To further advance the automation of business processes, Questetra BPM Suite has built-in Auto-steps that integrate with AI generation services as standard features. By using these, you can improve the efficiency of information processing in your business, such as automatic text generation, summarization, review, classification, and even image generation.
This article explains the features, configuration overview, and usage examples of each AI Auto-step available in Questetra. It also summarizes the differences and suggests how to use them. In particular, it focuses on the differences between Questetra’s own [AI Agent] step and the Auto-step linked to external services (OpenAI / Google Vertex AI / Amazon Bedrock), providing information that will be useful in actual operation scenarios.
※2025-08 This is a comparison at the time of writing, and features and configuration overviews may change due to updates to each generation AI base model or built-in process.
1. List of Available Services and Their Features
The table below summarizes the features of the AI Automated Steps built into Questetra BPM Suite by service. The Questetra [AI Agent] step can be used in all editions of Questetra BPM Suite, including the trial mode. To use other automated Steps, you need an Advanced or higher edition. In addition, you need a subscription to the target service.
| Service Name | Separate Service Contract | Features of Collaboration with Questetra | Difficulty of Implementation | Foundation Model |
| Questetra [AI Agent] | Not required | Flexible control is possible, such as processing multiple data items | Low: No special settings required | Anthropic Claude |
| OpenAI | Required | Highly accurate natural language response and image generation | Middle: Creating an API key | ChatGPT / DALL-E |
| Google Vertex AI | Required | Audio, video and PDF input available | High: Creating a project, registering a service account, etc. | Gemini |
| Amazon Bedrock | Required | Claude model for high-level reading comprehension and image generation | High: AWS environment configuration, IAM user creation, etc. | Anthropic Claude / Stability AI Stable Diffusion |
2. [AI Agent] Process Overview
2-1. Features
[AI Agent] is one of the automated processing steps built into Questetra BPM Suite. It reads information from specified Data Items and uses the power of generative AI to make edits and suggestions.
Unlike other AI automated processes described below, this process allows you to specify which Data Items to read in natural languages such as Japanese and which Data Items to output the processing results to. It is also possible at a single step to perform input/output on multiple Data Items, or instruct consecutive processing based on the processing results, such as summarizing sentences and then translating the summaries.
You can also flexibly add instructions to the prompts, such as sentence length, target audience, and keywords, to fine-tune the output results.
2-2. Configuration Overview
The following settings are available in [AI Agent]:
- C1: Instructions for AI：Describe the purpose, format, target data, etc. in an instructional format.
- C2: AI model：Cannot be edited (Claude-based model is used internally).
- C3: Temperature：Adjusts the degree of creativity. Higher values allow for greater freedom.
- Specifying the target Data Item：
- Display only: Can be read but not edited
- Editable: Read and write accessible
- Hidden: Unreadable
2-3. Usage Examples
- Read the inquiry body and output an appropriate reply in the response suggestion field
- AI reviews the report in the Data Items and records feedback in the AI review section
- The manuscript written in the article draft is evaluated on a 5-point scale and the evaluation results are output
2-4. Usage Restrictions and Security
There is a daily usage limit for the [AI Agent] Task depending on the contract edition and number of users. The limit is calculated based on the number of tokens, and if it is exceeded, the automatic process will fail and a notification email will be sent. It will reset the next day and become available again.
In addition, the input data will not be used to train the model. Amazon Bedrock, which is used internally as the foundation, explicitly states that it will not provide data to third parties or use it for training.
Help Page: AI Agent
3. External Service Integration AI Automated Process
Many AI services have a mechanism to send data using APIs in addition to Web UIs. To send an HTTP request to these APIs from Questetra’s workflow platform, you need a script that assembles and sends the request content.
External service integration AI automation is a package that provides scripts for sending requests to each service’s API. Authentication information such as API keys and parameter values can be easily set on the UI screen.
Here we will introduce the external service integration AI Auto-steps provided by Questetra, service by service.
3-1. OpenAI
You can use the API by subscribing to a plan such as “Plus” for individuals or “Team” for businesses. API usage is charged separately.
3-1-1. ChatGPT: Chat
Overview
This is an Auto-step that performs natural language processing using OpenAI’s ChatGPT-based model. It obtains output based on prompts and stores the results in a Data Item. It also supports attaching image files and analyzing those images.
Main Settings Items
- Authorization Settings (API Key)
- Model（gpt-4-turbo、gpt-4o, etc.）
- Token Limit / Temperature / Stop Sequence
- User Message (dynamically specified with #{EL})
- Specify the role to be given to ChatGPT
- Attachment: Image
- Image reading accuracy specification (model dependent)
Usage examples
- Draft a response to a request
- Evaluate the naturalness of the expressions in the specifications and provide feedback
Help Page: OpenAI ChatGPT: Chat
3-1-2. DALL-E: Image Generation
Overview
It uses OpenAI’s image generation model DALL·E to generate images based on prompts. The generated images are saved in a File-type data item.
