“AI as a tool” has already ended
``Learn the tips of prompts,'' ``Use generative AI to summarize meeting minutes,'' and ``Deploy a chatbot to search internal FAQs.'' If your company's AI strategy is still stuck at this level, I have to say something a little harsher. Most of that investment will likely be wasted.
What we are witnessing is not a story of "adding a convenient office-saving tool". The phase of ``AI as a tool'' where humans issue instructions and wait for output is already becoming a thing of the past. What's happening now is a complete shift to "autonomous AI agents" that, when given just a goal, can think autonomously, formulate plans, and complete a series of messy business processes on their own.
Let's make a concrete comparison to see how on-site operations will change.
[So far: Utilizing AI as a tool]
I want you to imagine a customer support workplace. A complaint email is received from a customer. The person in charge copies the text, opens the AI screen in a browser, pastes it, and types in the message, ``Create a draft reply in a polite tone.'' The AI reads the text generated in a few seconds, corrects the text, pastes it back into the email software, and sends it. It's true that the time I have to write has become shorter. However, human manual work and confirmation remains the same, from checking customer information, matching it with the core system, going back and forth between screens, and even the final sending process. AI is nothing more than a "high-performance ghostwriting tool."
[From now on: Utilizing AI as an autonomous agent]
In a world of autonomous agents, AI constantly monitors your inbox. The moment an email arrives, without any instructions from a human, they refer to their CRM (customer management system) to check past response history, and at the same time access delivery tracking data from the logistics system.
He decided that the best solution would be to apologize for the delivery delay and issue a 500 yen coupon that can be used next time. After issuing the coupon in the core system via API and completing the change in delivery arrangements, the "Summary of response results" and "Email scheduled to be sent" are placed on the person in charge's approval screen. All a person has to do is look at the processing details and click the approval button.
The difference between these two is not simply a difference in processing speed. While the former is ``improving efficiency by inserting AI into human-led work,'' the latter is ``transforming operations by rearranging the business process itself end-to-end based on AI.''
Information gathering, data collation, initial judgment, and data transfer between systems... Workflows that were once performed by humans are now running automatically behind the scenes. This is not just a matter of "improving time performance." This is a change that fundamentally overturns a company's fixed cost ratio, the source of its competitiveness, and the very concept of organizational productivity.
And this worldview is not science fiction set in the distant future. It's a reality that forward-thinking companies are already implementing and quietly but decisively separating themselves from the competition.
Reality on the ground - the ever-widening gap
However, I would like you to take your eyes off the screen for a moment and take a look at the "real world" of your company.
Contrary to the spectacular success stories talked about by vendors, what is happening in actual offices may be far different from the ideal.
Here, I would like to share the story of a ``stalling introduction of AI'' that actually occurred in a major manufacturing company.
[Real on-site: The truth behind “tool neglect” that occurred at major manufacturing company A]
At Company A, a major manufacturing company with over 10,000 employees, the management team said, ``Don't be late!'', and the company introduced a generation AI tool provided by a major IT vendor to all employees at once. At the announcement, executives declared, ``With this, we aim to reduce work by 100,000 hours a year,'' and the internal portal contained a neat PDF manual that summarizes how to write prompts.
When it was first introduced, the number of logins increased due to its novelty. The management team was satisfied, saying, ``It's off to a good start.''
However, a year later, he was left speechless when he saw the log data extracted by the IT department.
More than 70% of all accounts are dormant, either ``almost never used since initial setup'' or ``only entering a word in the search window once a month.'' The only people who used it on a daily basis were less than 10% of the total number of accounts, and only a small number of young people and technical workers who were originally fond of technology and were working to improve their operations independently.
The DX Promotion Office, in a panic, questioned the on-site business managers and managers, and what they received back was their honest feelings.
"It's completely disconnected from the daily work flow." "Our daily work is completed in Excel and the core system. It's much faster to enter things manually than to open a browser, switch to the AI screen, think up instructions, and then copy and paste them."
"The location of risk and responsibility is too vague." "I can't trust 100% the numbers produced by AI. If I accept incorrect data and cause damage to my business partners, I will be held responsible. In the end, all calculations are rechecked by hand. It's nothing more than double work."
