Thousands of Billions in AI Investment Returned with a 25% Stock Price Plunge

IBM’s stock price plummeted by 25% in just one day.

This translates to a loss of approximately $30 billion (about 4.5 trillion yen) in market capitalization in an instant. The cause was the Q2 2023 earnings report, which showed revenues of $17.2 billion, a mere +1% increase compared to the same period last year. Coupled with a downward revision of profit forecasts, the market reacted with relentless selling.

What we need to consider here is not just that “IBM failed.” It is about the structure of why “large corporations that have bet heavily on AI are unable to deliver results.”

Meanwhile, small businesses that invest around 50,000 yen a month in AI are quietly reducing their operational costs by several million yen annually. This asymmetry is, I believe, the real news happening right now.

The True Nature of IBM’s “Platform Dependency Syndrome”

IBM has continued to make massive investments in AI, led by Watson. In 2019, it acquired Red Hat for about $34 billion (around 5 trillion yen), promoting a strategy of hybrid cloud and AI. Judging by the amount invested, their commitment is beyond doubt.

But what are the results? The core mainframe business has contracted due to reduced customer spending. The voices from client companies regarding AI-related services have spread, stating, “We implemented it, but it hasn’t yielded results,” and it has failed to become a growth driver.

Why is that? The reason is simple.

IBM operates on a business model that sells ‘AI platforms.’

What customers need is not the implementation of an “AI platform” but rather specific solutions to challenges such as “I want to automate customer inquiries” or “I want to process invoices in half the time.” However, vendors like IBM first require the platform to be installed, then customize it, bring in consultants, and sign maintenance contracts. The installation takes six months, customization another six months, and measuring effectiveness yet another six months. Before you know it, a year and a half has passed, along with tens of millions to hundreds of millions of yen spent.

According to a McKinsey survey, 71% of companies reported that “less than 25% of the AI agents they implemented actually function as multi-step workflows.” In other words, out of four implementations, three are useless.

This is the essence of “platform dependency syndrome.” The means become the end, distancing from the challenges on the ground. Larger corporations are more prone to fall into this trap because there are more people involved in decision-making, leading to the mindset of “let’s first establish the foundation.”

Meanwhile, the Reality of Small Businesses ‘Ending on Their Own’ for 50,000 Yen a Month

Let’s discuss a contrasting story.

A manufacturing company in a rural area (with 30 employees) has automated the creation of estimates by combining ChatGPT Plus and Zapier. The monthly cost is 3,000 yen for ChatGPT Plus, about 5,000 yen for Zapier, and 1,500 yen for Google Workspace. The total is less than 10,000 yen per month. Previously, sales representatives spent an average of 40 minutes creating each estimate; now it is done in 5 minutes and is “ending on its own.” When converted to labor cost savings, this amounts to about 4 million yen annually. The investment was recouped in the first month.

Another example: A real estate company in a rural area (with 8 employees) implemented an AI chatbot on its website. They used the ChatGPT API, with initial setup costs (including outsourcing) of about 150,000 yen and a monthly running cost of about 5,000 yen. Before implementation, about 60% of phone inquiries were routine questions like “What are your business hours?” and “Is there parking available?” This has nearly disappeared, allowing staff to focus on property viewings and contract work. The average response time for customer inquiries has been reduced from 12 minutes to 30 seconds, and customer satisfaction survey scores have improved by 23% compared to before implementation.

What these two companies have in common is that they did not “install a platform” but rather “solved a problem.”

Why Small Businesses Can Achieve Results with AI—Three Structural Advantages

This is not coincidental. Small businesses have structural advantages in utilizing AI.

1. Quick Decision-Making

For large corporations, implementing a single AI tool requires reviews by the IT department, security checks, approvals, budget allocation, vendor selection… at least three months, and sometimes up to a year. In a small business, the president can say, “Let’s give it a try,” and start the next day. The pace of AI tool evolution is measured in months. A three-month delay can be fatal.

2. Specific and Small Challenges

Rather than “building a company-wide DX foundation,” it’s more about “I want to automate this invoice processing” or “I want to reduce this inquiry response.” Because the challenges are specific and small, selecting tools and measuring effectiveness can be done simply. And current AI excels at smaller challenges.

3. Low Cost of Failure

If a tool costing 5,000 yen a month doesn’t work out, you can cancel it the next month. Even if you fail, the loss is only 5,000 yen. Can a large corporation stop a project costing hundreds of millions of yen midway? How many cases are there where projects that yield no results continue due to the burden of sunk costs?

These three structural advantages are particularly effective now that the costs of AI have dramatically decreased. The API usage fee for GPT-4 is just a few yen to several tens of yen per process. Features that would have required hundreds of thousands of yen to develop in a dedicated system three years ago are now available through SaaS for just a few thousand yen a month. When costs decrease, those who benefit the most are the ones who previously could not reach them.

Think About ‘What Will Become Unnecessary’ Rather Than ‘What Can AI Do’

IBM’s sharp decline highlights the limitations of those who sell “AI platforms.” And this carries important implications for those implementing AI.

Thinking about “What can AI do?” usually leads to failure.

The correct question is, “What will become unnecessary due to AI?”

• Does that phone response really need to be handled by a person?
• Why are you writing that estimate from scratch every time?
• Why is someone spending 30 minutes manually creating that meeting minutes?

Identify what will become unnecessary and target AI precisely at that. You don’t need a grand platform. Most cases can be handled with tools costing a few thousand yen a month and half a day’s setup.

So, What Should Be Done?

There are only three things that small business owners should do.

1. List three “repetitive tasks” within the company. It can be anything: email replies, data entry, report creation. List tasks that you feel are similar every time.

2. Try an AI tool costing less than 10,000 yen on the one that takes the most time. ChatGPT Plus, Claude, Zapier, Dify—anything is fine. Don’t seek perfection; a 60% automation is sufficient. If it’s faster than a human doing it, that’s a win.

3. If it proves effective, document the method and roll it out. Don’t let it become personalized. Change “only that person can do it” to “anyone can do it.” This is how to systematize.

While IBM has wasted 4.5 trillion yen, there are small businesses saving 4 million yen annually with an investment of 10,000 yen a month. This asymmetry is the essence of the current AI era.

There is no need to mimic large corporations. Rather, with speed and agility that large corporations cannot achieve, tackle the challenges in front of you one by one. This is the AI strategy for small businesses and the only, most powerful way to compete.