AI is Trading Stocks on Its Own. Next, Your Company’s Accounting Will Also Be Done ‘Automatically’

AI agents are already operating in the market, trading stocks in real-time.

On the Sardine platform, AI agents autonomously analyze the market and execute trades. What human traders take hours to analyze, AI processes in seconds. This is a story from the financial industry. However, the essence of this technology is not limited to stock trading.

The structure of ‘AI completing judgment and execution without human intervention’ is beginning to spread across all business operations.

I want to ask small and medium enterprise owners: How many hours do you spend processing invoices each month? What about ordering tasks? Expense reimbursements? Can you imagine a state where these tasks are ‘automatically completed’?

What is the Difference Between ‘Automation’ and ‘AI Agents’?

‘Business automation’ has been around for a long time. Automating data entry for invoices with RPA tools or running routine processes with macros—these are examples of ‘rule-based automation.’ They operate according to rules set by humans.

AI agents are fundamentally different. According to an analysis by MIT Tech Review, AI agents do not ‘follow static rules’ but rather ‘understand situations, learn, and make optimal judgments.’ In other words, they can respond to unexpected situations.

Let’s consider a specific example.

Traditional RPA: ‘Create an order form for Company A on the 25th of each month and send it via email’—it operates according to the rules. Even if Company A raises prices or if there is excess inventory, it repeats the same order.

AI Agent: It comprehensively analyzes inventory data, past order history, price fluctuations from Company A, and information about alternative suppliers, concluding that ‘this month it would be 15% cheaper to place an order with Company B’ and generates the order form. Additionally, it sends a notification to the approver via Slack, and once approval is obtained, it automatically sends the order email.

This difference is decisive. RPA substitutes ‘manual tasks.’ AI agents substitute ‘judgment.’

What Changes in Accounting and Ordering for Small and Medium Enterprises?

The administrative departments of small and medium enterprises are chronically understaffed. It is not uncommon for there to be only one accounting staff member. If that person takes a day off, the invoice processing halts. This is the extreme of dependency on individuals.

If AI agents are introduced, the following tasks could become candidates for ‘automatic completion.’

① Invoice Processing
The AI reads the received invoices and automatically generates journal entries. It cross-references with past entry patterns and raises alerts for any anomalies. The three days of work that the accounting staff used to do at the end of the month could be reduced to 30 minutes of verification.

② Order Management
Based on inventory data and sales forecasts, the AI calculates the optimal timing and quantity for orders. It compares prices from multiple suppliers and creates the order form at the lowest price. The approval flow is also automated.

③ Expense Reimbursement
When employees upload photos of receipts, the AI automatically determines the amount, date, and account category. Expenses that do not comply with regulations are automatically returned. The nightmare of month-end expense reimbursements disappears.

④ Payment Reconciliation
The AI automatically reconciles bank deposit data with invoice data. Even in cases where the transfer name and the invoice recipient name are slightly different (e.g., ‘Ka Yamada’ and ‘Yamada Co., Ltd.’), the AI judges based on context and matches them.

What these tasks have in common is that they exhibit ‘a certain degree of established judgment patterns,’ ‘limited exception handling,’ and ‘relatively low risk of errors.’ This is the area where AI agents excel the most.

The Pitfalls of the ‘Automatically Completed’ Era—Audit Trails

It’s not all convenient. In a world where AI makes judgments and executes tasks on its own, there are significant risks.

‘Who made that judgment, and why?’ becomes unclear.

The Vorim AI platform addresses this very issue. It provides a service for ‘identity verification’ and ‘audit trails’ for AI agents. It records which AI agent made what judgment based on which data and what actions were taken—everything is documented.

For small and medium enterprises, this is extremely important from the perspective of tax audits and compliance. ‘The AI did it’ won’t suffice. There needs to be a trail that states, ‘This AI agent processed this invoice data according to this journal entry rule.’

Three Things Small and Medium Enterprises Should Prepare for Now

‘AI agents are still a distant reality’—you might think. However, they are already in operation in stock trading. The application to accounting and ordering is technically already possible. The question is not ‘when can we use it?’ but rather ‘when will we start using it?’

1. Articulate the ‘Judgment Rules’ of Your Operations
To delegate tasks to AI agents, you first need to clarify ‘the rules by which humans currently make judgments.’ Document the rules that exist only in the minds of accounting staff (e.g., ‘Be careful with Company A’s invoices as their consumption tax rounding is unique’). This is a prerequisite for AI implementation and a way to eliminate dependency on individuals.

2. Start Small with Low-Risk Tasks
Attempting to convert all operations to AI agents at once is risky. Begin with low-risk tasks like expense reimbursement or payment reconciliation, where the effects are more visible. There are increasingly more tools available that can be tried for a few thousand to tens of thousands of yen per month.

3. Implement with Awareness of Audit Trails
When selecting AI tools, always check whether ‘records of processing are retained’ and ‘the basis for judgments can be verified.’ Even if a tool is cheap and convenient, if it does not leave an audit trail, it could lead to serious issues later.

The Definition of ‘Human Jobs’ is Changing

Finally, let’s discuss the structural aspect.

In the era where AI agents will ‘automatically complete’ accounting and ordering tasks, will there be no human jobs left? The answer is No. However, the definition of ‘jobs that should be done by humans’ will fundamentally change.

Data entry, reconciliation, transcription, and routine judgments—these will no longer be jobs for humans. Instead, what humans should do is to ‘verify whether AI’s judgments are correct,’ ‘handle exceptions that AI cannot manage,’ and ‘make managerial decisions based on AI’s outputs.’

Accounting staff in small and medium enterprises will transition from being ‘people who make journal entries’ to ‘people who verify AI’s entries and provide insights for management.’ This is an upgrade.

The era of AI agents trading stocks has already arrived. The day when your company’s accounting will be ‘automatically completed’ is not far off. If you want to start preparing, today is the day.