使用 Amazon Quick Automate 构建端到端 RFI 问卷工作流

使用 Amazon Quick Automate 构建端到端的信息征询（RFI）问卷工作流，以解决各种规模组织都会面临的挑战。一个典型企业每年可能需要处理数百份 RFI 问卷，每份问卷都以复杂的多工作表工作簿形式提交，其中包含层级化的问题集、类别元数据和多种响应类型。这些文档的数量、多样性和复杂程度带来了显著的运营挑战。手动提取、结构化和处理 RFI 问题需要反复协调，延迟响应，而且每当问卷格式发生变化时，都可能引入错误。

Amazon Quick Automate 可处理受支持企业流程的多 Agent 自动化，这些流程可跨越部门、系统、UI 和 API 交互以及第三方系统。它利用一个 Agent 团队来简化业务流程管理，并减少维护开销。你只需用自然语言描述处理目标，Quick Automate 即可生成涵盖数据摄取、转换、验证和输出的可执行工作流。你可以通过对话优化工作流，针对你的数据运行它，并将经过验证的版本推广到不同 AWS Region。该方法可帮助你更快响应 RFI，并减少格式不一致和错误。自动化设计可随工作负载扩展，你也可以通过更新自然语言指令来适配新的问卷格式。

本文介绍如何构建端到端自动化：从 Amazon Simple Storage Service（Amazon S3） 读取多工作表 RFI 工作簿，提取并结构化问卷数据，然后将清洗后的输出写回 Amazon S3。本教程使用 Amazon Quick Automate 和 Amazon S3。有关成本详情，请参阅 Amazon Quick 定价 和 Amazon S3 定价。

解决方案概述

本演练将自动处理存储在 Amazon S3 中的 RFI 工作簿。采购团队收到 RFI 问卷时，需要提取每个问题，并保留其类别和响应类型。目标是生成一个供下游团队使用、解读和采取行动的结构化数据集。在本示例中，我们展示了使用 Amazon Quick Automate 自动化该流程的解决方案。首先，将 Amazon Quick Automate 连接到存储 RFI 工作簿的 Amazon S3 存储桶。然后，创建一个自动化项目，用于选择数据连接器并组织工作流。项目创建完成后，你可以使用自然语言描述处理逻辑：读取什么内容、如何将其结构化，以及将输出写入何处。最终得到的是一个可运行的自动化，它会读取多工作表 RFI 工作簿，提取并结构化问卷数据，并输出一个干净的逗号分隔值（CSV）文件。对于常见场景，这一过程无需自定义代码。下图展示了步骤的高级顺序。

图 1：RFI 工作簿自动化步骤的高级顺序

该图展示了从连接数据源到在生产环境中部署完成自动化的过程。请查看 AWS Regions 页面，了解最新可用性。

该解决方案遵循以下步骤：

1. 设置 Amazon S3 操作连接器 – 将 Amazon Quick Automate 连接到你的 S3 存储桶。
2. 将 S3 集成添加到自动化组 – 与将使用它的自动化组共享 S3 操作。
3. 创建自动化项目 – 创建用于描述处理逻辑的项目。
4. 描述处理逻辑 – 输入自然语言提示，让生成式 AI 助手构建工作流。
5. 通过对话优化 – 通过请求有针对性的变更，对生成的工作流进行迭代。
6. 验证结果 – 运行工作流，并在预生产（开发）AWS 账户中验证输出。
7. 使用 Import/Export 推广 – 导出经验证的版本，并将其导入生产账户或目标 AWS Region。

现在我们已了解该解决方案的高级步骤，接下来介绍环境要求。

前提条件

开始前，请确认你的环境满足以下要求。

1. 拥有 Amazon Quick Enterprise 订阅，并可访问 Amazon Quick Automate。
2. 一个与 Amazon Quick 应用位于同一 AWS Region 的 Amazon S3 存储桶。
3. 熟悉 AWS Identity and Access Management（IAM） 角色和策略。
4. 基本了解 Amazon S3 存储桶、前缀和对象。

本演练的 IAM 基础配置

你需要一个 IAM 角色，用于授予 Amazon Quick Automate 访问 S3 存储桶的权限。如果尚未创建，请按照以下说明立即创建：

1. 打开 IAM 控制台。
2. 在导航窗格中，选择 Roles，然后选择 Create role。
3. 对于 Trusted entity type，选择 Custom trust policy，并粘贴以下信任策略：

