PurchaseOrderOCRbyKoncile

Best purchase order OCR software: 10 tools compared in 2026

Published 2026-10-29 · Updated 2026-11-11 · By the Koncile team

Ten purchase order OCR and document processing tools, from a one cent per page cloud API to an $18,000 a year enterprise platform, compared on what they return for each line, what they check, what they cost and where each one falls short. Facts were read on each vendor's own site on 29 October 2026; Koncile publishes this guide and is ranked first, with its weaknesses written as plainly as the others.

Stack of diverse paper purchase orders next to a tablet displaying OCR software with a stylus hovering over it.

The short answer, by situation

Best overall for line items, totals and matching in one call: Koncile. Every line of a PDF, scanned, faxed or handwritten PO comes back as structured data, with the totals recomputed, each line matched to your catalog and a confidence score on every field, from 20 free credits and a public price of €199 per month for 1,000 pages.

Best developer API for customer POs: Mindee. A dedicated Purchase Order OCR API aimed at sales order entry, with line items, published accuracy figures and an entry plan at $59 per month ($53 billed annually). Catalog matching is code you write yourself.

Best if you want a PO page, workflows and three way matching in one platform: Nanonets. The product page describes matching POs with invoices and receipts. Pricing is per workflow block rather than per page, so budget a test month.

Best for a large AP department with a budget: Rossum (now part of Coupa). Two, three and four way matching on invoices, handwriting in 30 languages, a published entry price of $18,000 per year with a one year minimum.

Best low cost option for a mailbox of supplier POs: Parseur. A permanent free plan of 20 pages a month, then €49 per month for 100 pages. It prepares the data; matching and the ERP connection are yours to build.

Best if your team already runs on AWS and only needs tables: Amazon Textract. One cent per page on AnalyzeExpense, but there is no purchase order model: you get invoice and receipt fields, in English only, and everything else is code.

Update of October 2026: first version of this guide. Prices and claims are those displayed on each vendor's site on 29 October 2026; we correct them when a page changes. Update of 2 November 2026: pricing pages read again; Veryfi no longer states a trial length, so its free plan of 100 documents a month is listed instead. If you are new to the topic, start with the purchase order OCR overview on our home page.

How we ranked these tools

We read the purchase order page, the pricing page and, when there is one, the security page of each vendor on 29 October 2026, and kept only what the vendor states itself. Where a site does not say something, we write that it does not say it. We did not run a benchmark on a shared test set: every accuracy figure below is the vendor's own, which is also true of ours. Each tool was scored on five weighted criteria.

  1. Line items on real layouts (25 %). Does the tool return one row per line with SKU, description, quantity, unit of measure and unit price? Does it say anything about scans, faxes, handwriting and tables that run over several pages?
  2. Controls on the data (25 %). Are totals checked? Can each line be matched to an item catalog or a price list? Is two or three way matching with delivery notes and invoices built in, sold as a module, or left to you?
  3. Integration and intake (20 %). Email intake, API, webhooks, the ERPs named on the page, and whether connectors are included or reserved to a quote.
  4. Public price and trial (20 %). Can a buyer know what 1,000 POs a month will cost without a sales call? How much can be tested for free, and for how long?
  5. Data security (10 %). Certifications stated by the vendor, hosting regions, and what happens to documents after processing.

Koncile publishes this guide and is ranked first. We wrote its weaknesses as plainly as the others', and every claim about a competitor links to the page where we read it.

