Conclusion

To put it simply: “Reducing people with AI” is the wrong approach.

Ford replaced engineers with AI. And then, they rehired them.

This one sentence already provides the answer. “AI for reducing people” is structurally doomed to fail. What was intended as cost-cutting resulted in decreased quality, increased rework, and ultimately higher costs. Even a giant corporation like Ford faced this outcome. If small businesses follow the same path, they won’t even have the luxury of rehiring; they could go under in one fell swoop.

So how should AI be used? The answer lies not in “reducing people” but in “enhancing human value.” There is a clear order to follow in this process.

—

What Happened at Ford—The Pitfalls of “Replacement”

Ford introduced AI and automated parts of its engineering operations. The goal was simple: to cut labor costs. In the U.S., where the annual cost per engineer exceeds $100,000 (about 15 million yen), replacing them with AI could lead to savings in the tens of millions of dollars. From a management perspective, this seemed rational.

However, the results were the opposite. While AI excels at data-driven pattern recognition, it cannot make contextual judgments like “Why was this design chosen?” or “What is the background for selecting this material?” The tacit knowledge that engineers held—past failures, relationships with suppliers, and insights gained from experience—was entirely lost.

As a result, design errors increased, rework costs ballooned, and project delays cascaded. Ford found itself forced to rehire the engineers it had let go. The costs of hiring, onboarding, and lost time amounted to losses several times greater than the “costs that were supposed to be cut.”

This reveals a structural problem. To “replace” a person with AI assumes that all the knowledge, judgment, and relationships that person held can be fully transferred to the AI. In reality, only about 20-30% can be transferred; the remaining 70% resides solely in human minds. Thus, the moment a replacement occurs, the organization’s capability drastically diminishes.

—

76% of Nurses Distrust AI—Field Skepticism as “Correct Intuition”

While Ford’s story pertains to manufacturing, the same structure is spreading to other industries.

In healthcare, a survey found that 76% of nurses distrust AI. Some may dismiss this as a lack of literacy in the field, but that is incorrect.

Nurses distrust AI because it “cannot explain the basis for its judgments.” Even if AI flags a patient’s vital data as “abnormal,” it cannot answer questions like “Why is it abnormal?” “What level of urgency is required?” or “How should we respond considering this patient’s medical history?” In a life-or-death environment, it is impossible to follow judgments that lack a clear basis.

This skepticism is, in fact, justified. The rejection arises from attempts to introduce AI as a “replacement for humans.” If AI is introduced as a “tool to assist human judgment,” the narrative changes entirely. Automating the detection of vital abnormalities allows nurses to notice issues more quickly. Automating record-keeping increases the time spent with patients. This way, the value of nurses increases, and distrust in AI diminishes.

—

An Australian Report Highlights the Bias in “Disappearing Jobs”

An Australian government report revealed that the risk of job loss due to AI is concentrated among specific demographics, particularly women and university graduates. Administrative tasks, data entry, and routine analytical work—so-called “white-collar routine tasks”—are more susceptible to replacement by AI.

This is not just a distant concern for small businesses. With one accountant, one HR staff, and one sales administrator, if such a company casually decides to “automate administration with AI,” it risks losing not just those jobs but the “company memory” that those individuals hold. The history of detailed interactions with clients, the accumulation of verbal instructions from the president, and experiences dealing with past troubles—all of these have been accumulated as a byproduct of administrative tasks.

If only the administrative tasks are handed over to AI, this “byproduct” will disappear. One day, when a problem arises, there will be no one to ask, “What happened with that issue?”

—

The Order Small Businesses Should Follow—”Increase Before Reducing”

So how should small businesses utilize AI? The order is everything.

Step 1: Identify “Time Thieves” First (Cost: $0, Duration: 1 Week)

Simply ask employees. “What tasks do you do every day that are honestly exhausting?”

Transcribing estimates, entering daily reports, reconciling invoices, and sending standard email replies. These tasks can take 1-2 hours per person each day. In a company with 10 employees, that amounts to 200-400 hours per month. Even at an hourly wage of 1,500 yen, that translates to 300,000 to 600,000 yen per month spent on “non-value-adding tasks.”

Identifying these tasks costs nothing. It’s merely a matter of whether to do it or not.

Step 2: Automate Without Reducing Staff, Increasing Time (Cost: $50-$500 per Month, Duration: 1-2 Weeks)

Automate the identified “time thieves” using AI tools. However, do not reduce staff. Redirect the freed-up time to “higher-value work.”

For example, a local manufacturing company (with 15 employees) automated the creation of estimates with AI. A task that previously took 40 minutes per estimate was reduced to 5 minutes. With 60 estimates generated per month, that resulted in a reduction of 35 hours monthly. The sales staff used this time to visit clients, resulting in two new orders within three months and an increase in annual sales of 8 million yen. The cost of implementing AI was only 12,000 yen per month for the tool.

At this stage, it is crucial to recognize that it is not about “reducing people” but rather about “increasing human productivity.” Employee motivation also rises. They feel liberated from tedious tasks rather than fearing that “AI will take their jobs.”

Step 3: Once Results Are Achieved, Systematize Knowledge (Cost: $10-$30 per Month, Duration: 1-3 Months)

When results from Step 2 are achieved, a culture of “Wow, AI can make things this much easier” begins to emerge within the company. At this point, it is time to start systematizing the tacit knowledge that has been personalized.

Instead of teaching AI the judgment criteria that veteran employees have in their heads, first document them. Feed internal FAQs and manuals into an AI chatbot so that anyone can search for them. This can be set up for about 10,000 to 30,000 yen per month using Notion or Google Docs in conjunction with the ChatGPT API.

This is not for the purpose of “reducing people.” It is a system to ensure that “knowledge remains even if someone leaves.” The biggest risk for small businesses is that operations can grind to a halt if key personnel leave. By externalizing knowledge through AI, this risk can be mitigated.

Step 4: Only After These Steps Should You Consider “Optimizing Placement” (Duration: 6 Months to 1 Year Later)

After completing the three steps, only then does reconsidering personnel placement become an option. However, it should be about “reallocation” rather than “reduction.” Those who had their administrative tasks halved can be moved to customer success or quality control roles. Individuals freed from routine tasks can transition to jobs that generate higher added value.

As a result, sales can increase by 1.3 to 1.5 times with the same number of employees. Profit margins can rise without cutting staff. This is the true essence of “AI that enhances human value.”

—

The Critical Difference Between Ford and Small Businesses

Ford attempted “AI for reducing people” and failed, but had the resources to rehire. Small businesses do not have that kind of resilience. Therefore, they must not get the order wrong from the start.

Another advantage small businesses have that large corporations lack is the “speed of decision-making.” To change company-wide policies at Ford requires board approval. In a small business, if the president says, “Let’s use this starting next week,” they can act by the following week. In terms of the speed of AI implementation, small businesses are structurally more advantageous than large corporations.

Testing a 10,000 yen tool next week, measuring its effectiveness in two weeks, and switching to another tool if it doesn’t work—this cycle can only be maintained by small businesses.

—

So, What Should We Do?

1. By the end of this week, ask employees to list three “exhausting tasks” they do every day.
2. Automate one of those tasks using an AI tool costing less than 10,000 yen.
3. Reflect on what could be accomplished with the freed-up time two weeks later.

That’s all it takes. There’s no need for grand AI strategies or DX roadmaps. Just start by making one task easier.

What Ford’s failure teaches us is that “reducing people with AI” should be the last resort. First, it should be about “increasing people’s time with AI.” If this order is followed, small businesses can undoubtedly become stronger with AI.