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Principal": {
        "Service": "quicksight.amazonaws.com"
      },
      "Action": "sts:AssumeRole"
    }
  ]
}

4. 选择 Next，并附加一项权限策略，以授予目标存储桶上的 s3:GetObject、s3:PutObject 和 s3:ListBucket 权限。
5. 为角色命名（例如 QuickAutomate-S3-Role），然后选择 Create role。
6. 复制角色的 Amazon Resource Name（ARN）。在配置 S3 连接器时需要使用它。

创建 IAM 角色并复制角色 ARN 后，即可配置 Amazon Quick Automate 与 S3 存储桶之间的连接。

设置 S3 连接器

确认前提条件后，在 Amazon Quick 控制台中配置 S3 连接器。

1. 在 Amazon Quick 控制台中，选择右上角的用户图标。
2. 打开 Manage Account，然后转到 Permissions > AWS Actions。

图 2：Amazon Quick 控制台中的 AWS Actions 页面

3. 选择 New action，然后选择 Amazon S3。
4. 配置连接器：

• 输入连接器名称（例如 ExampleBucket-S3-Connection）和描述。
• 粘贴你在前提条件步骤中创建的 IAM 角色 ARN。

图 3：Amazon S3 操作连接器配置

5. 控制台会自动验证角色信任策略。确认状态显示连接成功。
6. 与构建和运行工作流的用户及组共享该连接器。

创建自动化组

1. 在 Amazon Quick 控制台中，转到 Automations > Create Group。

图 4：Automations 页面中的 Create Group 选项

2. 为该组命名（例如 RFI-Processing-Group）。
3. 在 Add actions 页面上，选择之前创建的 Amazon S3 连接，然后选择 Next。

图 5：为自动化组选择 Amazon S3

配置好连接器和自动化组后，创建将承载工作流的自动化项目。

创建自动化项目

自动化项目是你描述处理逻辑以及生成的工作流所在的位置。

1. 在 Automations 页面上，选择 Projects 选项卡，然后选择 Create Project。
2. 在 Provide project details 中，输入项目名称，选择创建的自动化组，并可选择添加描述和上传现有文档。
3. 选择 Next: Business case。
4. 业务案例部分为可选项，用于跟踪投资回报率。若要完成该部分，请输入每个案例节省的工时、每年案例数、项目优先级和目标上线日期。
5. 选择 Create。

项目将在摘要页面中打开，其中包含 Summary、Versions 和 Deployments 选项卡，以及一个显示项目当前状态的状态面板。

连接器、自动化组和项目均已准备就绪后，你可以开始描述工作流逻辑。

使用自然语言描述处理逻辑

在 Amazon Quick Automate 中，你可以使用自然语言描述自动化应执行的操作。你可以直接在提示区域输入内容，也可以上传包含需求的文件。AI 助手会解释你的指令并生成可执行的工作流步骤。

步骤 1：连接到 S3 并读取工作表

在自动化项目中，你会看到一个用于描述处理要求的文本区域。输入涵盖数据源、转换逻辑和所需输出的提示。例如：

“使用 ExampleBucket-S3-Connection 连接器连接到我的 S3 存储桶。从 sample-folder 前缀读取文件 SampleDepartment_Version2.xlsx。打开 Strategy 工作表并提取所有调查问题。根据编号和缩进识别主问题和子问题。使用 Serial Number、Category、Question 和 Response Type 列对数据进行结构化。将输出作为 CSV 文件写入同一存储桶中的 output-folder 前缀。”

图 6：在自动化项目中输入处理提示

步骤 2：查看生成的工作流

提交提示后，AI 助手会生成一个多步骤工作流。每个步骤对应你的部分指令：

1. 读取数据 – 连接到 S3、下载工作簿，并读取带有表头的目标工作表。
2. 提取和转换 – 根据编号和缩进识别主问题和子问题，并从列表头提取类别元数据。
3. 结构化输出 – 通过合并父级上下文，将子问题转换为独立记录。定义输出架构。
4. 写入结果 – 将结构化输出作为 CSV 文件保存到指定的 S3 位置。

