"Claude is overwhelming" and "Astra is the strongest" won't move large enterprises: Talent development and organizational development required in the generative AI era

Hello everyone! This is Ono from KIZASHI Partners.

Recently, when I talk to HR professionals at large companies about talent development related to generative AI, I feel something very strongly.

That is, "what is being discussed at the forefront of generative AI and what is needed in the corporate workplace are worlds apart." That is the case.

In the engineering community, enthusiastic topics are flying around every day, such as:

・"Claude Code has achieved an evolution of another dimension"

・"From now on, it is the era of autonomous AI agents"

・"We should develop business applications in-house using Dify"

・"The reasoning capability of the latest GPT-6 Astra is tremendous"

On the other hand, what we hear from the corporate HR front lines is:

・"We distributed Microsoft Copilot to the entire company, but employees don't use it on a daily basis in the first place"

・"The level of utilization varies completely depending on the department or individual"

・"Even if we conduct training, people return to their old ways in the workplace, and no behavioral change occurs"

・"Management tells us to 'plan AI training that fits the workplace,' but we don't know how to structure it"

Even though these two worlds are talking about the same "generative AI," the conversations are surprisingly disconnected.

And I feel that this gap is a major factor in why generative AI utilization is not progressing as expected in many companies.

Why does such a gap arise? And how should companies design their operations and develop talent to truly achieve results?

In this article, I would like to delve into this theme.

The "Three-Body Problem of Perception" faced by engineers, HR, and management

Recently, when I talk to engineers, corporate HR personnel, and management, I sometimes feel as if I have come to a different country.

Engineers talk about "what can be done." HR personnel talk about "how to make them use it." And management talks about

"how much effect it will produce."

Each is correct.

However, each is looking at a completely different landscape.

As a cause for the lack of progress in generative AI utilization in companies, a state of fragmentation in the worlds viewed and evaluation criteria has occurred among the three involved players, which can be called the "Three-Body Problem of Perception."

According to a survey by PwC Consulting, while the generative AI adoption rate in large companies has reached 68%, only 29% of companies feel they are "contributing to revenue." Furthermore, other surveys highlight the reality that 43.7% of companies that introduced AI face the issue of it "not being used or taking root," and 28.0% face the issue of "not knowing what to use it for."

No matter how many cutting-edge features you line up, if there is no one in the organization to translate and bridge these three perspectives, AI investment will end in failure.

Most of the value of AI comes from "people"

So, what is the decisive factor for successful AI adoption?

In the "10-20-70 Rule" advocated by the Boston Consulting Group (BCG), the success factors for AI utilization are organized as follows:

・10%: Algorithm (the AI model itself) ・20%: Technical infrastructure ・70%: People and processes

When it comes to discussions about generative AI, people tend to focus on model performance, such as

"Claude is excellent,"

"Astra is amazing,"

"Gemini is growing."

However, that is only 10% of the whole.

In companies, it is necessary to consider the technical infrastructure that accounts for 20%, namely:

・Integration with existing systems

・Security standards

・Governance

・Data management

And the largest 70% is accounted for by:

・Employee literacy

・Business processes

・Organizational culture

No matter how excellent an algorithm you introduce, value will not be created unless you address the remaining 70%: 'how to raise employee skills,' 'how to redesign existing business processes,' and 'how to build a culture of trial and error without fear of failure.'

I believe that this 70% area is precisely the work of those involved in human resources and organizational development.

The 'dual structure' of company-wide standard tools and on-site practical work in generative AI adoption

As I mentioned in my previous note article (below), there is a significant gap between a company's enterprise-wide standard tools and the individual usage by staff on the ground.

【Share of company-wide standard tool adoption】 1st: Microsoft Copilot (45.3%)

2nd: ChatGPT (45.0%)

3rd: Google Gemini (28.3%)

【On-site practical usage rate (percentage of people spontaneously using it for work)】 1st: ChatGPT (51.3%)

2nd: Google Gemini (41.2%)

3rd: Microsoft Copilot (30.1%)

Copilot, which leads in company-wide adoption, has fallen to 3rd place in on-site practical usage. Furthermore, 83.0% of users of standard tools feel that 'the features and accuracy have not reached a practical level,' and 70.3% of employees responded that they are self-defensively using non-standard tools in parallel.