Main Settings Items
- Authorization Setetings (API Key)
- Model（dall-e-2、dall-e-3）
- Image Size / Style / Quality
- File Name Specification
- Prompt content and negative prompts
Usage Examples
- Generates images for presentation materials and illustrations for articles
- Product concept visual creation
Help Page: OpenAI DALL-E: Generate Image
3-2. Google Vertex AI
Google Vertex AI is a pay-as-you-go Google product. To use it, create a Google Cloud account and register your billing information.
3-2-1. Gemini: Chat
Overview
Using the Gemini model from Google Vertex AI, AI processing is possible using inputs such as text, PDFs, images, and audio. Its strengths are its high versatility and multimodal processing.
Main Settings Items
- Google Service Account (OAuth2 JWT configuration required)
- Model（Gemini 1.5 pro、Gemini 1.5 Flash, etc.）
- Region/Project ID
- Token Limit / Temperature / Stop Sequence
- Enter a message
- Attachments: Images, videos, audio, PDF (depending on the model)
Usage Examples
- Summary of attached PDF contents, bulleted list output
- Generate UI comments based on screenshots
Help Page: Google Vertex AI: Gemini: Chat
3-3. Amazon Bedrock
By signing up for AWS (Amazon Web Services), you can use the APIs provided by Amazon Bedrock to access a variety of AI foundation models developed by a wide range of companies. Among these, Claude from Anthropic and Stable Diffusion from Stability AI are highly acclaimed AI foundation models. Both are available on a pay-as-you-go basis.
3-3-1. Anthropic Claude: Chat
Overview
We use the Anthropic Claude Sonnet and Haiku models via Amazon Bedrock. The Claude model supports complex reasoning by enabling Extended Thinking mode.
Main Settings
- Access Key/Secret Access Key
- Region Code
- Model（Claude 3.5 Haiku、Claude 3.7 Sonnet, etc.）
- Tokens, Temperatures, System Prompts
- Extended Thinking / Stop Sequence
- Attachments: Images, PDF
Usage Examples
- Quickly summarize contracts and lengthy reports
- Automated review and feedback of technical documentation and specifications
Help Page: Amazon Bedrock: Anthropic Claude: Chat
3-3-2. Stability AI: Generate Image
Overview
Images can be generated using Stability AI’s Stable Diffusion model.
Main Settings
- Access Key/Secret Access Key
- Region Code
- Model（Stable Diffusion 3.5 Large、Stable Image Ultra v1.1, etc.）
- Prompt / Negative Prompt
- Aspect Ratio/Image Format
- Setting the file save location and name
Usage Examples
- Print banners and images for advertising and marketing
- Creating concept art and prototyping visual effects
Help Page: Amazon Bedrock: Stability AI: Generate Image
4. Process Comparison and Selection Tips
Taking into account the characteristics of each AI Automated Step explained so far, we will compare aspects of use such as ease of setup, particularly between [AI Agent] and other Auto-steps that link to external services.
4-1. [AI Agent] vs. AI Automated Processes with External Services
The table below compares AI automated steps in Questetra BPM Suite in terms of features, contract terms, and technical infrastructure.
| Comparison Items | AI Agent | External Service Integration AI Automated Process |
| Reading and Writing Data Items | ◎ (Can be freely specified using words) | △ (defined in the Auto-steps settings) |
| Free to Use | 〇 (Daily limits apply) | × (Subscription plan contracts and pay-as-you-go billing) |
| Ease of Setup | ◎ (Process setting screen only) | △ (API key and service account must be prepared) |
| Security | ◎ (No learning use, Bedrock compliant) | 〇 (Model dependent) |
4-2. Use According to Intended Use
The table below shows the services that are suitable for each purpose of use.
| Intended Use | Recommended AI Auto-step |
| AI Agent |
| OpenAI ChatGPT: Chat |
| AI Agent Amazon Bedrock: Anthropic Claude: Chat |
| Google Vertex AI: Gemini: Chat |
| Open AI DALL·E: Generate Image / Amazon Bedrock: Stability AI: Generate Image |
Conclusion
Questetra’s generative AI services are provided to automate and streamline various business processes by harnessing the strengths of each service.
The [AI Agent] in particular combines no-code operability with flexible control at the Data Item level, making it an easy-to-use solution for process designers.
On the other hand, ChatGPT, Gemini, Claude, and other tools that can be used in Auto-steps linked to external services support more accurate natural language processing, advanced image analysis, and multimodal input, and by using them according to your requirements, you can build more advanced automated processing.
We hope this article will be helpful in selecting services for future business process design.