No-Code Conversational AI Building Service "miibo" Releases New Feature "Thinking Mode" Where AI Agents "Autonomously and Repeatedly Ponder"

miibo Inc. (Headquarters: Minato-ku, Tokyo, Representative Director: Masashi Kunugi) is pleased to announce the release of a new feature, "Thinking Mode," for our no-code conversational AI service "miibo," which allows AI agents to "autonomously and repeatedly ponder."

What is Thinking Mode? Conventional AI chatbots typically provide an immediate, single response to user queries. However, this approach may not provide sufficient answers for complex questions or tasks requiring multi-step processing.

In miibo's newly implemented Thinking Mode, as shown in the example below, the AI agent will autonomously make decisions internally and process tasks step by step. User: "Help me prepare for my business trip to Tokyo next week."

◼︎AI Agent's Thought Process: ├─ Step 1: Confirm the travel dates. ├─ Step 2: List necessary reservations (transportation, accommodation). ├─ Step 3: Create a packing checklist. ├─ Step 4: Check the weather forecast and advise on clothing. └─ Completion: Organize all information and provide a response. AI Agent: "Regarding your business trip to Tokyo next week, I have compiled the following preparations..."

Features of Thinking Mode 1. Planned Task Execution In Thinking Mode, the AI agent first creates a to-do list and processes each item reliably. This ensures that even complex requests are handled without omissions.

2. High-Quality Answers through Deep Thinking By considering multiple perspectives rather than a simple question-and-answer format, it generates more accurate and comprehensive responses.

3. Transparent Thinking Process The management screen's log feature visualizes how the AI agent thought and what steps it took. This allows confirmation of the AI's decision-making process, which can also be used for improvements.

4. Deeper Answers by Integrating with Knowledge, MCP, and Connectors Thinking Mode demonstrates its true potential when combined with miibo's other features: "Knowledge Data Store (RAG)," "MCP (Model Context Protocol)," and "Connector Function."

A. Deep Answers Across Multiple Information Sources Traditional RAG generated answers using only information retrieved in a single search. With Thinking Mode, as shown in the example below, the system can repeatedly search the knowledge base as needed, building answers by accumulating information. User: "Create a competitive comparison report for new product A."

◼︎AI Agent's Thought Process: ├─ Step 1: [Knowledge Search] Retrieve specifications and features of new product A. ├─ Step 2: [Knowledge Search] Retrieve information on competing products B, C, and D. ├─ Step 3: [Knowledge Search] Review the format of past comparison documents. ├─ Step 4: Compile the collected information into a comparison table. ├─ Step 5: Organize the advantages of our company's product. └─ Completion: Generate a competitive comparison report.

AI Agent: "I have created the competitive comparison report. New product A..."

B. Autonomous Task Execution Integrated with External Systems When combined with MCP and Connectors, the system can automate the entire workflow, from information retrieval to processing and action execution, as shown in the example below. User: "Aggregate this month's sales and share the report on Slack."

◼︎AI Agent's Thought Process: ├─ Step 1: [Connector] Retrieve sales data from the sales management system. ├─ Step 2: Analyze the data and calculate month-over-month changes and trends. ├─ Step 3: [Knowledge Search] Review the format of past reports. ├─ Step 4: Create a report including graphs and summaries. ├─ Step 5: [MCP] Post the report to a specified channel via the Slack API. └─ Completion: "I have created the sales report and shared it in the #sales channel."

How to Switch to Thinking Mode Thinking Mode can be easily configured from the management screen's sidebar. The setup procedure involves just two steps.

1. Open "Model/Prompt" in the sidebar and check "Use Thinking Mode" within "Thinking Mode Settings." 2. Configure "Thinking Mode Prompt" with instructions for agent operations such as tool usage and multi-step processing.

With these steps, the setup is complete. No coding or specialized knowledge is required. Anyone can start building AI agents utilizing Thinking Mode with no-code, immediately.

Use Cases This feature can be more conveniently utilized in the following scenarios:

1. Complex Research Requests Example: Automatic Generation of Market Research and Competitive Analysis Reports For requests like "Summarize the pricing strategies of competitors in the domestic SaaS market," the system repeatedly searches and integrates competitor information, past research reports, and industry trend data from the knowledge data store, automatically generating a comprehensive analysis report with a single instruction. While traditional RAG generated answers based on a single search result, Thinking Mode provides highly accurate research findings by accumulating information.

2. Multi-stage Business Support Example: Automation of Recruitment and Internal Application Processes With instructions like "Process applicant A's document screening and send the results to Slack," the AI agent will autonomously plan and execute a series of steps: ① Review applicant information, ② Compare against selection criteria, ③ Assist in judgment of acceptance/rejection, ④ Generate result notification text, ⑤ Send to Slack. This eliminates the need for the person in charge to give instructions at each step, allowing for entire complex business workflows to be delegated.

3. Specialized Consultations Example: Internal Consultation Desk for Legal and Accounting Departments For highly specialized questions such as "Please check if the clauses in this contract are compliant with our company's regulations," the system provides specific advice that goes beyond generalities, by cross-referencing internal regulations, past contract examples, and relevant legal information from multiple knowledge sources. This reduces inquiry workload for relevant departments while providing evidence-based answers instantly.

4. System Integration Tasks Example: End-to-End Automation from Sales Report Aggregation to Slack Sharing With instructions like "Aggregate this month's lead acquisition numbers from the CRM and report them to the sales channel along with the month-over-month comparison," the system automatically processes a series of tasks: ① Retrieve data from the CRM system (Connector), ② Aggregate data and calculate month-over-month changes, ③ Refer to past report formats (Knowledge Search), ④ Generate report, ⑤ Post to Slack (MCP). This allows routine tasks spanning multiple systems to be completed with a single instruction.

Summary Thinking Mode is a feature designed to enable AI agents to autonomously and repeatedly exercise their "thinking power."

* Autonomously plan and execute complex tasks. * Generate deep and accurate answers by searching knowledge multiple times. * Operate external systems in conjunction with MCP and Connectors. * Visualize transparent thinking processes.

Thinking Mode is where AI agents autonomously think and judge their way to the optimal answer. Expand the possibilities of AI agents further with miibo's Thinking Mode.

About miibo Inc. miibo Inc. is dedicated to "Enriching people's lives with AI technology" as its mission, and develops and provides the conversational AI building platform "miibo." As exemplified in this case, the company is engaged in solving social issues by leveraging AI technology. https://miibo.co.jp

About the No-Code Conversational AI Building Service "miibo" The no-code conversational AI building service "miibo," handled by miibo Inc., is a service that allows anyone to easily create practical conversational AI with no-code. Official Website: https://miibo.ai

1. Easy AI Application Creation for Everyone No difficult skills or languages required! It's a conversational AI building service that allows you to immediately create AI-powered applications by utilizing your existing databases and Large Language Models (LLMs).

2. Super Agile Development: Build and Test Rapid implementation, proof-of-concept testing, effect verification, and refinement. The entire development PDCA cycle can be repeated at high speed. This enables swift proposals to clients. Conversational AIs applicable to various purposes are being created daily using miibo, with adoption progressing in listed companies, government agencies, and local municipalities.