300,000 yen becomes 50,000 yen. Will you still outsource?

Mirror has laid off 220 editors. The reason is simple: readers have started to rely on AI summaries, leading to a dramatic decrease in article clicks. In other words, the very structure of paying for “long articles written by humans” has collapsed.

We must not dismiss this news as merely a story about large media companies. The essence lies elsewhere.

The cost of creating content has changed drastically.

This change affects not large corporations, but local small businesses the most.

—

What Mirror’s Layoffs Mean—The Arrival of the ‘Unread’ Era

In September 2026, the UK-based Mirror announced the layoff of 220 editorial staff. According to the company, the primary reason was a significant decrease in direct traffic to the site due to the proliferation of AI summarization services.

What does this mean?

Readers are no longer reading the “original articles.” AI summaries from tools like ChatGPT, Perplexity, and Google SGE provide answers above search results. Users are satisfied with that and do not click on the original articles. Advertising revenue decreases. Editors become unnecessary.

This structural change is not limited to the media industry.

Press releases, blog posts, newsletters—these must also be created with the premise of being “AI summarized”. Conversely, it means that as long as you provide information that can be accurately picked up by AI, the necessity of outsourcing long and elaborate writing is fundamentally questioned.

—

What Has Changed with Local LLMs—The Revolution of ‘Completing In-House’

Another simultaneous change is the practical application of local LLMs.

Local LLMs are large language models that operate on a company’s PC without connecting to the cloud. Open-source models like Llama 3, Mistral, and Phi-3 have been released one after another, allowing practical-level text generation to be completed in-house with a PC that has around 16GB of VRAM.

What’s crucial here is that “monthly subscription fees are not even necessary”.

Content production that used to cost 3,000,000 yen a month can now be done for less than 50,000 yen a month. This “two-digit change in scale” is the essence of local LLMs.

Of course, cloud-based AIs like ChatGPT Pro or Claude Pro can also do similar tasks for a few thousand to tens of thousands of yen a month. However, local LLMs offer significant advantages for small businesses: “You don’t need to send confidential information outside,” “It can be used without an internet connection,” and “No monthly subscription fees are required.” Many companies are resistant to sending information to the cloud when creating sales materials that include customer data or internal management documents. Local LLMs break through that barrier.

—

Let’s Calculate—For a Local Manufacturing Company with 30 Employees

Let’s visualize this concretely.

A local manufacturing company with 30 employees. The PR role is filled by someone who also handles general affairs. The previous information dissemination costs were as follows:

【Before: Mainly Outsourcing】

• Press Release Creation: 2 per month × 150,000 yen = 300,000 yen
• Company Blog Posts: 4 per month × 50,000 yen = 200,000 yen
• Recruitment Page Updates: 1 per quarter × 150,000 yen = 50,000 yen (monthly equivalent)
• Sales Material Design Updates: 1 per month × 100,000 yen = 100,000 yen
• Monthly Total: Approximately 650,000 yen (Annual: 7,800,000 yen)

【After: Utilizing Local LLM + AI Generation Tools】

• PC Purchase Cost (16GB VRAM): Approximately 200,000 yen (depreciation about 3,300 yen/month)
• Increased Electricity Costs: Approximately 2,000 yen/month
• General Affairs Staff AI Operating Time: 10 hours/month × 2,000 yen/hour = 20,000 yen
• Image Generation Tool (Canva Pro, etc.): 1,500 yen/month
• External Review and Proofreading (2 times/month): 20,000 yen
• Monthly Total: Approximately 47,000 yen (Annual: About 560,000 yen)

Annual Cost Difference: Approximately 7,240,000 yen.

From 7,800,000 yen to 560,000 yen. A 93% reduction.

Moreover, this calculation does not include “speed”. Outsourcing would take 1 to 2 weeks for a single press release, while in-house it can produce a first draft in 30 minutes. Reports can be generated the day after an exhibition. Recruitment pages can be updated on the same day the job market changes.

It’s not just about cost reduction. A structure emerges where small businesses can win on speed against large corporations.

—

How to Address the ‘Quality’ Issue

A common counterargument arises here: “Is the quality of AI-written content reliable?”

To be honest, the output from current local LLMs is not at a level that can be used as is. Errors in facts, stiffness in expression, and contextual discrepancies—these will certainly occur.

However, consider this: even when outsourcing to a writer, the first draft is never submitted as is. It is always checked and revised in-house. The process remains the same.

What differs is the “time and cost until the first draft is produced”.

Outsourcing: Order → Hearing → Writing → First Draft Delivery (1-2 weeks, 50,000 – 300,000 yen)
AI Utilization: Prompt Input → Generation → First Draft Completion (30 minutes, effectively a few hundred yen)

When the cost of the first draft becomes one-hundredth, human time can be concentrated on checking and revising. This is the correct way to use it.

In other words, AI should be used as a “first draft generation machine” rather than a “replacement for writers”. Humans focus on “judgment” and “final touches”. If this division of roles can be achieved, quality can be assured.

—

Reasons Small Businesses Can Win ‘Because They Are Small’

For large corporations to implement AI, it requires security reviews, legal checks, internal approvals, PoCs, and company-wide deployment—taking at least six months to a year.

In a company with 30 employees, the president can say, “Let’s start using this next week,” and it can begin the following week.

The speed of decision-making is the greatest weapon of small businesses.

Moreover, local small businesses face the challenge of “having content to share but lacking the resources to do so.” They have the technical skills, yet their websites are a decade old. They have great products, but their social media is not updated. They want to hire but their job pages lack appeal.

AI breaks down this “resource barrier”.

Few small businesses can spend 650,000 yen on information dissemination. However, if it’s 50,000 yen a month, the story changes. Companies that thought “information dissemination is impossible for us” can suddenly become players. This represents a true “reversal structure”.

—

So, What Should We Do?

Here are three actions you can take starting tomorrow.

1. First, experience ‘first draft generation’ with cloud AI
There’s no need to immediately build a local LLM. You can use the free version of ChatGPT or Claude. Try getting a first draft of your company’s press release or sales material generated. The first step is to grasp the sense of what can be used and what needs to be revised.

2. Designate someone in-house to ‘check’
You need a person to make the final judgment on the text generated by AI. Someone who accurately understands your company’s business is best suited, whether from general affairs, sales, or even the president themselves. Create a system where this person can check in 30 minutes.

3. Set up an environment that can start for less than 10,000 yen a month
ChatGPT Plus (20 dollars/month) + Canva Pro (1,500 yen/month). With this, you can establish an environment for generating first drafts of press releases, blog posts, social media posts, and sales materials. Once you get used to it, transitioning to a local LLM (like Ollama + Llama 3) can eliminate even monthly subscription fees.

—

What the 220 Employees at Mirror Teach Us

The layoffs at Mirror are not a tragic tale of “jobs taken by AI.”

It’s a story about the fact that “the cost structure of content has changed.”

When costs change, players change. The era is coming when small businesses that could not afford to spend on information dissemination can stand on the same stage as large corporations.

The question is whether we recognize this change and take action.

The technology is in place. Costs have decreased. It all comes down to “whether to do it or not.”

Next week, try having AI write the first draft of a press release. That 30 minutes could lead to a difference of 7 million yen annually.