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

0. Introduction

The AI ​​Agent feature has been released in version 17.0 of Questetra BPM Suite.

This new feature is available at no extra cost to all editions of Questetra, and does not require a separate contract with an AI model vendor such as OpenAI. In addition, instructions to AI can be given in natural English. This makes it easy to incorporate AI into business processes, potentially enabling dramatic improvements in business efficiency.

In this article, we will explain the overview of the AI Agent step and basic usage, and introduce how to make use of it.

This article is intended for system administrators and business App designers at companies who are considering automating or improving business processes.

1. Overview of AI Agent

AI agents can be used by using the new automatic processing step [AI Agent] that can be placed in a workflow app. [AI Agent] executes processing based on the instructions set in the step.

Main Features of [AI Agent]

• Instructions in Natural Language
  Based on instructions written in natural languages ​​such as English, it references data from specified Data Items and automatically performs tasks such as generating sentences, summarizing, and classifying.
• No Authentication or Communication Settings Required
  No authentication or communication settings are required to use the generative AI, and it can be easily set up on the step settings screen of the modeler.
• Free to Use
  It is included in the Questetra usage fee and there is no additional charge.
  (However, there are restrictions on usage depending on the contract.)

2. Basic Usage

The steps to use [AI ​​Agent] are as follows:

• a. Add the [AI Agent] Step to the workflow diagram
• b. Prepare Data Items for AI to reference/edit
• c. Enter instructions for the AI in the configs

So let me explain step by step.

a.  Add the [AI Agent] Step to the Workflow Diagram

First, let’s add an [AI Agent] Step to the workflow diagram of the business App.

• a-1. Launch the modeler and display the [Workflow Diagram] editing screen
• a-2. Select [AI Agent] from [Auto-Processing Task] on the palette
• a-3. Place it where you want the process to occur on the workflow diagram
• a-4. Connect the flows appropriately and incorporate them into your business process

b. Prepare Data Items for the AI to Reference/Edit

Next, prepare Data Items to store the data that the AI ​​will read and the data that you want the AI ​​to edit.

• b-1. Display the Data Item editing screen on the [Data Item] tab
• b-2. Add the required Data Items
• b-3. Set the details of each Data Item
• b-4. Set the editing permissions for each Data Item
  • Similarly, for [AI Agent], set [Editing permission for AI]
    • AI can only refer to the values ​​of Data Items that are specified as [Display only] and [Editable]
    • AI can only change the values ​​of Data Items that are specified as [Editable]

The data entered into the AI ​​will not be used to train the AI ​​model. Please use it with confidence. For details, please refer to the Notes section of the Auto-Processing Step [AI Agent].

c. Fill in Instructions to the AI in the [AI Agent] Step

• c-1. Open the detailed settings screen for the [AI Agent] Step
• c-2. In [C1: Instruction for AI], write the instructions for the AI to perform
  • Example: “The text of the inquiry email is stored in the data item “Inquiry Content”. Please summarize “Inquiry Content” in 50 characters or less and write it in the data item “Summary”.”
  • I recommend that you give as specific instructions as possible, as if you were asking a subordinate, a colleague, or an outsourcer to do it for you
    • Please specifically explain what data is in each Data Item and how each Data Item will be processed
• c-3. Set the Temperature
  • Temperature is the parameter that controls the balance between solidity and creativity in the output
    • When creating output, the AI ​​can choose between solid answers and more challenging/creative answers
  • The appropriate range varies depending on the application
    •  Low temperature (0.0 to 0.3): Solid ⇒ Summary
    •  Medium temperature (0.3 to 0.7): Balanced type ⇒ Creating article ideas
    •  High temperature (0.8 to 10.0): Creative ⇒ Idea generation

3. Precautions for Use

While AI Agents are useful, you should be aware of the following points:

• Accuracy of generated content
  Due to the nature of the generative AI model on which it is based, the content may contain errors. Please be cautious when using it directly for important decision-making.
• Usage limits
  [AI Agent] can be used free of charge, but there are usage restrictions depending on the contract edition and number of contracts.
  For details, please check the [AI Agent] help

4. Examples of Use

Below are some examples of how [AI Agent] can be used.