图 7：生成的多步骤工作流

由于 AI 助手依赖生成式 AI，其在不同运行中生成的具体步骤和措辞可能有所不同。你看到的工作流可能与本示例略有差异，但会根据你的指令涵盖相同的逻辑操作。

步骤 3：运行并验证

选择 Run，针对你的工作簿执行工作流。检查输出以确认：

1. 已提取所有问题（主问题和子问题）。
2. 类别文本与源列表头完全一致。
3. 子问题包含父级上下文。
4. 响应类型按源内容原样保留。

图 8：工作流运行输出

通过对话优化工作流

当输出需要改进时，可使用对话界面请求有针对性的变更。请引用具体步骤，并描述预期行为。

优化提示示例：

1. “读取 Strategy 工作表的步骤还应将列表头捕获为元数据。”。
2. “类别字段与源表头不匹配。请更新提取逻辑，以使用完全一致的表头文本。”。
3. “在子问题转换步骤中，将父级上下文合并到问题文本中。”。

遵循以下迭代周期：

1. 识别问题 – 将输出与你的预期结果进行比较。
2. 请求有针对性的更新 – 使用自然语言描述变更，并引用具体步骤。
3. 查看更新后的工作流 – AI 助手会修改相关步骤并向你展示变更。
4. 重新运行并验证 – 再次执行并确认修复结果。

最佳实践

在将自动化推广到生产环境前，请遵循以下最佳实践：

1. 渐进式验证 – 每次优化后都运行并检查输出，而不是批量进行多项变更。
2. 明确处理边缘情况 – 描述当数据缺失、重复或格式不一致时工作流应如何运行。

Amazon Quick Automate Import/Export 支持从预生产环境到生产环境的受控推广，以及在不同 AWS Region 中的一致复用。

导出版本

1. 打开源自动化。
2. 转到 Versions。
3. 选择…

Build an end-to-end Request for Information (RFI) questionnaire workflow using Amazon Quick Automate to solve a challenge organizations face at every scale. A typical enterprise might handle hundreds of RFI questionnaires each year, and each arrives as a complex, multi-tab workbook packed with hierarchical question sets, category metadata, and varied response types. The volume, variety, and complexity of these documents create a significant operational challenge. Manually extracting, structuring, and resolving RFI questions requires repeated coordination, delays responses, and can introduce errors each time the questionnaire format changes.

Amazon Quick Automate handles multi-agent automation for supported enterprise processes that can span across departments, systems, UI and API interactions, and third-party systems. It uses a team of agents to simplify business process management and cut down on maintenance overhead. You describe your processing goals in plain language, and Quick Automate can produce an executable workflow covering ingestion, transformation, validation, and output. You can refine the workflow through conversation, run it against your data, and promote validated versions across AWS Regions. This approach can help you respond to RFIs faster and reduce formatting inconsistencies and errors. The automation is designed to scale with your workload, and you can adapt it to new questionnaire formats by updating the natural-language instructions.

In this post, we show you how to build an end-to-end automation that reads a multi-tab RFI workbook from Amazon Simple Storage Service (Amazon S3), extracts and structures the questionnaire data, and writes clean output back to Amazon S3. This tutorial uses Amazon Quick Automate and Amazon S3. Review Amazon Quick pricing and Amazon S3 pricing for cost details.

Solution overview

This walkthrough automates the processing of an RFI workbook stored in Amazon S3. When procurement teams receive an RFI questionnaire, they need to extract each question and preserve its category and response type. The goal is to produce a structured dataset that downstream teams can consume, interpret, and act on. In this example we present a solution that automates the process using Amazon Quick Automate. You start by connecting Amazon Quick Automate to the Amazon S3 bucket where your RFI workbooks are stored. Then you create an automation project, which is where you select your data connectors and organize your workflows. With the project in place, you can describe your processing logic in plain language: what to read, how to structure it, and where to write the output. The result is a working automation that reads a multi-tab RFI workbook, extracts and structures the questionnaire data, and outputs a clean comma-separated values (CSV) file. It does this without requiring custom code for common scenarios. The following diagram shows the high-level sequence of steps.

Figure 1: High-level sequence of steps for the RFI workbook automation

The diagram takes you from connecting your data source through deploying the finished automation in a production environment. Check the AWS Regions page for the latest availability.