Quick comparison table

Read on each vendor's own pages on 29 October 2026. Prices excluding tax, entry plan only.
ToolTypePO modelLine itemsMatchingFree trialEntry price
KoncileExtraction + controls, API and webYesYes, with UoM splitCatalog, price list, PO / delivery note / invoice20 credits€199/month, 1,000 pages
MindeeExtraction APIYes (customer POs)YesLeft to your code14 days$59/month, 500 pages
NanonetsIDP platform + APIYesYesThree way matching announced$50 of credits$100/month, 100 credits
Rossum (Coupa)Enterprise IDPNo PO pageYes2, 3 and 4 way on invoices14 days$18,000/year
DocsumoIDP platformNo PO pageYesEnterprise plan only14 days, 1,000 pagesQuote
Doxis (Klippa)IDP platform + APIYesYesTwo or three way module€25 of creditsQuote
ABBYY VantageEnterprise IDPPaid skill, 4 countriesYesBuyer and supplier lookupOn requestNot published
VeryfiExtraction APIYesYesTwo and three way announcedFree plan, 100 docs a month$500/month minimum
ParseurEmail and document parserYesYesNot built in20 pages/month free€49/month, 100 pages
Amazon TextractCloud APINo (invoices, receipts)Yes, invoice fieldsNo3 months, 100 pages/month$0.01/page

1. Koncile: best overall for line items, totals and matching in one call

Type: document extraction with a control layer, REST API, web interface, email intake. PO model: yes, header fields plus every line item. Matching: item catalog, price list, and PO matching against delivery note and invoice. Entry price: 20 free credits, then Advanced at €199 per month excluding tax for 1,000 credits, 1 credit per extracted page (pricing). Security: SOC 2 Type II, GDPR.

One response, header and lines. Koncile's purchase order template returns the PO number, date, buyer, vendor, bill to and ship to addresses, delivery date, payment terms, FOB or Incoterms, subtotal, tax and total, then one row per line with SKU, description, quantity and unit of measure as two separate fields, unit price and line total. "10 cases of 6" comes out as a quantity and a pack size, which is what an ERP needs. You add your own fields (cost center, GL code, job site) in the interface, without a deployment.

The control layer is the difference with a pure OCR. Quantity times unit price is recomputed on each line, the sum of the lines is compared with the subtotal, subtotal plus tax with the total. Each line is then matched to your item catalog from the customer's SKU or description, by exact match, fuzzy match or AI similarity, with a cross reference table for descriptions that differ too much. Unit prices can be compared with the customer's contract price, and a PO received twice is flagged by duplicate detection.

Built for the three documents, not only the PO. Delivery notes and invoices are read with templates from the same library, so quantity received can be compared with quantity ordered and price billed with price ordered, line by line. A second invoice on a PO that is already closed shows up before it is paid twice.

Handwriting and bad scans. Order books filled in by hand, faxes and phone photos are read as a whole rather than character by character. Koncile says plainly what still degrades the result: cramped handwriting, low resolution scans and photos taken at an angle.

Published accuracy. Koncile publishes a success rate above 99 % on the key fields of invoices and purchasing documents. Like every figure in this guide, it is the vendor's own measurement.

Strengths

  • Line items, recomputed totals and catalog matching in the same response
  • Reads handwritten, faxed and skewed POs, and says what degrades them
  • PO, delivery note and invoice from one template library, for three way matching
  • Public price per page and 20 free credits without a card
  • Confidence score on every value, with routing to human review below your threshold
  • Documents never used to train shared models; dedicated retraining only on Enterprise, private to your account

Weaknesses

  • Not a procurement suite: it does not create, approve or send purchase orders
  • Not an AP suite: no payment, no approval workflow
  • No native app inside NetSuite, Dynamics or Odoo: orders go through the API, webhooks, Make, n8n, Zapier or Power Automate
  • Prices are in euros, which US buyers must convert
  • The 20 free credits are given once, not every month like Parseur's free plan
  • Cramped handwriting and low resolution scans remain the main sources of error

Our verdict

Koncile is made for teams that key purchase orders into an ERP today, on the sales side or the AP side, and want the lines checked before they move: distributors that want sales order automation for the customer POs they receive by email, AP teams matching invoices to orders, buyers comparing confirmations with what they ordered. If you need a tool that also issues POs and routes approvals, pair it with your procurement suite or look at Rossum inside Coupa: our page on purchase order processing software separates the two jobs.

