat window for 10 turns, 30 turns, even 100 turns.

This process greatly accelerates understanding of complex topics such as global economic outlooks, industry analysis, and AI investment strategies.

5. AI is not a tool for saving time, but a tool for changing the order of work

The traditional way of writing reports was mostly linear.

• In week 1, you gather materials.
• In week 2, you structure the table of contents and message.
• In week 3, you review and revise the draft.
• In week 4, you refine the design and presentation materials.

But once AI is used, this order is completely disrupted.

During research, you can already build the presentation structure.

Before drafting, you can review the visualization direction first.

While designing the report, you may even change the core message again.

While analyzing the market, you can simultaneously create expected questions and test counterarguments.

This is the point where AI transformation differs from simple task automation.

AI does not merely make existing work a little faster; it changes the order of work and the way people collaborate.

That is why leaders must use AI themselves.

Only then can they see where their company’s reporting system and meeting structure are bottlenecks.

6. Vice President Kim Ji-hyun’s actual AI report-writing process

Vice President Kim says that most of her work is reporting and presentations.

The biggest difference before and after using AI is not just report-writing speed.

It is that her understanding of the market and the completeness of her messaging became much deeper during the report-making process.

Step 1: Fact-finding for 3 to 4 days with Perplexity

Once a topic is decided, she first uses Perplexity to research global articles, blogs, reports, YouTube content, and market materials.

She does not simply search once and stop.

She keeps asking questions for 3 to 4 days, 10 hours a day, to broadly verify the facts.

She keeps digging into opposing views, cases from other countries, industry-specific perspectives, and differences between companies.

Step 2: Verify the story and framework with ChatGPT

After reviewing enough material, she uses ChatGPT to verify the report’s storyline.

She checks whether the outline she first came up with is right, whether there are other perspectives, how the supplier perspective differs from the demand-side perspective, and how it looks from a government policy perspective.

In this process, the message she first had in mind may change.

The important thing here is not to use the structure AI created as-is, but to let her own thinking mature by debating with AI.

Step 3: Write directly for a full day without AI

Once things are reasonably organized in her head, she opens a notepad and writes directly.

At this stage, she intentionally does not use AI.

That is because the core structure and message of the report must be organized entirely in her own language.

Step 4: Generate a document draft with Genspark

She puts what she wrote directly into ChatGPT and organizes it into a markdown form that Genspark can easily understand.

Then she uses Genspark to create a draft in PPT or document form.

At this stage, a report draft of around 30 pages is quickly produced.

Step 5: Supplement visualization and structure with NotebookLM

She puts the generated document into NotebookLM and looks for improvements page by page.

She checks which pages need diagrams, which parts are better shown as graphs, and which messages can be conveyed more visually.

This stage is not just design; it is a process of improving the structure so the message can be delivered more clearly.

Step 6: Verify logic and facts with Gemini

Once about 80 percent of the version is complete, she puts the document into Gemini for verification.

She checks page by page for logical leaps, incorrect numbers, and weakly supported messages.

She does not accept everything Gemini points out as-is.

After searching and confirming directly, she reflects only the content that is correct.

Step 7: Refine sentences and presentation messages with Claude

When it reaches the 90 percent stage, she uses Claude to refine the title, subtitle, core keywords, and the points to emphasize during the presentation.

She does not change the content of the document; she adjusts the wording so it lands more sharply with the audience.

Step 8: Use Claude inside PowerPoint to improve final quality

Finally, inside PowerPoint, she calls on Claude again to recheck slide flow, whether pages should be removed, diagram improvements, and how to connect the context before and after.

From this stage on, there is no end.

You can keep pushing 1.0 to 1.5, to 2.0, to 3.0.

So the leader must judge when to stop.

7. The most important points that other YouTube channels or news sources do not explain well

Many pieces of content emphasize “creating a report with AI in 10 minutes.”

But the really important content is almost the exact opposite.

Core point 1: When you use AI well, you do not end up with extra time; you work more deeply

Using AI reduces the time spent on simple tasks.

But a good leader does not use the saved time only for rest.

They use that time to review more perspectives, examine more counterarguments, and think more deeply.

In the end, even if the same two days are used, the density of the output and the strategic implications are different.

Core point 2: AI can explode the productivity of one outstanding person, but it can also weaken the whole organization

A person who uses AI well can do in four weeks what three people used to take ten weeks to do.

From a company’s perspective, that is a massive productivity gain.

But if work becomes concentrated on that person, the organization is not sustainable.

The moment that person is absent, organizational capability can shake dramatically.

Core point 3: The reason AX fails is not lack of technology, but the existing reporting system

AI speeds up work, but if the company’s approval structure and meeting structure remain unchanged, bottlenecks will not disappear.