``No matter how hard you work to save time, no one will appreciate it.'' ``Even if you use AI to reduce your work time by one hour, you'll just end up adding new chores. Your evaluation and salary won't go up.In that case, it would be a smarter way to live within the company if you took your time the traditional way and put on a face that said, ``I did it carefully.''''
What remained as a result were license fees in the tens of millions of yen that piled up every year without being utilized, and the excuse that ``literacy at my workplace is low'' could be seen as a shift of responsibility.
It's easy to blame the scene. But what would you do if you were the manager at the site? They should take the same action. Who would want to use a tool that doesn't improve evaluations, requires double effort, and only involves risk?
There is an endless gulf between the potential of technology and the value created in the field. Tools are constantly evolving. However, organizations and people cannot keep up.
Golden rule of transformation: “30% technology, 70% people”
Why does our organization come to a halt even though we have acquired a weapon with such overwhelming power? Why does a large investment end up as a ``toy for some technology-loving employees''?
The cause is clear. This is because we hold a vague illusion that ``introducing technology will change the organization,'' and ignore the golden rule of resource allocation in transformation.
Global firms including McKinsey & Company in the US state that ``70% of changes fail due to human resistance and cultural barriers.'' A well-known example of AI-driven transformation is the ``10:20:70 rule'' advocated by the Boston Consulting Group (BCG) (10% for technology such as algorithms, 20% for data and infrastructure development, and 70% for process and human transformation), or the principle of ``30% technology, 70% people''.
Among the components needed to achieve AI adoption (fixation, fleshing out, and conversion into business value), the role played by "technical systems" such as selecting an AI model, building a security environment, linking APIs, and designing prompts accounts for only 30% of the overall success factors.
The remaining 70% is accounted for by human-related change management, including the psychological hurdles faced by the people who deal with it, existing evaluation systems, long-cultivated work practices, vertical divisions between departments, and an organizational culture that dislikes change.
However, what is happening in many companies is a ``complete reversal'' of this ratio.
Investment budgets, project man-hours, and most of the management's attention are focused on the ``30% technical elements'' such as ``which AI vendor to choose'' and ``what kind of accuracy was obtained in PoC?''
On the other hand, only a few percent of the costs are paid for ``changes in people and organizations,'' which determine 70% of the success or failure of change. After just holding one training session and posting a Q&A on the portal site, they say, ``We've set up the environment. Now it's up to the people in the field to use their wisdom.'' This is not reform, but just neglect.
Measures that prepare only 30% of the foundation and leave 70% of the mechanisms (human emotions, incentives, authority design, etc.) unresolved will not lead to business results. The low utilization rate and the obedience (actually pretending to be obedient on the outside, but not doing the work on the inside) are all the result of ``neglecting the human system'' in the initial structural design.
This series reveals “Management Change Management”
As AI evolves and becomes more autonomous, the bottleneck for business will shift from the technology itself to the adaptability of humans and organizations.
No matter how good AI is introduced, it will be meaningless unless behavioral patterns on the ground change. Furthermore, depending on the case, it may be necessary to thoroughly review decision-making authority and the evaluation system itself.
What is required to link AI investments to business results is not the introduction of tools by the IT department, but management-led ``transformation of organizations and people.''
In this series, I'm not going to talk about the temporary prompting techniques that are abundant in the world or compare tools.
Based on the lesson that many companies have faced in the process of promoting DX: "The biggest barrier is not technology, but the organization and people," we will unravel the muddy but essential "management change management theory" to incorporate new intelligence called AI into organizations, redefine the division of roles between humans and AI, and lead organizations to sustainable growth.
In the next (part 2), we will examine the mechanisms behind why organizations stagnate due to the same structural flaws, comparing the causes of failure in promoting DX in the past with the current reality of AX (AI transformation).
The inconvenient truth of AI adoption
The barrier that prevents companies from introducing AI lies not in technology but in "organizations and people." In this series, we will unravel the essence of organizational transformation through management-led "change management," which is essential in the era of autonomous AI.
See more articles in this series
Author of this article
Related articles