2. Mindee: best developer API for customer purchase orders

Type: extraction API with ready models or a custom schema, SDKs in Python, .NET, Node.js, Java, Ruby and PHP. PO model: yes, the page is titled "Purchase Order OCR API for sales order automation". Matching: not built in. Entry price: Starter at $59 per month billed monthly, or $53 per month billed annually ($26 platform fee plus $27 for 500 credits a month), on the US pricing page read on 4 November 2026; 1 page = 1 credit, 1.5 with confidence scores (pricing). Security: SOC 2 Type II infrastructure, GDPR, EU or US hosting.

A clean schema for the order entry use case. The Purchase Order OCR API returns line items with SKU, description, quantity, unit price and total, and handles tables that run over several pages. Email intake goes through Make, Zapier or n8n, with NetSuite used as the example workflow.

Published accuracy figures, which few vendors do: "99.8% Accuracy on line items" and "96.1% Field-level accuracy on header fields", in Mindee's own words and on its own measurements.

Matching is your code. Mindee writes that comparing the extracted lines with your price list and stock "takes a few lines of code". For a developer, that is honest and flexible; for an order desk without engineers, it is the part that is missing.

Strengths

  • Dedicated PO model aimed at sales order entry
  • Published accuracy on line items and header fields
  • Low published entry price and a 14 day trial without a card
  • SDKs in six languages and no code connectors

Weaknesses

  • No catalog matching, price check or three way matching built in
  • Handwriting not mentioned on the PO page
  • Email intake requires an automation tool in between
  • Starter plan limited to one seat and 500 credits

Our verdict

For a developer who wants clean PO JSON at a known monthly price and is ready to code the matching, Mindee is a sound choice. Teams that want the checks without writing them should compare it with Koncile on twenty of their own POs.

3. Nanonets: best platform with a PO page and workflow blocks

Type: IDP platform with an API and workflows built from "blocks". PO model: yes, a purchase order OCR page. Matching: announced, "Match POs with invoices and receipts automatically" and three way matching. Entry price: Starter is free with $50 of credits, then $100 per month for 100 credits; extraction blocks cost $0.30 each (pricing). Security: SOC 2, HIPAA, GDPR and ISO listed on the pricing page.

The broadest workflow story. Upload by email, API, desktop, Drive or Dropbox; validation against the supplier and product masters of the ERP is described in the PO page FAQ; the ERPs named include SAP, Oracle, NetSuite, Dynamics 365, Infor, Sage, Coupa, Ariba, QuickBooks and Xero.

Pricing by block, not by page. Nanonets bills each block a workflow runs: $0.02 for a simple block, $0.10 for standard AI, $0.30 for complex AI including extraction. Its own pricing page says a typical invoice workflow "runs 4-6 blocks per document". The cost of one PO therefore depends on how you build the workflow, which makes a test month the only reliable estimate.

Connectors live in the higher plans. "ERP & database integrations" appear in the Growth plan, and Salesforce, SAP and Oracle connectors in Enterprise, both on quote.

Strengths

  • Dedicated PO page with line items and three way matching announced
  • Email, API and cloud drive intake
  • Long list of named ERPs
  • $50 of free credits without a card

Weaknesses

  • No price per page: the cost per PO depends on the blocks in your workflow
  • ERP connectors reserved to plans on quote
  • Accuracy given as a range ("90-98%+") on the PO page
  • Handwriting cited only as a generic buying criterion, not as a Nanonets feature

Our verdict

Nanonets suits teams that want to assemble a full PO workflow in one platform and can spend a month measuring the real cost per document. If a fixed price per page matters to your budget, compare the block count of your workflow with a per page tool, or read our page on Nanonets alternatives for purchase orders.

4. Rossum: best for a large AP department with a budget

Type: enterprise IDP; the site banner reads "Coupa Acquires Rossum". PO model: no dedicated PO page; document types listed are invoices, bills of lading and sales orders. Matching: "2/3/4-way matching" on invoices, master data matching on the Business plan. Entry price: Starter "Starting at $18,000 per year", one year minimum (pricing). Security: ISO 27001, ISO 42001, SOC 2 Type II, HIPAA; AWS hosting in Europe by default, US and Japan options.

