300万円のRPAが、月5,000円になった世界

The era of spending between 3 million to 10 million yen to implement RPA (Robotic Process Automation) is coming to an end.

A new type of tool, known as the “Poke” tool, allows businesses to automate processes simply by sending a one-line text message. With zero initial costs and a monthly fee of just a few thousand yen, there’s no need to even open a settings screen. By simply typing in a chat interface like Slack or LINE, “Please compile last week’s sales report and email it every Monday,” automation begins to run.

This is not just about a tool; it represents a structural change in what happens when the cost of business automation drops to one-hundredth of its previous level.

RPAが中小企業に届かなかった本当の理由

Traditional RPA was a “dream technology” for small and medium-sized enterprises (SMEs). To be precise, it was a technology that remained a dream.

The reason is simple: the costs were prohibitive.

The implementation costs for typical RPA tools, including design, construction, testing, and operation, start at a minimum of 2 million to 5 million yen. When engaging major vendors, it is not uncommon for costs to exceed 10 million yen. Moreover, every time the business flow changes, modifications are necessary, leading to annual maintenance costs that can range from hundreds of thousands to millions of yen.

Additionally, RPA required “specialized knowledge to operate.” This meant drawing flowcharts, setting conditional branches, and creating exception handling. Without personnel capable of handling RPA in-house, outsourcing was the only option, and if that person left, the system would cease to function. This is a classic case of dependency on specific individuals.

As a result, only large enterprises with IT departments could benefit from RPA. Many SMEs either gave up, thinking it was irrelevant to them, or attempted to implement it but ended up neglecting it due to their inability to utilize it effectively.

Poke型ツールが変えたゲームのルール

Text-based automation tools like Poke fundamentally rewrite the rules of the game.

Let’s clarify the differences from traditional RPA:

The costs have changed by two orders of magnitude. This is not just a matter of being “cheaper”; it signifies a change in the range of people who can use it.

For instance, a task that took 30 minutes daily to transcribe invoice data. With RPA, it would take three months to implement and cost 3 million yen. With a Poke-type tool, you simply give the text command, “When the PDF of the invoice arrives, extract the amount and date, and append it to column B of the spreadsheet.” The setup takes just five minutes, and it starts running automatically the next day.

The important point is that this automation can be done by the “person in charge on the ground” themselves. There is no need to request the IT department or outsource it. The person who knows the business best can automate it themselves. This is not just about eliminating dependency on specific individuals; it is about the “democratization of automation itself.”

AnthropicのAIエージェント簡素化が追い風になる

What is accelerating this trend is the simplification of AI agent development by Anthropic.

Anthropic is steering towards making the construction of AI agents using Claude easier. Previously, integrating AI agents into business required the hands of engineers for API design, prompt engineering, error handling, and more. Now, it is becoming possible to build them with less code and in a more intuitive manner.

This means that the “backend” of Poke-type tools is being strengthened. The accuracy with which AI can convert user-written text commands into business workflows is improving. As a result, the practicality of “automating with a single line of text” is further enhanced.

Astropadの「Workbench」——AIエージェントをどこからでも監視する

Another noteworthy development is Astropad’s announcement of “Workbench.” This platform allows users to remotely monitor and operate AI agents running on a Mac Mini from an iPhone or iPad.

What is significant about this?

Traditional RPA was predicated on the assumption that it would run on “office PCs.” This meant that to check on automated processes, one had to be in front of that PC. With Workbench, users can check the operational status of AI agents even while on the go and intervene as necessary.

Business owners of SMEs are not glued to their offices; they are often out in the field. The experience of “AI running on its own, with results only needing to be checked on a smartphone” aligns well with the working style of small and medium-sized enterprises.

「勝手に終わっている」が当たり前になる未来

The combination of Poke-type tools, the simplification of AI agents, and remote monitoring fundamentally changes the experience of business automation.

When arriving at the office in the morning, the sales aggregation from the previous day is already complete. The processing of invoices is finished. Inventory alerts are sent to Slack. Who did it? No one did. The AI did it on its own.

This experience of “having it done automatically” is the true essence of automation for SMEs.

The era of spending hundreds of thousands of yen on RPA, hiring dedicated personnel, and outsourcing for every modification is over. For just a few thousand yen a month, with a single line of text, the person in charge on the ground can automate processes themselves. This will become the new norm.

で、結局どうすればいいのか

There are three things to do immediately.

1. Write down one “repetitive task you do every day.” Whether it’s transcribing invoices, aggregating daily reports, or checking inventory — anything will do. Just start with one.

2. Ask a Poke-type tool, Claude, or ChatGPT, “Can this be automated?” You will receive surprisingly specific answers.

3. Test on a small scale. Instead of a full company rollout, start with one task and one person. If it works well, expand it.

Should you spend 3 million yen on RPA or just a few thousand yen a month for a single line of text? The answer is clear. The only question is whether you realize this and take action.