跳到主要内容

AI税务文档解析器 Power Automate

AI税务文档解析器 是一个 Power Automate 采取行动 PDF4me 它利用人工智能读取税务文件,并返回表格类型、纳税年度和一个结构化字段字典。可用于 W-2、1099、1040、1098、1095 或 W-4 表格,将数据直接导入。 SharePointExcelDynamics 365 或 Power BI,无需手动输入。

此操作的作用

PDF4me AI税务文档解析器 阅读一份税务文件(PDFPNGJPG, 或者 JPEG)和返回 表单类型纳税年度,以及一个 字段 提取值的字典。支持 W2、1099、1099-SSA、1040、1040-SR、1040-NR、1098、1098-E、1098-T、1095A、1095C、W-4 和 UnifiedTaxUS。可选择性地设置 主体/税收模型 为了在已知形状上获得更高的精度,或者 正文/自定义字段键 提取标准集之外的其他字段。

相关博客文章(1)

我如何验证我的身份 Power Automate 流动?

每一个 PDF4me 行动 Power Automate 需要有效的连接。创建或选择一个包含您数据的连接。 PDF4me API 关键在于确保流程能够安全地验证 AI 税务文件处理请求。

您不容错过的重要事实

始终需要填写两个字段。
文件内容(文档字节)和正文/文档/名称(带扩展名的文件名)。其他所有内容都位于“高级参数”下,且为可选参数。
Body/TaxModel 只是一个提示,而不是一个筛选器。
留空的话,解析器仍然会自动检测表单类型,并在 formType 中返回。只有将其设置为已知代码(W2、1040、1099 等)才能提高字段识别的准确性。
在信任输出结果之前,请检查警告和回退方案。
非空的警告数组或 fallbackUsed: true 表示 AI 遇到了置信度低的字段或异常的布局。在对合规性要求较高的流程中,应将这些情况路由至人工审核。
Power Automate AI - 税务文档解析器操作已配置文件内容、正文/文档/名称和高级参数,显示正文/税务模型和正文/自定义字段键,包含一个项目

映射文件内容和正文/文档/名称,然后展开正文/税务模型和正文/自定义字段键的高级参数。

参数

必需的: 文件内容和正文/文档/名称必须始终提供。正文/税务模型和正文/自定义字段键是可选的高级参数。

范围必需的它的作用例子
File ContentRequiredBinary content of the tax document, mapped from a previous action (Get file content, SharePoint, OneDrive, email attachment). Supports .pdf, .png, .jpg, .jpeg.body('Get_file_content')
Body/Document/NameRequiredFilename of the tax document including its extension. Used for format detection and processing.tax_document_2024.pdf
Body/TaxModelOptionalAdvanced parameter. Known form type to improve extraction accuracy. Leave blank to let the AI auto-detect the form type.W2
Body/customFieldKeysOptionalAdvanced parameter. Array of additional field names to extract beyond the standard tax data set.["customField1", "customField2"]

支持的税务表格

身体/税务模型代码典型领域
Wage statementW2Wages, federal/state tax withheld, Social Security wages, Medicare wages.
Information returns1099, 1099-SSAIncome amounts, payer information, tax withheld, recipient information.
Individual return1040, 1040-SR, 1040-NRAdjusted gross income, taxable income, tax owed, refund amount, filing status.
Interest and education1098, 1098-E, 1098-TMortgage interest paid, student loan interest, tuition payments.
Health coverage1095A, 1095CHealth insurance coverage information, employer information.
WithholdingW-4Withholding allowances, additional withholding, filing status.

输出字段

场地类型它包含什么
formTypeStringThe detected or specified tax form type, for example W2, 1099, or 1040.
taxYearStringThe tax year the document applies to, for example 2024.
fieldsObjectDictionary of extracted values, keyed by field name, plus any custom fields requested.
warningsArrayMessages flagging potential data quality issues or extraction challenges. Empty when none.
fallbackUsedBooleantrue if fallback extraction logic was used for this document.
rawOcrTextStringThe raw OCR text extracted from the document.
jobIdStringUnique identifier for the AI processing job.
successBooleantrue if extraction completed, false if it failed.
messageStringDescriptive result message for the processing run.