Strong on invoices and matching. Rossum's AP page promises to "validate invoices against any set of documents" with two, three or four way matching, and its Aurora engine reads handwriting in 30 languages and long tables of line items.

POs as inputs, not as a product. Purchase orders appear as a document Rossum receives in the order management FAQ, and sales orders have their own page, but there is no purchase order page or model presented as such.

The price of an enterprise contract. The Starter plan includes unlimited seats, email and API ingestion, the validation screen and a 12 month archive. Master data matching and most integrations (SAP, Coupa, Workday, Oracle, NetSuite, Dynamics, QuickBooks) belong to the Business plan, on quote. With the Coupa acquisition, buyers already on Coupa get a natural path.

Strengths

  • Two, three and four way matching on invoices
  • Handwriting in 30 languages, long line item tables
  • Strong certification list, European hosting by default
  • 14 day free trial

Weaknesses

  • No purchase order model presented
  • Entry at $18,000 per year with a one year minimum
  • Master data matching and most connectors on the Business plan, on quote
  • Product direction now tied to Coupa

Our verdict

Rossum fits a large AP department, ideally on Coupa, that wants invoice matching at scale and has an annual budget for it. For a mid size order desk keying customer POs, the entry price is hard to justify.

5. Docsumo: best if invoices come first and POs second

Type: IDP platform with API. PO model: no dedicated page; the purchase orders URL redirects to the accounts payable page, which says "Docsumo reads purchase orders, delivery slips and receipts as well as invoices". Matching: line matching on the Enterprise plan. Entry price: not published, Business and Enterprise on quote (pricing). Security: SOC 2 Type 2, ISO 27001:2022, HIPAA, GDPR; EU and UK regions available.

AP first. Docsumo's accounts payable page sends every field and line item to the ERP with a confidence score, reads handwritten amounts, and names NetSuite, SAP, Oracle, Dynamics 365, Sage Intacct, QuickBooks and Xero. It states "99% field-level accuracy on 250+ document types in production".

Matching behind the Enterprise plan. "On the Enterprise plan, each invoice line is matched to what was ordered", together with master data lookup and database mapping.

A generous but guided trial: 14 days and up to 1,000 pages, which starts with a short call.

Strengths

  • Reads POs, delivery slips, receipts and invoices
  • Confidence score per field and line
  • 1,000 page trial
  • Long certification list, EU and UK hosting

Weaknesses

  • No purchase order page or model presented
  • Line matching only on the Enterprise plan
  • No published price
  • The trial starts with a sales call

Our verdict

Docsumo is a good fit for an AP team whose priority is invoices, with POs as supporting documents, and who will negotiate an Enterprise plan. Teams that need PO line items and matching without a quote can test a per page tool first: our page on Docsumo alternatives compares seven of them.

6. Doxis (formerly Klippa): best for matching as an add on module

Type: IDP platform (Doxis AI.dp) with API and SDK; the site reads "Klippa is now Doxis". PO model: yes, a purchase order page. Matching: "two/three-way matching" as a module. Entry price: on quote; €25 of free credits at signup. Security: ISO 27001 and ISO 9001, GDPR, SOC 2 stated on the pricing request page.

PO extraction with email parsing. Supplier name, PO number, dates, line items, quantities, prices and totals; incoming POs parsed from connected mailboxes or FTP; SAP, Oracle, Dynamics, QuickBooks, NetSuite and Xero among "50+ integration options". Doxis announces "99% accuracy" on the PO page.

Modules on top. Two or three way matching and fraud detection are presented as modules you connect, which keeps the base product simple and the final price dependent on what you add.

A brand in transition. The site still runs on klippa.com while presenting itself as Doxis; contracts and support may move under the new name.

Strengths

  • Dedicated PO page with line items and email parsing
  • Matching and fraud detection available as modules
  • Many integration options
  • €25 of free credits to test

Weaknesses

  • No published price; the pricing form starts at companies of 50 to 99 employees
  • Handwriting not mentioned on the PO page
  • Rebranding from Klippa to Doxis in progress
  • Hosting regions not stated on the pages we read

Our verdict

Doxis suits a European mid size company that wants PO extraction now and matching later as a module, and is comfortable with a quote. Smaller teams will find the entry harder than a self serve tool.