A 'dual structure' of tools (Shadow AI) is occurring daily, where 'employees on the ground are spontaneously using tools that are easier to work with behind the scenes because the tools distributed uniformly by the company are not enough to handle daily tasks.' However, very troublingly, if this is overlooked, it could lead to serious security risks such as the leakage of confidential information.

Why do large enterprises and public institutions center their strategy on 'Microsoft Copilot'?

Even though the front lines feel that 'ChatGPT or Claude are easier to use and superior,' why do large companies place Microsoft Copilot at the core of their company-wide deployment?

The reason is very simple. It is not 'because it is overwhelmingly superior as an AI,' but 'because it is overwhelmingly easier to control and manage as an enterprise.'

For large companies that have already introduced Microsoft 365, Copilot operates within existing security boundaries. 'Enterprise Data Protection (EDP)' is guaranteed, ensuring that internal data is not used for AI training without permission, and access rights management for each employee (who is allowed to see which files) and data loss prevention mechanisms are integrated from the start.

The larger the enterprise, the more likely it is that 'security control' will be prioritized over 'on-site practicality.' That is precisely why, no matter how much the front lines may desire other companies' tools, they often end up being consolidated into the Microsoft foundation.

In some local governments like Kobe City and advanced companies, initiatives to develop in-house AI applications for specific tasks while using Copilot as a foundation have begun. You should refer to such cases for the technical infrastructure part.

The focus of generative AI training is shifting from tool operation to 'practical application' and 'improving core human skills'

In the early days of generative AI, training that taught 'prompt techniques' or 'tool introduction and operation procedures' was mainstream. However, I believe that the training required from now on will change significantly.

The theme for the future is not 'which button to press,' but 'how to utilize generative AI in on-site business operations.'

As AI performance improves, humans will be required to take on more advanced roles.

Generative AI can now instantly search for knowledge, write text, and draft documents. However,

・Is the question appropriate?

・Is the answer valid?

・Is the information reliable?

・How should the final decision be made?

remain the roles of humans.

In other words, generative AI training must evolve from

'training on how to use AI'

to

'training on how to achieve results using AI'.

To achieve this, I believe it is essential to cultivate core human skills in parallel with practical business application, such as:

・The ability to question

・The ability to discern

・The ability to decide

・The ability to drive action.

I believe that generative AI will not eliminate human jobs, but rather shift the focus of human work.

Until now, the core of work was 'creating answers' to given questions. From now on, the work will be 'formulating appropriate questions, evaluating answers, selecting, and executing'.

That is precisely why human thinking and judgment skills will likely become even more important than before.

Are you satisfied with 120% efficiency? — Business redesign in the AI era

I sometimes hear stories from the field where AI is introduced, saying 'document creation time has been halved' or 'meeting minutes are easier to create'.

Of course, these are wonderful results, but I also feel a sense of crisis about being satisfied with that. This is merely improving the efficiency of existing tasks, not changing the work itself.

While people are happy that 'AI has automated data creation and halved work time,' humans are still endlessly checking the output data visually. There are by no means few such workplaces.

Even if you use AI to streamline some work processes, if the business processes before and after them do not change, it will not lead to significant improvements for the organization as a whole. In other words, while it may be partial optimization, it is not overall optimization.

What we really need to ask is not 'Can this task be done faster?' but 'Is this task really necessary?'

The true value of generative AI lies not in simple work efficiency, but in the ability to redesign the business itself (BPR). Only by stepping into that realm will business productivity jump by orders of magnitude to '200% or 500%'.

And I believe this trend will accelerate further in the future.

In recent years, the evolution of AI agents has been remarkable, and they are now being entrusted with a series of tasks, not just text creation support, but also information gathering, analysis, report writing, customer support, and routine task execution.

