AI’s Utilization Cost is Approaching Zero. The Question is “What Comes Next?”

ChatGPT has opened unlimited text chat for free users. OpenAI has announced a $300 AI smart speaker. Alongside major AI companies like Google, Anthropic, and Microsoft, they have also agreed on a standard protocol for AI agents (Agent2Agent Protocol).

When we line up these three pieces of news, a clear direction emerges. “The cost of using AI itself is approaching zero.”

Just a year and a half ago, if you wanted to use GPT-4 for business, the minimum subscription was $20 per month. If you used it via API, costs could easily reach tens of thousands of yen depending on usage. Now, for text-based interactions, it is free and unlimited. By purchasing a speaker, you can access AI via voice at home or in the office. $300—about 45,000 yen. That’s cheaper than a business phone.

This is undoubtedly a tailwind for small businesses. However, we should pause and think.

“When AI becomes free, what will small businesses pay instead?”

When Costs Approach Zero, What Value Increases and What Value Decreases?

Looking back at the history of technology, every time something becomes dramatically cheaper, the value of something else skyrockets.

When the cost of disseminating information on the internet became zero, the value that increased was “attention.” Because anyone can publish, getting noticed became rare.

When server costs plummeted due to cloud computing, the value of ideas and execution regarding “what to create” increased.

The same structure is occurring now.

As the cost of using AI approaches zero, the value of “the ability to design what to have AI do” skyrockets. Furthermore, “the human decision-making ability to judge the results of what AI has done” becomes more important than ever.

Conversely, it is clear what value is decreasing. The ability to “use AI” itself is no longer a differentiator. Summarizing meeting minutes with ChatGPT or drafting emails—everyone can do that now. In other words, we have entered an era where “just using AI” generates no value at all.

The Three Costs Small Businesses Will “Really Pay”

1. The Invisible Currency of Data

There is a principle behind free services: “If you are not buying the product, you are the product.”

You paste your company’s sales data into ChatGPT for analysis. You hand over a customer list for classification. You have it summarize internal meeting minutes. It’s convenient. However, all those interactions are going to OpenAI.

Of course, OpenAI claims a certain level of protection in its privacy policy. However, in the free plan, the setting for using chat content for model training is turned on by default.

A common scenario in small businesses is as follows:

• A sales representative pastes customer inquiries directly into ChatGPT to generate responses.
• An accountant pastes billing data from clients for aggregation.
• The CEO has it summarize the contents of a management meeting.

Each instance may seem trivial. However, when accumulated, core business information flows to the platform. This is the price of “free.”

Specific Measures: Handle highly confidential information via API (which is not used for training). The $20 Plus plan also has an opt-out setting. Alternatively, use locally running open-source models (like Llama or Mistral). If you can protect your company’s data sovereignty for the difference between “free” and “5,000 yen a month,” it’s a bargain.

2. The Quiet Constraint of Lock-In

You start using ChatGPT for free and begin to accumulate internal knowledge. You create custom bots with GPTs and integrate them into your workflow. Within six months, the number of “businesses that cannot operate without ChatGPT” increases to five or ten.

This is lock-in.

What’s frightening is that lock-in deepens “the more convenient it becomes.” Because it’s user-friendly, you become dependent. Because you are dependent, you cannot leave. Because you cannot leave, you have no choice but to accept sudden price increases.

In fact, OpenAI has already hinted at the possibility of raising the Plus plan from $20 per month in the future. The Team plan for businesses is $25 per user per month. For ten employees, that’s $250 a month, about 4.5 million yen a year. While it may seem “cheap” now, what if it becomes $50 per user? What if it becomes $100?

This is where the AI agent standard protocol (A2A) comes into play.

The Agent2Agent Protocol, proposed by Google and agreed upon by OpenAI, is a common standard that allows different AI agents to collaborate. If this becomes widespread, agents created with “ChatGPT” and “Gemini” will be able to interact with each other.

In other words, the possibility of operating AI agents without being tied to a specific platform opens up. This is good news for small businesses. A structure that reduces the risk of lock-in is being established as an industry standard.

However, having a standard protocol does not mean that the transition cost becomes zero. The “design philosophy” of how to incorporate AI into your business tends to depend on the platform. Therefore, it is crucial to be aware of “designing for portability” from the start.

3. The Loss of Decision-Making Ability, the Most Expensive Asset

This is the most serious and least visible cost.

You use the answers provided by AI as they are. What started as “for reference” gradually turns into “since AI said it, it must be correct.”

The strength of small businesses lies in the “local knowledge” possessed by managers and frontline staff. This proposal resonates with this customer. This is the time to make this purchase. This employee should be entrusted with this task—such judgments are based on context and experience that cannot be measured by numbers alone.

AI is good at finding patterns, but it does not understand “this region, this customer, this context.” If you continue to adopt AI’s suggestions without question, your company’s decision-making criteria will become hollow.

The analysis that used to cost 3 million yen with a consultant can now be done in the equivalent of 50,000 yen worth of time with ChatGPT. This is fantastic. However, when relying on a consultant, there was a process of “thinking for ourselves and scrutinizing the consultant’s proposals.” With AI, that process tends to be skipped. It ends with “because AI produced it.”

Specific Measures: Always treat AI outputs as “drafts.” Establish a rule within the company that final decisions are made by humans. Additionally, cultivate the habit of questioning “why this conclusion” in response to AI’s proposals. This is not a cost but an investment.

The Reverse Structure That Only Small Businesses Can Achieve

This is not just a pessimistic narrative. Rather, this change presents a historic opportunity for small businesses.

For large companies to adopt AI, security reviews, legal checks, IT department approvals, and budget applications take at least 3 to 6 months. When deploying to 1,000 employees, training alone can cost several million yen.

What about small businesses? If the CEO says, “Let’s start using it tomorrow,” they can start using it tomorrow. In a company of ten, everyone can master it within a week. The speed of decision-making is on a different level.

In a world where the cost of using AI approaches zero, the competition will not be about “who can use AI” but “who can use it quickly and deeply.” And in terms of speed, small businesses can beat large corporations.

What should be done specifically? Here are three points.

① First, automate one task. Instead of a company-wide DX (I dislike this term), choose something small and specific, like “the invoice processing that takes three hours every month” or “the weekly report email creation.” There are countless examples where just integrating ChatGPT with spreadsheets can reduce three hours to 15 minutes.

② Establish rules for handling data. Decide what can be shared with AI and what cannot. A three-line rule is sufficient. “Do not include customer personal information,” “anonymize information containing amounts before inputting,” “confidential documents only via API”—this is enough.

③ Create a culture of questioning AI outputs. AI makes mistakes. It confidently makes mistakes. Based on this premise, make it standard practice to “check AI’s answers.” Organizations that can do this will be able to leverage AI as an ally. Those that cannot will be at the mercy of AI.

Conclusion: There’s No Such Thing as a Free Lunch. But Depending on How It’s Used, It Can Become the Ultimate Weapon

The cost of using AI is approaching zero. This is a fact. However, “being able to use it for free” is different from “being free of costs.”

Understanding these three “invisible costs”—data, lock-in, and decision-making ability—will allow AI to become the most powerful leverage for small businesses. If used without understanding, you will unknowingly sell off your company’s competitiveness.

Whether to buy a $300 speaker is honestly inconsequential. What should be questioned is, “In a world where AI is free, what will we spend our time and intellect on?”

Small businesses that can answer this question will survive the next decade.