7. ABBYY Vantage: best if you already run ABBYY

Type: enterprise IDP platform with a Marketplace of skills. PO model: a paid "Purchase Order Processing" skill (classification plus US and European models). Matching: buyer and supplier detection or matching against an existing database; two or three way matching not mentioned on the skill page. Entry price: not published. Security: SOC 2 certified cloud instances in Europe, USA or Australia, ISO 27001, on premises or private Azure possible.

Mature platform, narrow skill. The purchase order skill extracts buyer, order number, order date, ship to and bill to addresses and line items. Its coverage is four countries (United States, Germany, France, Spain) and four languages.

Handwriting and intake. Vantage reads handwritten text from forms and documents and takes documents by email, with REST API and connectors to RPA, BPM and ERP systems, none of them named on the pages we read.

Pricing on request. The skill is marked "Purchase terms: Paid" and the Vantage pricing page returned a 404 on the day we checked.

Strengths

  • Mature enterprise platform, on premises possible
  • Handwriting supported
  • Database matching for buyers and suppliers
  • Cloud regions in Europe, USA and Australia

Weaknesses

  • PO skill limited to four countries and four languages
  • No published price, trial on request
  • No ERP named on the pages we read
  • "90% accuracy at the start" on pretrained skills, by ABBYY's own words

Our verdict

ABBYY makes sense when the company already licenses Vantage or FlexiCapture and wants POs in the same platform. For a new project, the narrow PO coverage and the absence of public pricing weigh against it.

8. Veryfi: best API if you also process receipts

Type: extraction API with SDKs and a "Day-1 Ready Pre-Trained PO (Purchase Order) AI Model". Matching: two way matching and a three way matching section on the PO page; product matching as an option on the Growth plan. Entry price: Starter "$500/mo minimum commitment"; POs fall under "100+ Other Docs", price on request (pricing). Security: SOC 2 Type 2, GDPR, CCPA, HIPAA.

A full line item schema. Veryfi's PO page lists SKU, description, quantity, UoM, unit price and total for each line, and "Handwritten Fields" among the fields returned. It claims "99%+ accurate data from any document type".

Strong on receipts and invoices. Published per document prices exist for invoices ($0.16) and receipts ($0.08); purchase orders are quoted separately.

Free to start, expensive to commit. A 14 day trial and a free plan of 100 documents a month, then a $500 monthly minimum on Starter.

Strengths

  • Pretrained PO model with UoM on each line
  • Two and three way matching announced
  • Free plan of 100 documents a month
  • SOC 2 Type 2, HIPAA

Weaknesses

  • PO price not published
  • $500 per month minimum commitment on the first paid plan
  • No ERP named on the PO page
  • Email intake not mentioned on the PO page

Our verdict

Veryfi is a good choice for a product team that already processes receipts and invoices with it and wants POs in the same API. For POs alone, ask for the PO price before planning on the $500 minimum.

9. Parseur: best low cost parser for a PO mailbox

Type: email and document parser with AI (Vision AI, Text AI or templates). PO model: yes, a purchase order automation page. Matching: not built in; PO, receipt and invoice are extracted in the same format "so the three can be compared field by field". Entry price: free plan of 20 pages every month, then Micro at €49 per month for 100 pages. Security: GDPR, SOC 2 Type II, data possibly stored in the EU.

Email first. You send POs to a dedicated mailbox and Parseur extracts "repeating table rows as structured line items". Exports go out as JSON, CSV or XLSX, or through Power Automate, Zapier, Make, webhooks and the API.

Honest about matching. Parseur prepares the three documents for comparison but does not run the comparison itself. It cites "up to 99.9% accuracy on purchase order data" from internal testing and customer reporting.

The cheapest permanent free plan in this list: 20 pages every month, without time limit.

