跳到主要内容

AI发票解析器 Make

本模块的功能

PDF4me AI发票解析器 从中提取结构化数据 PDF 或者在您的内部添加图片发票 Make 使用人工智能的场景,无需模板设置、字段坐标或维护规则。提供一个 PDFPNGJPG, 或者 JPEG 模块返回 供应商详情、明细项目、金额、日期、税费、货币和付款条款 作为可单独映射的字段。非常适合应付账款自动化、费用管理、财务对账和发票数字化工作流程,可扩展至数百家供应商和多种格式。

相关博客文章(2)

验证您的身份 API 要求

每一个 PDF4me 模块 Make 需要有效 联系创建或选择一个可以容纳您的 PDF4me API 关键在于确保该方案能够安全地验证 AI 发票提取请求。

您不容错过的重要事实

支持 PDFPNGJPG, 和 JPEG 输入
将发票二进制数据映射到 DropboxGoogle Drive电子邮件附件,或 HTTP 将模块直接添加到“文档”字段中。设置发票名称及其正确的扩展名。 PDF 对于电子发票, PNG/JPG/JPEG 适用于扫描或拍摄的发票。
每个输出结果都包含置信度评分。
每次提取都会返回一个 信心 对象包含每个字段的分数(0-1)。使用 Make 筛选低置信度结果,将其路由至人工审核队列 Airtable 或者 Slack 与其直接将数据推送到会计系统,不如避免代价高昂的后续更正。
对非标准发票使用自定义字段键
通过以下方式添加额外的字段名称 自定义字段键 要提取标准发票架构之外的字段,例如 vendor_phone、tax_id、purchase_order_number 和 store_number。每个自定义键都会作为可单独映射的令牌返回。 Make 输出包。
配置 PDF4me AI 发票解析器模块,将“连接”设置为“我的 PDF4me 连接”,“文件”设置为“映射”,并映射发票名称和文档(映射自之前的 Dropbox 下载步骤),还可以为其他发票字段添加“自定义字段键”(可选)。

将文件设置为映射,从前一个模块导入发票名称和文档,还可以选择添加自定义字段键来提取标准发票架构之外的字段。

参数

必需的: 必须始终提供连接号、发票名称和文档。匹配的输入字段(二进制属性, Base64 内容或文件 URL根据所选的输入数据类型,此字段为必填项。自定义字段键是可选的,可以根据具体情况添加。

范围必需的它的作用例子
ConnectionYesPDF4me API connection. Click Add and paste your API key the first time. Reused automatically across all PDF4me modules in your scenarios.My PDF4me connection
Input Data TypeYesHow the invoice is supplied. Binary Data reads from a previous module (most common). Base64 String accepts encoded payload. URL downloads from a public link.Binary Data
Input Binary FieldConditionalBinary content of the invoice file. Required when Input Data Type is Binary Data. Wire from Dropbox, Google Drive, OneDrive Data field, or any HTTP/email module that returns file bytes.1. Data
Base64 Invoice ContentConditionalBase64-encoded invoice content. Required when Input Data Type is Base64 String. Useful when invoice arrives from a webhook or API call that delivers encoded content.JVBERi0xLjQK...
Invoice URLConditionalPublic URL to the invoice file. Required when Input Data Type is URL. The URL must not require authentication.https://example.com/invoice.pdf
Invoice NameYesFilename of the invoice with proper extension (.pdf, .png, .jpg, .jpeg). Used for format detection and AI processing optimization.invoice.pdf
Custom Field KeysNoOptional list of custom field keys to extract beyond the standard invoice schema. Each key returns as an individual mappable token in the output bundle.["vendor_phone", "tax_id", "purchase_order_number"]

输出字段

场地类型它包含什么
invoiceNumberStringExtracted invoice number or reference identifier (e.g. INV-2024-001).
vendorNameStringComplete name of vendor or supplier as printed on the invoice.
vendorAddressStringComplete vendor address including street, city, state, and postal code.
vendorEmailStringVendor email address when present on the invoice.
vendorPhoneStringVendor phone number when present on the invoice.
billToNameStringName of the entity being billed (your company or end customer).
billToAddressStringComplete billing address.
invoiceDateStringInvoice issue date in ISO 8601 format (YYYY-MM-DD).
dueDateStringPayment due date in ISO 8601 format.
subTotalNumberSubtotal amount before taxes.
totalNumberFinal total amount including all taxes and adjustments.
currencyStringCurrency code in ISO 4217 format (USD, EUR, GBP, INR, CHF, etc.).
paymentTermsStringPayment terms and conditions (e.g. Net 30, Due on receipt).
lineItemsArrayArray of line items with description, quantity, unitPrice, total, and HSN code when present.
cgst / sgst / igstNumberIndian tax fields when present on the invoice: central, state, and integrated GST amounts.
confidenceObjectPer-field confidence scores (0-1). Use a Make Filter to route low-confidence extractions for human review.
jobIdStringPDF4me job identifier for audit trail and support tickets.
successBooleanTrue if the AI extraction completed successfully; false if it failed.
messageStringHuman-readable status message confirming success or describing the error.

