AI Agent

AI Agent

AI エージェント

This is an automated process in which AI reads and updates data from specified fields (Data Items), comprehends the content, and performs analysis. Based on the input, it generates appropriate responses. In addition, it can flexibly adapt to specific instructions, such as adjusting the length of the output text, incorporating designated keywords, and tailoring the content for target readers, enabling information processing suited to your specific objectives.

Notes

• The AI models currently available are: Claude Sonnet 4 (deprecated), Claude Sonnet 4.5, Claude Sonnet 4.6, Claude Haiku 4.5, Claude Opus 4.5, Claude Opus 4.6
  • Claude Sonnet 3.5 v2, Claude Sonnet 3.7 and Claude Sonnet 4 are deprecated
• When additional models are released or updates are applied, the latest information will be provided in the Questetra announcements
• Thinking effort
  • If you select Sonnet 4.6 or Opus 4.6, you can set the depth of Thinking effort (adaptive thinking) with C3a: Thinking effort (low / medium / high / max)
  • The deeper the thinking process, the longer the processing time and the more tokens are consumed
• Extended Thinking (other than Sonnet 4.6 / Opus 4.6)
  • If you select a model other than Sonnet 4.6 / Opus 4.6 and set the temperature to 1.0, the Extended Thinking mode will be automatically enabled
  • Extended Thinking is recommended for tasks that require step-by-step reasoning, such as mathematics, coding, and analytical processing
  • When Extended Thinking Mode is enabled, overall response time may increase due to additional reasoning processes
• The Title, String-type, Numeric-type, Date-type, Datetime-type and Select-type Data Items can be set to “Display Only ” in the “Editing permission for AI” section
• The Title, String-type, Numeric-type, Date-type Datetime-type and Select-type that meet the following conditions can be set to “Editable ”
  • [Type of data source] is “Fixed choice list”, number of choices is 20 or less, and there are no Dependent parent
• AI can only modify the values of Data Items set to “Editable ”
  • The maximum number of data fields that can be set to “Editable” is 30 (the Title is counted as one field)
  • Exceeding this limit will result in a setting validation error
• AI cannot read the values of Data Items set to “Hidden ” regardless of the instructions
• If you specify a Discussion-type (text) Data Item in C4, the AI’s responses (intermediate and final responses) will be saved in that data item
  • The post submitter will be the Step name (or “AI Agent” if no Step name is set)
  • Messages exceeding 1 million characters will not be saved
• Usage is limited depending on the contract edition and number of contract users
  • Set a daily usage fee limit for foundation models
    • The usage fee for foundation models is calculated based on the total number of input and output tokens
    • If the limit is exceeded, the service task will fail to process
      • An error notification email will be sent with a message indicating that the usage limit has been reached
    • The limit resets every day, allowing the service task to be used again
  • The limit is calculated by multiplying the number of user accounts (number of seats for annual contracts, actual number of users for monthly contracts) by a coefficient specific to each edition
  • You can check the usage percentage relative to the usage limit in [System Settings – System Summary]
• Treatment of data input to the AI in this step
  • Questetra, Inc. will not use the input data to train its AI models
    • The [AI Agent] uses foundation models provided by Amazon Bedrock
    • Amazon Web Services, Inc. states that Amazon Bedrock does not provide the input data to third parties or use it to train its own AI models
      • For details, refer to the following AWS official documentation
        • Amazon Bedrock FAQs
        • Amazon Bedrock User Guide > Data protection

Capture

Prompt Examples

These are examples of “Instructions for AI”.

Please draft a reply based on the content of the “Inquiry Body” Data Item and enter it into the “Draft Reply” Data Item.

The activity report details are recorded in the “Details” Data Item. Please review the “Details” Data Item for any unclear expressions or typos and save the review results, in Markdown format, to the “AI Review” Data Item. If there are any feedback comments or suggestions for improvement, please include them, also in Markdown format, in the “AI Review” Data Item.

Please evaluate a draft article in the “Draft” Data Item. Rate the draft article based on clarity of summary, logic of explanation, originality, and compatibility with target readers, using a 5-point scale. Provide reasons for the rating. The target reader persona is provided in the “Persona” Data Item. The evaluation should be entered in the “Evaluation Result” Data Item.

See Also

• No Additional Cost & Easy Setup! Let’s Introduce AI Agents!