The important thing is not 'what to make AI do.'

'How to break down your own or your team's work and how much to entrust to AI' is what matters.

That is why future managers and HR personnel will be required not only to learn the latest technology but also to have a perspective that re-examines the business structure of their own departments.

So, what kind of development is necessary?

So, what kind of training is necessary to cause behavioral change in the field and establish generative AI?

I am sure you are all well aware that lecture-based group training or e-learning that only involves watching videos will not easily change employees' daily work habits.

In addition to group training, I think it is best to organize and prepare in the following three ways, such as adding hands-on support directly linked to practical work.

Measure "behavioral change," not "satisfaction," in AI training

I sometimes receive inquiries from HR personnel asking, "How should we measure the effectiveness of generative AI training?"

What I tell them is that "it cannot be measured by participant surveys alone."

I will explain this based on Kirkpatrick's four-level evaluation model, which is a framework for measuring the effectiveness of talent development.

In generative AI training, the most important thing is Level 3, "behavioral change."

・Have employees actually started using it?

・Has the frequency of use increased?

・Have work processes changed?

It is meaningless if you do not measure these points.

And beyond that lies the management perspective.

What management really wants to know is the return on investment (ROI).

Management will not pay for "it was fun."

・What they want to know is revenue

・Cost reduction

・Productivity improvement

is what they want.

That is why the HR department is required not only to visualize behavioral change but also to connect it to the final ROI.

Conclusion: Design a "mechanism for anchoring in the field" rather than chasing the latest technology

In the world of generative AI, remarkable technological innovations are happening almost every week.

"Claude is overwhelming," "GPT-6 Astra is the strongest," "Agents will take our jobs"—such cutting-edge discussions are stimulating, and it is important to keep them in view as the future of technology.

However, from the perspective of a company's sustainable growth, the real question is not "what can it do," but "how will it continue to be used in the field?"

No matter how excellent the latest model is, it is just a cost if employees do not use it. Conversely, even if a tool meets internal security standards, if employees acquire the "ability to ask," "ability to discern," "ability to decide," and "ability to drive," and can redesign their daily work processes, the organization can achieve dramatic evolution.

Utilizing generative AI is not an IT tool implementation project. Essentially, it is "organizational development itself" that updates employees' work styles and organizational culture.

And in the future, with the evolution of AI agents, we will be able to entrust more and more tasks to AI.

The important thing is not whether you know the new technology at that time. It is whether you can continue to think about "what humans should do" and "what should be left to AI" by taking a bird's-eye view of your own and your team's work.

I believe that the role we, who are involved in talent development and organizational development, should play is not just to chase the evolution of tools, but to design a mechanism that allows people and organizations to accept those changes and connect them to results.

The evolution of generative AI will not stop here.

That is precisely why I want to continue supporting people and organizations in truly changing, while staying close to the realities of the business.

KIZASHI Partners Inc. Representative Director Kenshi Ono Bridging the sincerity of the front lines with the resolve of management. A partner for organizational transformation.

◆ Regarding consultations with KIZASHI Partners

At KIZASHI Partners Inc., in addition to strategic design for hiring science and professional roles and building organizations that involve the front lines, we provide support for 'generative AI organizational adoption and talent development programs that actually move the front lines.'

Rather than ending with a one-size-fits-all company-wide training, we realize behavioral change in employees and effective operational reduction through a 'small-group hands-on approach' tailored to your company's security environment and actual business practices.

For companies struggling with 'we introduced tools company-wide but they aren't taking hold' or 'we want to build practical AI training that produces results on the front lines,' we will first help you organize your current situation and desired future state as a sounding board.

▶ Click here for inquiries and details

https://kizashi-partners.co.jp/

★ If you found this article helpful or interesting, please be sure to 'Like (♡)' and 'Follow'!

I will continue to share insights felt on the front lines and tips for organizational development under the theme of 'signs for the future of people and organizations.'

(PR) We are cooperating with listings on the job site Green.