Both corporations and the public sector are diving headfirst into the AX full-scale initiative!
These days, businesses large and small, public institutions, and local governments all seem to be racing to establish dedicated AI teams as if by agreement. AI Solution Lab, AX Team, AI Frontier TF, AX Initiative Team, AgenTic AI Innovation Team, National AI Service Innovation Initiative. The names may differ, but the work is similar. They establish CEO-level task forces, put AX initiative teams at the top of organizational restructuring, assign employee IDs to AI agents, and conduct company-wide AX training for 500 people. This itself is proof that AI has become synonymous with innovation itself.
However, while we have established dedicated teams and are doing the various things mentioned above, when we actually think about whether the organization's productivity—or even the organization's DNA—has truly changed, there doesn't seem to be any real difference. We talk as if AI will change everything, but what has actually changed in our daily work besides creating documents and generating images or videos? I'm even starting to feel skeptical. Why is this happening? Today, I'd like to check five key points that those of you working in AI-related organizations must examine.
Nowadays, virtually every organization is operating a dedicated AI team. A major chemical materials company established a CEO-level task force and even brought affiliate companies under its umbrella, and a certain financial group explicitly listed 'AX Initiative Team' as the first item in its organizational restructuring. One telecom company granted an employee ID to an AI agent and declared 'AX Innovation 2.0,' and a well-known public institution newly created an AgenTic AI Innovation Team early this month. And that's not all—the government allocated 2.4 trillion won out of 10 trillion won in cross-governmental AI budgets this year to the industry-wide AX transformation, and mid-sized manufacturers and their partners are also establishing AX teams in response to customer demands.
Local governments are no different. OO County conducted prompt practice training for about 80 employees over four days at the end of August. The main subjects were prompts for administrative document and press release writing, and prompts for card news image generation. OO County is currently teaching prompt writing and repetitive task automation to 160 people divided into four sessions this week, and another OO County is even conducting company-wide AX transformation training for 500 people. At the metropolitan city level, they are outsourcing 'AI utilization capability enhancement training,' and educational institutions are offering rank-specific prompt packages, AX capability diagnostics, and in-house AI champion cultivation programs as products.
Having an AI team in the organization has become equivalent to declaring that we are innovating. Everyone is working hard. They're spending budgets, assigning people, and writing reports. But the AX practitioners we meet in the field tell us a similar story. Training was provided, accounts were distributed, but the organization's way of working is the same as last year. Only a handful of people have become good at using it.
The reason is simple: 'AI is being used as a tool.' Tools begin and end in the hands of the person using them. The key is to make AI become the organization's environment. Work should run on that environment, and people should shift toward designing that environment. But you might ask, isn't that just easy to say? No, it's actually really possible. We're proving it ourselves.
Companies That Use AI and Companies That Run on AI
In companies that use AI, people work and AI assists. The person writing the report delegates the draft to AI, and the person designing pulls mockups with AI. Work still begins and ends with human hands. The organization, workflow, and meeting formats remain unchanged. AI is a good tool placed on an individual's desk.
In companies that run on AI, work flows on top of AI. People design how that work should flow, and they adjust it as they watch it run. Agents do the execution, and people become designers. Data and experience accumulate in systems instead of in individual minds.
Most current AX approaches are being refined to make the former type of company work better. They make people write better prompts, provide better tools, and get more people to use them. If we continue in that direction, we'll end up with 'a company where all employees use AI well.' And that company still only accomplishes as much as its number of people. The five points below are questions that AX managers should ask themselves as they open their own plans.
Checkpoint 1. Instead of training completion rates, you should check the number of workflows running in the system after training ends
The outputs of prompt training are typically three things: completion rate, satisfaction score, and a handful of employees who have gotten better at it. Even in the training service RFP, attendance records and satisfaction statistics are listed as deliverables.
The problem is that all three of these remain with individuals only. When an employee who has gotten better at using it switches teams or leaves the company, that know-how disappears with them. AI training actually reinforces 'the fundamental problem of the existing organization—the structure where data and experience accumulate with individuals rather than systems.'
The opposite approach is to change the training's deliverables. On the day training ends, each team's automated workflow for one of their tasks should be registered in the company system. When you start evaluating training based on what the organization gained rather than what individuals learned, the entire training content changes.
Look at the 'expected effects' section of the training plan you're preparing right now. If there's only 'capability enhancement' written there, this checkpoint is going in the wrong direction.
Checkpoint 2. Instead of adoption rate or number of licenses, look at how many times per day employees access the tools we created
A major survey found that 38% of companies had adopted generative AI at the corporate level, and the #1 reason for not adopting was concern about information leakage (41.9%). So many AX teams tackle security review and enterprise license adoption as their first task. It takes six months. Then on the day accounts are distributed, they report that adoption is complete.
Account distribution is just the starting point. If you only give employees accounts, the way of working remains the same. Each person works a bit faster in their own seat, but the results don't remain with the company. From the company's perspective, this stage is more of a cost than an investment.
The metric needs to change. Instead of adoption rate or number of licenses, look at how many times per day employees access the tools your company has created. UniqueGood Company considers this access frequency to be that company's AX level. It's the number of times per day that employees open the workflow we directly created for our work. If this number is zero, you've distributed accounts, but the company hasn't actually moved yet.
Checkpoint 3. Instead of time saved, look at how many tasks run entirely without human hands
When an AX team opens a project call, similar lists come up. 30% reduction in report writing time, automatic meeting minute organization, promotional image generation, draft customer inquiry responses. In a Korea Chamber of Commerce and Industry survey, the #1 effect companies gained from AI was time savings (45.8%).