如何在 AI 税务文档解析器中设置 AI 税务文档解析器 Power Automate

  1. 添加 PDF4meAI税务文档解析器 致你 Power Automate 流动。
  2. 地图 文件内容 来自之前的操作(SharePointOneDrive或电子邮件附件)。
  3. 主体/文件/名称 文件扩展名(.pdf、.png、.jpg 或 .jpeg)。
  4. (可选)扩展 高级参数 并设置 主体/税制模型 使用已知表单代码以提高准确性。
  5. (可选) 正文/自定义字段键 对于标准集之外的任何字段。运行流程,解析结果将返回如下内容 表单类型纳税年度, 和 字段

典型配置

工作流程示例Common Power Automate flow patterns using AI-Tax Document Parser.
W2摄入量到HR系统
  1. 当员工将 W2 表格上传到系统时,流程将被触发。 SharePoint 收据。
  2. AI税务文档解析器将Body/TaxModel设置为W2以达到更高的准确率。
  3. 提取的工资和预扣税项被写入一个 SharePoint 列表。
  4. 条件检查警告数组,并将置信度低的行路由到人工审核。
1099 批次核对
  1. 每收到一个 1099 附件,就会触发一个流程。 Outlook
  2. AI税务文档解析器将付款人信息和收入金额提取到字段字典中。
  3. 数据被插入到一个 Excel 会计对账工作簿。
1040 返回存档
  1. 当已提交的 1040 表格进入记录库时,就会触发一个流程。
  2. AI税务文件解析器提取申报状态、调整后总收入和退税金额。
  3. A Dynamics 使用提取的纳税年度和汇总字段创建 365 记录,用于审计查找。

实用技巧

Set Body/TaxModel when you already know the form
Leaving it blank still works via auto-detection, but a known code sharpens field recognition, especially on scanned or lower-quality images.
Always branch on success and warnings
Check success before writing extracted data downstream, and inspect warnings for any fields the AI flagged as uncertain before treating the output as final.
Custom fields extend, they do not replace
Body/customFieldKeys adds extra entries to the fields dictionary alongside the standard form fields, it does not narrow or filter the standard set.
rawOcrText is useful for debugging
When a field looks wrong, compare it against rawOcrText to see whether the source scan quality, not the parser, is the root cause.
Match the file extension to the real format
Body/Document/Name must carry the true extension (.pdf, .png, .jpg, .jpeg) of the mapped File Content, a mismatched extension can affect processing.

速查表

场地价值
ActionAI-Tax Document Parser
Required fieldsFile Content, Body/Document/Name
Optional advanced fieldsBody/TaxModel, Body/customFieldKeys
Supported form codesW2, 1099, 1099-SSA, 1040, 1040-SR, 1040-NR, 1098, 1098-E, 1098-T, 1095A, 1095C, W-4, UnifiedTaxUS
Supported file types.pdf, .png, .jpg, .jpeg
OutputformType, taxYear, fields, warnings, fallbackUsed, rawOcrText, jobId, success, message

常见问题

Which tax forms does the AI-Tax Document Parser support?+
W2, 1099, 1099-SSA, 1040, 1040-SR, 1040-NR, 1098, 1098-E, 1098-T, 1095A, 1095C, W-4, and UnifiedTaxUS. Set Body/TaxModel to the exact form code to improve extraction accuracy for a known type.
Do I have to tell the parser which tax form type I am sending?+
No. Body/TaxModel is optional. Leaving it blank lets the AI detect the form type automatically and return it in the formType output field. Setting it improves accuracy when you already know the form.
How do I extract fields that are not part of the standard tax data set?+
Populate Body/customFieldKeys with an array of the field names you need. Each key you list is extracted alongside the standard fields dictionary, in addition to the built-in form fields. See <a href="https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/prebuilt/tax-document" target="_blank" rel="noopener noreferrer">Microsoft's tax document data extraction reference</a> for the kind of structured fields these forms typically expose.
What does the warnings array in the output mean?+
Warnings flag potential data quality issues, such as a low-confidence field or an unusual layout, without failing the run. Check the warnings array in automated pipelines before trusting extracted values without review.
Can I process scanned images as well as digital PDFs?+
Yes. File Content accepts .pdf, .png, .jpg, and .jpeg formats, so scanned tax documents work the same way as digitally generated PDFs. Retention requirements for the underlying tax records should still follow guidance such as the <a href="https://www.irs.gov/businesses/small-businesses-self-employed/how-long-should-i-keep-records" target="_blank" rel="noopener noreferrer">IRS recordkeeping guidelines</a>.

行业应用案例及应用

  • 税务申报处理利用提取的税务文件数据实现税务申报自动化处理
  • 报税利用提取的税务表格信息自动准备税务申报表
  • 税务合规:利用提取的税务文件数据确保税务合规性
  • 税务申报根据提取的税务文件信息生成税务报告
  • 审计准备利用提取的税务文件数据准备审计工作

相关操作

在其他平台上执行相同任务

获取帮助