快速设置

  1. 添加 PDF4meAI发票解析器 致你 Make 设想。
  2. 选择 联系 (或点击) 添加 用你的 PDF4me API 钥匙)。
  3. 在下面 文件, 选择 地图 显示发票名称和文档字段。
  4. 地图 发票名称 到源文件的文件名(包含) .pdf.png.jpg, 或者 .jpeg)。
  5. 地图 文档 从源模块获取二进制内容:通常是 数据 领域 DropboxGoogle DriveOneDrive, 或者 Gmail 附件模块。
  6. (可选) 自定义字段键 对于超出标准发票架构的字段(例如, purchase_order_numbertax_idstore_number)。
  7. 点击 节省 并运行该场景。输出包会暴露所有提取的字段:将它们连接到 QuickBooksXeroAirtableGoogle SheetsSlack 批准,或 HTTP 模块。

工作流程示例

工作流程示例Common Make scenario patterns using AI-Invoice Parser.
供应商邮件 → AI发票解析器 → QuickBooks 账单
  1. Gmail 当收到来自已知供应商地址的电子邮件时触发,该邮件包含以下内容: PDF 发票附件。
  2. 附件将传递给 AI 发票解析器。输出结果包含供应商名称、发票编号、明细项目、总计、货币和付款条款。
  3. A Make 过滤器检查总字段的置信度,提取低于 0.85 的数据,并将其路由至 a Slack 用于人工审核的渠道。
  4. 高置信度发票将发布到 QuickBooks 生成一张新账单,自动应用供应商、到期日、明细项目和账户映射。
  5. 一个 Airtable 日志记录会捕获发票编号、供应商、金额、状态和时间戳,以便进行审计报告。
Dropbox 监视文件夹 → 批量解析 → Google Sheets 账簿
  1. A Dropbox 监视文件会在新发票生成时触发。 PDF 已放入会计收件箱文件夹。
  2. AI发票解析器提取完整的发票信息。自定义字段键添加了采购订单号和部门代码,以便逐行记录成本分配情况。
  3. A Google Sheets 添加行模块会根据发票、供应商、总计、货币、采购订单、部门和到期日,追加一个规范化的账簿行。
  4. 第二个模块遍历每个项目,并将其添加到单独的“项目”工作表中,以进行支出分析。
  5. 发票 PDF 已移至按年月划分的子文件夹中进行存档,以便下次运行时保持收件箱干净整洁。
高价值发票 → AI解析 → 多步骤审批
  1. webhook 从内部门户接收发票。文档二进制文件被传递给 AI 发票解析器。
  2. 路由程序会根据提取的总金额进行分支。超过 10,000 美元的发票需要首席财务官 (CFO) 批准;超过 1,000 美元的发票需要经理批准;低于 1,000 美元的发票会自动批准。
  3. 每个分支机构发送一个 Slack 包含供应商、总计、明细项目以及指向原始文件的链接的审批消息 PDF
  4. 收到批准后,发票将过账到 Xero 根据自定义字段键(部门代码、项目代码)生成具有正确总账科目的账单。
  5. 最后 Airtable 行记录审批人、审批时间戳和最终处理结果,用于合规性审计跟踪。

常见问题解答

What invoice fields does the AI-Invoice Parser extract?+
The module returns a complete structured set: invoiceNumber, vendorName, vendorAddress, vendorEmail, vendorPhone, billToName, billToAddress, invoiceDate, dueDate, subTotal, total, currency (ISO 4217), paymentTerms, and a lineItems array with description, quantity, unit price, total, and HSN code per row. Indian tax fields (cgst, sgst, igst) appear when present. Every field comes with a confidence score so you can decide which extractions to trust automatically and which to route for human review.
What file formats does AI-Invoice Parser support in Make?+
PDF, PNG, JPG, and JPEG. Digital PDFs (born-digital invoices generated by ERPs or invoicing software) deliver the highest extraction accuracy. Scanned PDFs and photographed invoices work too: PDF4me applies AI OCR before semantic extraction. Always set the Invoice Name with the correct file extension so the module detects the format properly and applies the right processing pipeline.
How do I extract fields that are not in the standard invoice schema?+
Use the Custom Field Keys parameter. Add keys such as purchase_order_number, store_number, department_code, vendor_tax_id, gst_number, or any other field your accounting workflow needs. Each custom key returned appears as an individually mappable token in the Make output bundle and can be wired directly into accounting, database, or ERP modules without any additional parsing.
How should I handle low-confidence extractions in production scenarios?+
The confidence object returns per-field scores from 0 to 1. Add a Make Filter immediately after the parser to route low-confidence invoices (e.g. total confidence below 0.85, or any critical field below 0.90) into a human review queue in Airtable, Slack, or a dedicated approval workflow. High-confidence results go straight to QuickBooks, Xero, NetSuite, or your ERP. This pattern eliminates costly downstream corrections while still scaling to thousands of invoices per month.
Can AI-Invoice Parser process invoices in different languages and currencies?+
Yes. The AI is trained on multilingual invoices and handles English, German, French, Spanish, Italian, Portuguese, Dutch, and other major business languages without any extra configuration. Currencies are returned in ISO 4217 format (USD, EUR, GBP, INR, CHF, JPY, etc.) so downstream conversion or reconciliation logic always has a normalized code regardless of the source invoice format.

相关模块

获取帮助