How AI Agents Are Transforming Business Automation: Next-Gen Strategies from Oita Prefecture and US E-commerce Giants

The Essence of AI: Creating Time to “Engage with Questions”

Oita Prefecture has solidified its policy to utilize AI for creating all council responses. The noteworthy point is that its purpose is not merely “efficiency in document creation” but “increasing time to engage with questions.” This insight captures the essence of AI utilization. While many companies tend to view AI as a “tool to speed up tasks,” Oita Prefecture’s case demonstrates a more strategic perspective: using AI to “reallocate resources to high-value-added tasks that humans should inherently focus on.”

Meanwhile, major US e-commerce platforms like Salesforce and Shopify are focusing on developing and utilizing “AI Agents.” AI Agents refer not to conventional AI that executes single tasks, but to advanced AI that autonomously operates multiple tools and systems to complete a series of business processes. These two trends indicate that AI utilization is evolving from “efficiency in individual tasks” to “autonomization and redesign of entire processes.”

What Are AI Agents? 3 Core Concepts Business Leaders Must Know

AI Agents are the next evolution of generative AI. Specifically, they possess the following three capabilities.

1. Autonomous Judgment and Action

Conventional AI merely executed tasks instructed by humans (e.g., summarize the minutes). However, AI Agents autonomously perform a series of judgments and actions—such as “analyzing minutes, referencing related past responses, drafting replies to new questions, and prompting relevant departments for confirmation”—with minimal initial instructions. In Oita Prefecture’s case, an agent is envisioned that analyzes council questions, automatically retrieves related ordinances and past minutes, and drafts response outlines.

2. Coordinated Operation of Multiple Tools

AI Agents do not operate in isolation but integrate with existing business systems. For example, they can perform cross-functional operations like retrieving materials from cloud storage, sending confirmation requests to stakeholders via email systems, and registering tasks in project management tools. In our own practice, we have built an agent that integrates Claude with the Slack API, Google Drive, and GitHub to seamlessly handle everything from automatic generation of business reports to version control. The monthly cost, including AI tool fees, is only around a few hundred dollars, yet this creates over 1,500 hours of business time annually.

3. Learning and Adaptation

Excellent AI Agents learn through feedback and improve their output. In Oita Prefecture’s example, the very process of staff revising the AI-generated response drafts becomes learning data for the AI. By incorporating this “human revision” into a learning loop, the agent’s accuracy improves over time, ultimately drastically reducing the revision work itself.

AI Agent Implementation as Business Strategy: 3 Practical Steps

To utilize AI Agents within your company, a phased approach is more effective than rushing into large-scale development.

Step 1: “Visualization” and “Decomposition” of Processes

First, write out and visualize the entire business process you want to automate, such as “the flow from new customer reception to initial follow-up.” Next, decompose that process into smaller tasks like “information gathering,” “judgment,” “execution,” and “recording.” For Oita Prefecture’s “council response creation,” this could be decomposed into “receipt of question text,” “search for related materials,” “draft creation,” “legal compliance check,” and “supervisor confirmation.” This decomposition is the first step in defining what to have the AI do.

Step 2: Identifying Integration Points with Existing Tools

Identify which tools (email, CRM, internal databases, Google Workspace, etc.) are currently used for each decomposed task. The value of an AI Agent lies in automating the “manual data transfer” between these tools. In our experience, at many companies, 20-30% of employee time is spent on such simple copy-paste tasks between tools. The key is preparing an environment where AI can cross-functionally access this “siloed data.”

Step 3: Conducting a Small-Scale Pilot and Building a Learning Loop

Attempting to automate all operations at once is risky. First, select a process that is most labor-intensive yet has low risk if it fails (e.g., compiling and reporting internal survey results) and conduct a small-scale pilot. At this stage, it is essential to incorporate “human checking and revision” into the process and design a feedback loop where the revision data is used to train the AI. For tools, starting with platforms that require minimal advanced coding (e.g., Microsoft Copilot Studio, or enhancing Make.com/Zapier with AI integration features) is practical. Initial investment can start from a few hundred to a couple thousand dollars per month.

The Common Challenge for Public Sector and Companies Highlighted by the Oita Prefecture Case

Oita Prefecture’s initiative proves that AI Agent utilization is practical even in the public sector, where legal compliance and accountability are paramount. Similar challenges abound in corporate management. For example, creating compliance documents, conducting regulatory research, and preparing audit response materials are high-load tasks requiring accuracy and consistency with vast amounts of past documentation.

The key to applying AI Agents to these tasks is not aiming for “full automation.” Just as Oita Prefecture states it will “use AI for all council response creation” while humans retain final responsibility, a realistic division of roles in companies is also for AI to handle “draft creation and presentation of related information,” with humans making final judgments and approvals. This enables shifting professionals’ time from “information gathering and drafting” to “advanced judgment and coordination.”

Future-Oriented Investment: Also Considering Escape from SaaS Dependence

The fact that platformers like Salesforce and Shopify themselves are developing AI Agents provides an important implication. In the future, there is a risk of being locked into AI features provided by various vendors, leading to a “recurrence of SaaS dependence” that is not optimized for a company’s unique business processes.

From a mid-to-long-term perspective, possessing AI Agents that autonomize core business processes, either built in-house or in a customizable form as much as possible, leads to competitive advantage. The development of code-generation AI (Claude Code, GitHub Copilot, etc.) has significantly lowered the barrier to in-house development. The crucial point is not to internalize everything, but to make the strategic decision to keep under your control the AI governing parts that are the source of your competitive advantage, such as “unique decision-making processes” or “customer experience.”

The decision by Oita Prefecture and the movements of US e-commerce giants signal that AI utilization has entered a new phase. It is no longer merely the introduction of “convenient tools,” but a management strategy itself to “redefine the very meaning of human work and optimize organizational resource allocation.” The time has come for business leaders and CTOs to reconsider AI not as a technical challenge, but as a “strategic lever” enabling the reconstruction of business processes and organizational design.