Use case 1: Automatic summarization of inquiries

By automatically creating summaries and naming cases according to their contents, the person handling the inquiry can quickly gain an understanding of the overview.

We will explain how to set up the parts of the above workflow that relate to [AI Agent].

Here we use Message Start Event (email). For information on how to use Message Start Event (email), please see M218: Auto Start Triggered by Incoming Email.

1. On the Modeler’s Workflow Diagram Editing Screen, place the Auto-Processing Step [AI Agent] next to [Message Start Event (Email)]
2. Connect the Steps that have been placed with a flow
3. On the Data Item editing screen, add the following data items
   • String (single line): Summary
   • String (multiple lines): Email Body
4. Return to the workflow diagram, open the settings screen for [Message Start Event (Email)], and set the following:
   • Body: Data Item “Email Body”
5. Open the setting screen of the [AI Agent] Step and set [Editing permission for AI] as follows:
   • Title：Editable
   • Summary: Editable
   • Email Body: Display only
6. Set [C1: Instruction for AI]
   • Example settings:
     The body of the inquiry email is stored in the data item “Inquiry Content”. Please summarize the “Email Body” in 100 characters or less and write it in the data item “Summary”. Also, please create a subject that matches the content in 20 characters or less and write it in the data item “Title”.
7. Set [C3: Temperature] to 0.3

Example 2: Proofreading of approval documents

By having AI review approval applications, we can improve the quality of the approval documents. This also improves the speed of approval and reduces the burden on the approver.

We will explain how to set up the parts of the above workflow that relate to AI Agent.

1. On the Modeler’s Workflow Diagram Editing screen, place an Auto-Processing Step [AI Agent] next to the Human Task that creates the approval document.
2. Connect the Request Form Creation Task and the [AI Agent] Task to each other via a flow.
3. On the Data Item editing screen, add the following data items.
   • String (multiple lines): Request for Approval
   • String (multiple lines): Issues
4. Return to the Workflow diagram, open the settings screen for the [AI Agent] Step, and set [Editing permission for AI] as follows:
   • Request for Approval: Display only
   • Issues: Editable
5. Set [C1: Instructions for AI]
   • Example settings:
     Please tell us what needs to be improved about the content of the approval request in the data item “Request for Approval”. Please point out any errors or omissions. Please enter the improvements and suggestions in the data item “Issues”.
6. Set [C3: Temperature] to 0.5

Example 3: Automatic generation of comparison tables of old and new data

When updating regulations, we use AI to create a comparison table between the old and new regulations, allowing reviewers to instantly and visually understand the changes, reducing the burden on reviewers and the risk of oversights and misunderstandings.

We will explain how to set up the parts of the above workflow that relate to AI Agent.

1. On the Modeler’s workflow diagram editing screen, place an Auto-Processing Step [AI Agent] next to the Human Task that inputs the proposal items for new and old regulations.
2. Connect the proposal input task and the [AI Agent] Task together.
3. On the Data Item editing screen, add the following data items
   • String (multiple lines): New Rules
   • String (multiple lines): Old Rules
   • String (Markdown): Old and New Comparison Table
4. Return to the Workflow diagram, open the settings screen for the [AI Agent] Step, and set [Editing permission for AI] as follows:
   • New Rules: Display only
   • Old Rules: Display only
   • Old and New Comparison Table: Editable
5. Set [C1: Instruction for AI]
   • Example settings:
     Please create a comparison table by comparing the Data Item “Old Rules” with the Data Item “New Rules” and enter it in the Data Item “Old and New Comparison Table”. However, if “Old Rules” is empty, there is no need to create a comparison table. Also, if there are any typos or unnatural points in the “New Rules”, please list them, propose improvements for each, and enter them in the Data Item “Improvement Proposals”.
6. Set [C3: Temperature] to 0.5

---

5. Summary

The AI ​​Agent introduced in Questetra Ver. 17.0 is a powerful tool for easily incorporating generative AI into workflows. Please try incorporating this feature into your operations, so it can contribute to automating processes and improving work quality.
Also, if you find a good way to use it, please post it and share it with the Community.