Strengths

  • Permanent free plan of 20 pages a month
  • Low published entry price
  • Email intake at the core of the product
  • SOC 2 Type II and GDPR

Weaknesses

  • No matching, price check or totals check built in
  • No ERP named: every connection goes through an automation tool or the API
  • Handwriting not mentioned
  • Entry plan limited to 100 pages a month

Our verdict

Parseur is the right first step for a small team with a few dozen supplier POs a month that wants them out of the inbox and into a spreadsheet. As soon as the lines must be checked against a catalog or an invoice, it becomes one piece of a pipeline you assemble.

10. Amazon Textract: best raw building block on AWS

Type: cloud API (Detect Document Text, Analyze Document, Analyze Expense, Analyze ID, Analyze Lending). PO model: none; AnalyzeExpense is for invoices and receipts, from which it returns a PO number and an order date. Matching: not mentioned. Entry price: AnalyzeExpense $0.01 per page up to one million pages, Analyze Document Tables $0.015 per page (pricing). Security: HIPAA eligible, PCI, ISO, SOC.

Cheap pages, expensive engineering. At one cent a page, Textract is the lowest raw price here. What you get is invoice and receipt fields (item, product code, quantity, price, unit price) or generic tables, not a purchase order schema.

English only on the useful parts. The FAQ states that handwriting, invoices and receipts are supported in English only.

Everything else is yours: PO specific fields, totals checks, catalog matching, review screens, email intake and the ERP connection. Teams with AWS engineers sometimes prefer this; order desks rarely do.

Strengths

  • Lowest price per page
  • Pay as you go, three month free tier of 100 pages a month
  • Handwriting supported in English
  • AWS compliance programs

Weaknesses

  • No purchase order model
  • Invoices, receipts and handwriting in English only
  • No matching, no review screen, no email intake
  • Every business rule must be developed

Our verdict

Textract is a component, not a purchase order product. Choose it if you have engineers, run on AWS and want to own the whole pipeline; otherwise its low page price is offset by the code you will write around it.

Two more worth knowing: Google Document AI and Docparser

Google Document AI has no public purchase order parser. Its Invoice Parser returns a purchase order reference and line items, and a PO needs a Custom Extractor you configure, at $30 per 1,000 pages up to one million; a procurement document splitter is listed for limited access customers only (pricing). New Google Cloud customers get $300 of credit. Like Textract, it is a building block.

Docparser is a rule based parser: its own page recommends "one parser per supplier layout", with table rules for line items. It starts at $39 per month for 100 credits, where one credit is a document of up to five pages, with a 14 day trial (pricing). It works well for a handful of stable supplier layouts and less well for customer POs that change every week: our page on Docparser alternatives shows a layout change a rule can miss and seven tools that need no parser per layout.

What a purchase order OCR tool must return

Whatever the vendor, the output is only useful if it maps to what your ERP expects. On the sales side, a customer PO becomes a sales order; on the AP side, the PO is the reference an invoice is checked against. The fields below are the minimum we would ask any tool to return, with the trap each one hides.

Minimum fields for PO entry and matching, and where tools usually go wrong.
FieldWhy it mattersCommon trap
PO numberTies the order to the delivery note and the invoiceConfused with a quote or customer reference number
Buyer, vendor, ship toDecides who is billed and where goods goBill to and ship to addresses merged into one
SKU or item numberKey for catalog matchingCustomer SKU returned instead of yours, with no link between them
Quantity and UoMWhat the warehouse picks"10 cases of 6" read as 10, or as 60, without the pack size
Unit price and line totalWhat will be billedDiscounts applied on the line total but not on the unit price
Subtotal, tax, totalThe check on everything aboveShipping or tax lines read as item lines
Delivery date and termsPlanning and paymentLead time in days returned as a date

A tool that returns these fields with a confidence score lets you set a simple rule: above the threshold, the order goes straight to the ERP; below it, a person checks the flagged lines only. That rule, more than the headline accuracy, decides how many orders someone still touches each day.

How to test purchase order OCR on your own documents

Vendor accuracy figures are measured on vendor test sets. Yours will differ. A fair test takes an afternoon and twenty documents.