The solution follows these steps:

1. Set up an Amazon S3 action connector – Connect Amazon Quick Automate to your S3 bucket.
2. Add the S3 integration to an automation group – Share the S3 action with the automation group that will use it.
3. Create an automation project – Create the project where you describe your processing logic.
4. Describe the processing logic – Enter a natural-language prompt and let the generative AI assistant build the workflow.
5. Refine through conversation – Iterate on the generated workflow by requesting targeted changes.
6. Validate results – Run the workflow and verify output in your pre-production (development) AWS account.
7. Promote with Import/Export – Export the validated version and import it into your production account or target AWS Region.

Now that we have reviewed the high-level steps for the solution, we will go over the environment requirements.

Prerequisites

Before you begin, confirm that your environment meets the following requirements.

1. An Amazon Quick Enterprise subscription with Amazon Quick Automate access.
2. An Amazon S3 bucket in the same AWS Region as the Amazon Quick application.
3. Familiarity with AWS Identity and Access Management (IAM) roles and policies.
4. Basic understanding of Amazon S3 buckets, prefixes, and objects.

IAM baseline for this walkthrough

You need an IAM role that grants Amazon Quick Automate permission to access your S3 bucket. If you don’t already have one, follow these instructions to create it now:

1. Open the IAM console.
2. In the navigation pane, choose Roles , then Create role.
3. For Trusted entity type , choose Custom trust policy and paste the following trust policy:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Principal": {
        "Service": "quicksight.amazonaws.com"
      },
      "Action": "sts:AssumeRole"
    }
  ]
}

4. Choose Next and attach a permissions policy that grants s3:GetObject, s3:PutObject, and s3:ListBucket on your target bucket.
5. Name the role (for example, QuickAutomate-S3-Role) and choose Create role.
6. Copy the role Amazon Resource Name (ARN). You will need it when configuring the S3 connector.

With the IAM role created and the role ARN copied, you’re ready to configure the connection between Amazon Quick Automate and your S3 bucket.

Set up the S3 connector

After prerequisites are confirmed, configure the S3 connector in the Amazon Quick console.

1. In the Amazon Quick console, choose the user icon in the upper-right corner.
2. Open Manage Account , then go to Permissions > AWS Actions.

Figure 2: AWS Actions page in the Amazon Quick console

3. Choose New action , then select Amazon S3.
4. Configure the connector:

• Enter a connector name (for example, ExampleBucket-S3-Connection) and a description.
• Paste the ARN of the IAM role you created in the prerequisites step.

Figure 3: Amazon S3 action connector configuration

5. The console validates the role trust policy automatically. Confirm that the status shows a successful connection.
6. Share the connector with users and groups who build and run workflows.

Create an automation group

1. In the Amazon Quick console, go to Automations > Create Group.

Figure 4: Create Group option on the Automations page

2. Name the group (for example, RFI-Processing-Group).
3. On the Add actions page, select the Amazon S3 connection you created earlier, then choose Next.

Figure 5: Selecting Amazon S3 for the automation group

With the connector and the automation group configured, create the automation project that will hold your workflow.

Create an automation project

The automation project is where you describe your processing logic and where the generated workflow lives.

1. On the Automations page, choose the Projects tab, then choose Create Project.
2. In Provide project details , enter a name for the project, select the automation group you created, and optionally add a description and upload existing documentation.
3. Choose Next: Business case.
4. The business case section is optional and is used for tracking return on investment. To complete it, enter hours saved per case, cases per year, project priority, and a target launch date.
5. Choose Create.

Your project opens on a summary page with Summary , Versions , and Deployments tabs, and a status panel showing the current state of the project.

With the connector, the automation group, and the project in place, you are ready to describe your workflow logic.

Describe the processing logic in natural language

In Amazon Quick Automate, you describe what the automation should do using natural language. You can type directly into the prompt area or upload a file with your requirements. The AI assistant interprets your instructions and generates executable workflow steps.

Step 1: Connect to S3 and read a sheet

In the automation project, you see a text area where you describe your processing requirements. Enter a prompt that covers the data source, transformation logic, and the output you want. For example:

“Connect to my S3 bucket using the ExampleBucket-S3-Connection connector. Read the file SampleDepartment_Version2.xlsx from the sample-folder prefix. Open the Strategy sheet and extract all survey questions. Identify main questions and subquestions based on numbering and indentation. Structure the data with columns for Serial Number, Category, Question, and Response Type. Write the output as a CSV file to the output-folder prefix in the same bucket.”