Creating outputs faster and better is just the first step. If you stop there, the company is just making the same work a bit less difficult. The truly new project is establishing a structure for repeatedly executing that work entirely.
Let me give a marketing agency as an example. When there's a division between the intake organization and the execution organization, the execution employee's new task becomes automating the execution work itself. It's making campaign planning, content creation, and performance reports run without human hands. Then the company can focus on customer consultation and meetings, and with the same number of people, it can handle multiple times more customers.
Delete the word 'time savings' from your project list. If there are still projects remaining, those are the real AX projects. If nothing remains, this checkpoint is also going in the wrong direction.
Checkpoint 4. Rather than AX team's AI expertise, someone who understands the entire workflow should be the leader
An AX team typically consists of field practitioners along with IT departments, developers, and external experts. Sometimes people who understand AI take the lead, and sometimes people are just assigned without regard. Of course, organizations with development teams would be more successful at AX, right? No, they're the same.
The key is someone who properly understands the entire workflow and can make things move. Most employees don't know where their company's work starts and ends. How customer inquiries come in, through whom quotes go out, where it always gets stuck, and which steps are actually unnecessary. There are people who know this. People who have experienced that flow from beginning to end.
Just having someone who can understand and adjust the overall flow isn't enough, and just having practitioners who can convert it to AX isn't enough either. But when these two people don't meet, a laughable situation unfolds. The field requests one step they can see. Please make report drafts come out automatically. The AX team creates what was requested. But what really should have been changed might have been why that report was needed in the first place. The person receiving the report might not have wanted a report but just one number, and that number might have already been somewhere in the system. These are things only visible from a position where you can see the entire workflow. So even if tools built step-by-step upon request are made very well, they just solidify the existing flow as is. They even automate unnecessary steps and make things more rigid.
The ability to see structure is more about position than capacity. The position that performs tasks and the position that sees the entire flow are different. Whether an entire task can be eliminated, whether flows should be merged, whether a decision can be delegated to an agent—these can only be judged from a position that can see the whole picture. In a small company, that position is the CEO. In a large organization, it's the leader responsible for that process end-to-end. That person needs to be in the design seat. The AX team handles the implementation with AI from beside them.
Look at your AX project list now. For each project, check whether 'someone who understands this entire flow' is participating in the design. If only requests are being exchanged, this checkpoint is going in the wrong direction.
Checkpoint 5. Instead of how much people used it, see if the workflows we created are still being used? And whether they're improving on their own
Completion rate, satisfaction, usage rate, time saved. Almost all metrics in AX performance reports count 'how much people used something.' When you measure performance by the numbers of people, you plan in a way that gets more people to use things. So next year's plan is more training and more accounts.
There are two things to look at. First, are the workflows we created still being used? Something that shines during the pilot period then stops is not a result. In a 2025 study MIT published, 95% of corporate generative AI pilots ended without results. Because nobody opens them the moment the pilot ends. Whether a workflow created three months ago is still running today—that one metric is more honest than a hundred usage rates.
Second, is that workflow improving on its own? A bigger problem is hidden here. AI produces better models every few months. If you create tools using AI and then just leave them, the moment the next model comes out, that tool becomes outdated. It stays frozen at the AI level at the time it was created. So management tools companies create with AI, even under the name of 'process improvement,' eventually get abandoned. If you leave tools as results, the more AI advances, the more what we've created becomes obsolete.
The opposite approach is to leave loops instead of tools. It's a structure where AI doesn't just run the workflow, but checks and corrects it on its own. Looking at this week's execution results, finding where it got stuck, proposing a better method, and running that method the next week—a cycle. We call this a learning loop. There are levels to using AI. Passing through simple direction, repeating loops, rule-based triggers, the highest is the stage of 'using it to improve and refine on its own.' If you have this loop, every time a new model comes out, our workflow improves with it. The organization's AX rides along with AI's development.
So the metrics for companies running on AI are different. Is the workflow created three months ago still running today? Has that workflow improved compared to last month? Are there parts that improved without people touching it? It's a metric that counts systems instead of people.
You don't have to change all five things at once. You only need to change one perspective. Looking at AI not as a tool but as an environment. Tools require someone to pick them up to work and stop when set down. An environment is where work flows on its own, and people design and fix that environment. With just this one perspective shift, the same training, the same accounts, and the same project calls start leaving entirely different things behind. Training leaves workflows, accounts become readable as access frequency, and projects become the work of redrawing entire flows. What creates big change in the current transformation is not budget or personnel—it's this perspective.
I'd like to suggest one thing you can do right now. Pick one of your own tasks as an AX manager. Make that task run from start to finish without human hands and upload it to your company system. That becomes your organization's first environment. From the second one onward, team members will follow and create.
In an era where one person moves ten thousand people,
The inspiring journey that created full-speed AX <Nanoventure>
UniqueGood Company's remarkable AX journey has been a hot topic every day. This book vividly records how UniqueGood Company transformed crisis into opportunity through AX. When you see how, even with a small number of people, they redefined existing work and processes, established structures for execution and verification, and transformed through agents, you'll discover that not only can a single person now create remarkable productivity, but a new environment has opened where small organizations can accomplish the work of over 100 people!
Easy ordering via KakaoTalk
Ordering via Yes24
Ordering via Kyobo Bookstore