  1. Pick 20 real POs, not the cleanest ones: at least three scans or faxes, two handwritten orders if you receive any, one PO of more than one page, and two from your most difficult customer or supplier.
  2. Write down the expected values for the header and every line in a spreadsheet before you upload anything.
  3. Run each tool with its default PO template, then once more after the setup the vendor recommends (fields, rules, catalog).
  4. Count at three levels: characters tell you little; fields (PO number, each unit price) are the right level to compare tools; the share of POs with nothing to fix is the figure that sets your staffing.
  5. Check the controls, not only the reading: change one quantity in a copy of a PO so the line total no longer matches, send the same PO twice, and see which tool tells you.

Most tools on this list offer enough free pages for this test: Koncile's 20 credits, Parseur's 20 monthly pages, Mindee's 14 day trial, Veryfi's free plan of 100 documents a month, Nanonets' $50 of credits.

Sources

All pages read on 29 October 2026. Prices exclude tax and may change; we update this guide when they do.

Frequently asked questions

01What is the best purchase order OCR software?

It depends on what happens after the reading. For line items, recomputed totals and catalog matching in one call at a public price, Koncile. For a developer API aimed at customer POs, Mindee. For a large AP department with matching on invoices and an annual budget, Rossum. For a small mailbox of supplier POs, Parseur's free plan is enough to start.

02Can OCR read the line items of a purchase order?

Yes, modern AI based tools return one row per line with SKU, description, quantity, unit price and line total. The differences are in the details: whether the unit of measure is split from the quantity, whether tables over several pages are joined, and whether a confidence score tells you which lines to check.

03Which tools match purchase orders with invoices?

Koncile reads POs, delivery notes and invoices with the same template library and compares them line by line. Nanonets and Veryfi announce three way matching on their PO pages, Rossum offers two, three and four way matching on invoices, Docsumo on its Enterprise plan and Doxis as a module. Mindee, Parseur, Textract and Document AI leave the comparison to you.

04Is there a free purchase order OCR?

Parseur has a permanent free plan of 20 pages a month. Koncile gives 20 free credits once at signup, Nanonets $50 of credits, Veryfi a free plan of 100 documents a month, and Mindee, Rossum, Docsumo and Docparser 14 day trials. Open source engines such as Tesseract or PaddleOCR are free but return text, not PO fields.

05Can these tools read handwritten purchase orders?

Koncile, Rossum (30 languages), ABBYY Vantage and Amazon Textract (English only) state that they read handwriting; Docsumo mentions handwritten amounts on invoices and Veryfi lists handwritten fields. Mindee, Parseur, Doxis and Docparser do not mention it on their PO pages. In every case, cramped handwriting lowers the confidence score: test with your own order books.

06How much does purchase order OCR cost per page?

From the prices published on 29 October 2026: Amazon Textract AnalyzeExpense $0.01 per page (no PO model), Google Document AI Custom Extractor $0.03, Koncile about €0.20 on Advanced and €0.16 on Pro, Parseur €0.49 on its Micro plan, Mindee about $0.11 to $0.12 on Starter including the platform fee (rechecked on 4 November 2026). Rossum starts at $18,000 a year; Docsumo, Doxis, ABBYY and Veryfi quote PO prices on request.

07Do I still need OCR if my customers use EDI?

Keep EDI with the accounts that already use it: the order arrives structured. OCR is for everyone else, the customers and suppliers who still send a PDF, a scan or a photo, so their orders join the same flow without asking them to change tools.

08Why is Koncile ranked first in a guide published by Koncile?

Because this site is Koncile's and we say so at the top of the page. To keep the comparison useful, every fact about another tool comes from that vendor's own page, read on the date shown and linked in the sources, and Koncile's weaknesses are listed with the same detail as the others'. If a vendor's page changes, write to contact@koncile.ai and we will correct the guide.

Test Koncile on twenty of your own purchase orders

Create an account, upload your POs and compare the header, the lines and the checks with what you expected. No card, no call.

Create a Koncile account, 20 free creditsTalk to an expert

20 free credits, no credit card. For professional use only.