The Cost of a Single Call Reduced to One-Sixth

I want to ask small business owners: how much are you currently spending on phone support each month?

If you assign one part-time employee to handle phone calls, it costs around 150,000 to 200,000 yen per month. For a full-time employee, it amounts to 290,000 yen (including social insurance). That’s 3,480,000 yen annually. Moreover, they can only respond during business hours. They take breaks and can fall ill. If that person leaves, the quality of service drops to zero.

This structure is beginning to break down.

With Google’s “Call for Me” feature in the 2025 Pixel series, ElevenLabs’ voice AI agents, and AI systems that automatically detect scam calls, it becomes clear that we have entered the implementation phase of an era where AI can make, receive, and assess calls.

Furthermore, the cost is less than 50,000 yen per month. That’s 600,000 yen annually—one-sixth of the traditional cost.

Gemini’s “Call for Me”—AI Takes on the Role of Making Calls

First, let’s clarify the facts.

Google’s “Call for Me” feature, integrated into the Pixel, allows AI to make calls on behalf of users. Whether it’s making restaurant reservations, checking availability at hospitals, or inquiring about product stock—when users instruct, “Call XX and ask YY,” Gemini automatically dials out, navigates automated voice menus, and converses with the human on the other end.

There are three key points:

1. Real-time text of the conversation is displayed. Users can monitor the conversation on their smartphones and intervene only when necessary.
2. It can handle interactions with humans. It can manage not only simple IVR (Interactive Voice Response) operations but also engage in natural conversations with human operators.
3. It operates on Google’s infrastructure. This means no special hardware or expensive API costs are required.

While this was announced as a consumer feature, that’s not its essence.

The important fact is that “AI has taken on the role of making calls.”

Sales representatives at small businesses typically make 30 to 50 appointment calls a day. With each call taking 3 to 5 minutes, 2 to 4 hours can be consumed just on the phone. What if we delegated this task to AI? Humans would only need to speak with “the right people.” The cost of making appointments would dramatically decrease.

ElevenLabs’ Voice AI—Operating 24/7 for Just a Few Thousand Yen

Next, let’s discuss the “receiving side” of phone calls.

ElevenLabs is at the forefront of voice synthesis AI. Their voice AI agents can respond to calls with a level of natural speech indistinguishable from humans.

Let’s break down the costs. ElevenLabs’ API charges about $0.3 (approximately 45 yen) for generating 1,000 characters. Assuming an average phone interaction involves 500 characters (about 1 to 2 minutes of conversation), that’s around 22 yen per call. Handling 500 calls a month would cost about 11,000 yen. Even when combined with phone line services like Twilio, the total would remain within 20,000 to 30,000 yen.

Moreover, it operates 24/7. Late-night inquiries and weekend reservation changes can be handled by AI. The era of missed opportunities due to “we are currently closed” voicemails is over.

The 290,000 yen labor cost is replaced by 30,000 yen for AI, a cost ratio of about one-tenth.

Of course, it’s not necessary to handle all calls with AI. The key is “triage.” AI can handle initial responses, assess the content, and escalate only the cases that require human intervention. With this structure, humans can focus solely on “calls that require judgment.”

CEO Mati Staniszewski states, “We should make it clear that customers are speaking with AI.” This is not only ethically correct but also practically rational. “AI will handle initial responses. We will connect you to a representative if necessary”—this simple statement sets appropriate customer expectations.

Scam Call Detection AI—The Evolution of the “Judging Side”

As AI begins to make and receive calls, the risks inevitably change.

The biggest concern is scam calls. As AI voice synthesis technology advances, scammers will also start using AI. The era where a call saying, “Your son has been in an accident” comes in the real voice of your son is just around the corner.

Here, the focus should be on honey pot-style scam call detection systems. Researchers are collecting vast amounts of conversation data from actual scam and spam calls and training AI on it. If this system is integrated into a company’s phone lines, it can assess incoming calls and indicate, “This call has an 87% chance of being a scam” at the point of ringing.

In other words, AI has entered the phase of not just “making” and “receiving” calls but also “judging” them.

This is particularly important for small businesses. Large corporations have security departments; small businesses do not. CEOs may directly answer calls. If AI can automatically filter out scam calls, that alone makes it worthwhile to implement.

Cost Comparison: Current Situation vs. Post-AI Implementation

Let’s compare the numbers concretely.

Annually, this shifts from 5,400,000 yen to 600,000 to 720,000 yen. The difference is approximately 4,700,000 yen.

What can you do with 4,700,000 yen? Invest in new equipment, raise employee salaries, or launch new businesses. For small businesses, this difference is not just a “nice to have” but a matter of “survival.”

Reasons Why Small Businesses Should Move First

What I want you to consider here is that this is not about “small businesses following what large corporations do.”

It’s actually the opposite. Large corporations invest hundreds of millions in existing call centers. Changing systems involves multiple levels of approval and contracts with vendors. Their movements are slow.

Small businesses are different. If the CEO says, “Let’s do it,” they can start next week. Without existing systems, they can build phone support that is AI-native from the ground up. The absence of legacy systems becomes their greatest weapon.

A company of 10 can achieve the same quality of service as a 100-person call center at one-tenth the cost.

This is not a pipe dream. The technology is already in place. It’s just a matter of “whether to do it or not.”

So, What Should We Do?

Here are three things you can start doing tomorrow.

1. First, visualize your company’s phone costs. How many hours are spent on phone support each month? Of the number of calls, what percentage should be handled by humans? In many cases, 70-80% are routine inquiries.

2. Start small. There’s no need to switch everything to AI all at once. Let AI handle only after-hours calls. Have AI respond to only specific categories of inquiries. There are already services available that can start for under 10,000 yen a month.

3. Start questioning the “personalization of phone support” today. If it’s said, “Only that person can handle it,” that’s a flaw in the system. If you textify the response manual, it can directly become AI training data. The work of eliminating personalization becomes the preparation for AI implementation.

The phone, the oldest and most neglected area of business operations, can fundamentally change the labor cost structure for small businesses when AI is introduced. The annual difference of 4,700,000 yen translates to 14,000,000 yen over three years and 23,500,000 yen over five years.

Looking at these numbers, can you still say, “It’s too early”?