Ultimately, the speed created by AI gets absorbed by the organization’s processes.

If this is not changed, AI investment only increases costs and limits the effect.

Core point 4: If junior employees rely on AI, the fundamentals of work can disappear

AI reduces simple repetitive work, but if learners delegate everything from the beginning, their ability to think does not grow.

Leaders must design the scope of AI use differently depending on employees’ skill levels.

Core point 5: Agent automation is convenient, but it can weaken a leader’s thinking ability

Tasks like calendar coordination, repetitive emails, and file organization can be automated.

But if even market analysis, industry outlook, semiconductor cycle interpretation, and AI bubble judgments are handed over to automation, the leader’s value can weaken.

Leaders must not be ruled by AI; they must rule AI.

8. AI usage also changes in customer response analysis and marketing

When companies sell consumer products or do marketing, analyzing customer sentiment is very important.

In the past, focus group interviews, surveys, and outsourcing to external research firms were common.

But this approach is costly, time-consuming, and can miss small signals.

With AI, it is possible to closely analyze responses from Facebook, Instagram, communities, blogs, reviews, and social networks.

For example, companies like Xiaomi, which quickly reflect community feedback in product updates, can turn customer feedback into competitiveness.

This approach accelerates the digital transformation of marketing organizations.

Also, during periods of uncertainty in the global economy, changes in consumer response can directly affect sales and inventory strategy, making this even more important.

9. AI burnout: the better a leader uses AI, the more dangerous it can be

AI makes work endlessly improvable.

Even if report 1.0 is complete, you can keep revising it into 2.0 and 3.0.

As long as tokens and time allow, quality improvement can continue.

The problem is that this process wears down leaders and organizations.

If the person who uses AI well keeps getting more work and is expected to deliver higher quality, burnout can happen.

In particular, a structure that depends on one high performer is risky.

That is why, in the AI era, CHOs and executives should not only ask “Who uses AI well?” but also “How do we spread that capability across the whole organization?”

A company’s real capability is not determined by its single best person.

Long-term competitiveness is determined by the organization’s average level, especially how much it raises the lowest level.

10. The trap of agent automation: if you automate everything, you may disappear

Recently, tools that automate work, such as Claude Code, Codex, vibe coding, and agent builders, have been spreading rapidly.

These tools are certainly powerful.

But leaders must distinguish between tasks to automate and tasks that should not be automated.

Scheduling and repetitive tasks can be automated without reducing knowledge.

But if market analysis, industry outlook, and strategic judgment are all handed over to agents, a leader’s thinking muscles may weaken.

Vice President Kim gives the example of looking up words in a dictionary.

When you search words in a paper dictionary, you sometimes happen to see other words and gain new learning.

An electronic dictionary is faster, but it reduces those incidental discoveries.

AI agents are the same.

If they give only the exact answer you want too quickly and too precisely, leaders can lose the insights found along the side path.

11. Checklist for judging whether I am a leader who uses AI well

• Before telling others to use AI, I have applied it deeply to my own work.
• I do not use AI outputs as-is; I repeatedly check opposing views and evidence.
• I see AI-generated drafts as a starting point for discussion, not the final answer.
• I set the scope of AI use differently depending on employees’ skill levels.
• I use AI not to shorten report-writing time, but to deepen the message.
• I think about how AI use should change existing reporting systems and meeting methods.
• I design a structure so work does not become concentrated on one person who uses AI well.
• I remain wary that agent automation can weaken my thinking ability.
• I can judge when to stop using AI.
• I, not AI, hold the final decision-making authority.

If you can answer more than half of these items with confidence, you are close to being a leader who uses AI fairly well.

Conversely, if you strongly believe “AI does everything, so employees can just use it too,” that is a warning sign.

< Summary >

A leader who uses AI well is not someone who knows many tools, but someone who changes their own thinking and the organization’s workflow with AI.

You must distinguish between tasks to delegate to AI and tasks you must do yourself.

Leaders should not use AI like a magic wand; they should keep pounding away like a washboard to create insights.

Generative AI is not just a simple work automation tool; it is a core point of digital transformation that changes reporting systems, meeting methods, and even organizational structure.

However, if work becomes concentrated on one person who uses AI well, organizational burnout and concentration of capability can occur.

In the end, a good leader in the AI era is someone who masters AI while creating a structure in which the whole organization can grow together.

[Related Articles…]

• AI Transformation in Global Corporate Productivity Strategy
• The Core Changes in Semiconductor Cycles and AI Investment Flows

*Source: [ 티타임즈TV ]

– 내가 AI 잘 쓰는 리더인지 판단해 보세요 (김지현 SK AI위원회 부사장)

---