Figure 6: Entering the processing prompt in the automation project

Step 2: Review the generated workflow

After you submit the prompt, the AI assistant generates a multi-step workflow. Each step maps to a part of your instructions:

1. Read data – Connects to S3, downloads the workbook, and reads the target sheet with headers.
2. Extract and transform – Identifies main questions and subquestions by numbering and indentation, and pulls category metadata from column headers.
3. Structure output – Converts subquestions into standalone records by merging parent context. Defines the output schema.
4. Write results – Saves structured output as a CSV file to the specified S3 location.

Figure 7: Generated multi-step workflow

Because the AI assistant relies on generative AI, the exact steps and wording it produces might vary between runs. The workflow you see might differ slightly from this example, but it covers the same logical operations based on your instructions.

Step 3: Run and validate

Choose Run to execute the workflow against your workbook. Review the output to confirm:

1. All questions are extracted (main and subquestions).
2. Category text matches the source column headers exactly.
3. Subquestions include parent context.
4. Response types are preserved as-is from the source.

Figure 8: Workflow run output

Refine the workflow through conversation

When the output needs improvement, use the conversation interface to request targeted changes. Reference the specific step and describe the expected behavior.

Example refinement prompts:

1. “The step that reads the Strategy sheet should also capture column headers as metadata.”.
2. “The category field does not match source headers. Update extraction to use exact header text.”.
3. “In the subquestion transformation step, merge parent context into the question text.”.

Follow this iterative cycle:

1. Identify the issue – Compare output against your expected results.
2. Request a targeted update – Describe the change in natural language, referencing the specific step.
3. Review the updated workflow – The AI assistant modifies the relevant step and shows you the change.
4. Re-run and validate – Execute again and confirm the fix.

Best practices

Before promoting your automation to production, follow these best practices:

1. Validate incrementally – Run and check output after each refinement, rather than batching multiple changes.
2. Handle edge cases explicitly – Describe how the workflow should behave when data is missing, duplicated, or formatted inconsistently.

Amazon Quick Automate Import/Export supports controlled promotion from pre-production to production and consistent reuse across AWS Regions.

Export a version

1. Open the source automation.
2. Go to Versions.
3. Choose Export version for the validated release.

Figure 9: Exporting a validated automation version

4. Copy the generated version link.

Note: The export link expires after a set period. Copy and use it promptly, or generate a new link if it has expired.

Figure 10: Generated export version link

Import a version

1. Open or create an Amazon Quick Automate project in the destination account or AWS Region.
2. Choose Import version.
3. Paste the version link and choose Start.
4. Review version details and complete the import.

Figure 11: Importing an automation version

5. Configure required connectors, credentials, and deployment settings in the target environment.

Figure 12: Configuring the target environment after import

Clean up

1. Delete the automation project and the automation group.
2. Remove the S3 connector if no other workflows use it.
3. Delete test IAM roles and test S3 objects when no longer required.

For detailed cleanup guidance, see Using Amazon Quick Automate.

Conclusion

In this post, you built an end-to-end automation for processing RFI questionnaires stored as multi-tab Excel workbooks using Amazon Quick Automate. You connected your data through an Amazon S3 action connector, described your processing logic in natural language, refined the automation through conversation, and promoted it across AWS accounts and Regions using Import/Export.

You can extend this approach to other business workflows, such as:

1. Intelligent document processing (IDP) – Extract structured data from invoices, contracts, or forms.
2. Report consolidation – Merge multi-sheet financial reports into a unified dataset.
3. UI automation – Automate repetitive browser-based tasks that feed into downstream analytics.
4. Survey data aggregation – Consolidate survey responses from multiple sources into a single output.

For more on automation capabilities, action connectors, and natural-language workflows, see the Amazon Quick documentation. Visit Getting started with Amazon Quick to start using Amazon Quick Automate today.

---

About the authors

Chaytanya Kumar

Chaytanya is an AI Builder at AWS Professional Services, helping enterprise and public sector customers accelerate their generative AI and cloud journeys. With deep expertise in AI enablement, agentic AI, and digital transformation, he builds intelligent solutions that drive measurable business outcomes. Outside work, he recharges by exploring oceans and mountain trails.

Anneline Sibanda

Anneline is an AI Builder at AWS, specializing in the architecture and delivery of agentic and generative AI solutions. With 10+ years of experience delivering solutions for Healthcare, Higher Education and FSI customers, she is a key technical partner for enterprises bridging the gap between innovative concepts and production-ready applications.