Cut Invoice Processing Time to Just 3 Seconds With This Invoice OCR Tool

‍

Invoice processing remains a time-consuming and expensive task for many businesses. According to Business Insider’s accounts payable process automation report, 36% of firms still use paper invoicing and 47% rely on manual processes for approval. Automation solutions for the payable process can bring tremendous business benefits.

In this article, you’ll find out about the problems faced by manual invoice processing and semi-automated processing that use traditional optical character recognition (OCR) and invoice templates. You’ll see how, by using fully automated invoice OCR tools instead, you get compelling business benefits and amazing productivity improvements.

Problems With Manual Invoice Processing

Traditionally, invoices were processed manually by accounts payable (AP) teams. They reviewed invoices, checked for errors, verified details with other departments, entered data into an accounting system, and sent them for review by senior accountants or management. Many companies use this approach even now but it brings many long-term business challenges:

• Demands more staff: A large, skilled support staff has to be hired for manual data entry, multiple rounds of verification, and coordination. This increases direct employment costs and increases the workload on management and HR.
• High per-invoice processing cost: More staff means processing an invoice manually turns out to be quite expensive.
• Slow: Manual invoice processing slows down your ability to make business decisions quickly in a competitive environment. Accounting practices like invoice matching and ledger reconciliation also take a long time.
• Format limitations: Vendors send invoices in many ways — as paper invoices by fax or mail, as email attachments, as digital formats like PDF files, as special formats like electronic data interchange (EDI) or XML, and more. Handling such a variety of formats manually can result in missed invoices and coordination problems.
• Error-prone: Manual processing frequently suffers from typos, date and currency mistakes, misinterpreting handwritten text, and similar problems.
• Inefficient at detecting fraud: Manual verification of data across multiple invoices is very inaccurate and inefficient. Fraud and unintentional mistakes may remain undetected for long periods and complicate financial audits.
• Needs physical storage: Paper invoices require storage space and filing practices, which increases both real estate and staff.
• Spoil vendor relationships: Manual workflows can result in late payments and payment disputes with vendors.

These challenges prevent manual invoice processing from scaling up your business beyond a few vendors.

Problems With Traditional OCR Invoice Processing and Invoice Templates

Due to the challenges of manual invoice processing, most businesses have been using semi-automated data capture that combines traditional OCR software and invoice templates.

Given the wide variations in invoice layouts, businesses need some way to help software identify invoice text as appropriate invoice fields. Invoice templates help address this problem.

An invoice template consists of positional data and parsing rules that enable the invoice tool to identify unstructured text in an invoice as structured data as long as the invoice conforms to that template. Some invoice scanning software provides visual editors to create invoice templates easily.

Two Critical Limitations of Invoice Templates and Traditional OCR

But such capture solutions have two fundamental limitations:

1. Traditional OCR is not very accurate: Traditional OCR technology recognizes text using only the pixels and contours of individual characters in an image, which makes it prone to typos and recognition errors. People don’t interpret text that way — in addition to visual features, we also use our language skills, vocabulary, and real-world knowledge of invoices to understand the text correctly.
2. Invoice templates don’t scale: The variety in invoice layouts is just too high for templates to be practical. If your business has more than a few dozen vendors, template management becomes unwieldy. Creating, searching, and updating hundreds of invoice templates correctly increases the manual effort with diminishing returns.

More Problems With Invoice Templates and Traditional OCR

The two critical limitations of templates and traditional OCR result in many more problems:

• Deteriorated accuracy under noisy data: Real-world invoices differ widely in ambient lighting and perspective angles. They may contain handwritten text, colors, or unimportant text. There may be small variations in the positions of invoice fields. The result is inaccurate OCR and template mismatches.
• Lack of deep understanding of fields and text: Templates don’t understand the relationships between the text, positions, and fields deeply. Even minor variations in positions may lead to template mismatches. Downstream tasks like invoice matching become inaccurate and unreliable without such in-depth understanding.
• Slow: Too many invoice templates means slow template matching and reduced accuracy.
• Only marginally better at finding mistakes and fraud: Because this approach does not understand invoices semantically, its ability to detect mistakes and fraud is only marginally better.
• High per-invoice costs: Compared to manual processing, semi-automation can speed up data entry. But its overall low accuracy and unreliability force organizations to hire review and verification staff just as in manual processing. So per-invoice costs remain quite high.

Fully Automated Invoice OCR Tool Is the Best Solution

Fully automated invoice OCR tools that understand invoices the way people do — using machine learning and artificial intelligence — provide solutions to all these challenges. They combine state-of-the-art deep learning models for vision, text recognition, language understanding, and spatial understanding to provide highly intelligent and accurate invoice processing.

Width.ai’s Automated Invoice OCR Tool Streamlines Business Costs

The biggest business benefit of Width.ai’s automated invoice OCR tool is drastically lower costs. Automated invoice processing doesn’t require much supervision to begin with and even that reduces as the tool automatically adapts to your invoices. Costs of other tasks like invoice matching are also reduced thanks to the high reliability of automated invoice OCR. 

Per-invoice cost reduces by as much as 85% compared to manual processing and 70% compared to the template approach.

Invoice automation reduces the end-to-end time for processing one invoice to just three seconds compared to two hours for manual processing and 20 minutes for the template approach.

Why You Should Use Width.ai’s Automated Invoice OCR Tool

Reduced costs and time are not the only benefits. Let’s understand all the primary advantages and functions of Width.ai’s automated invoice OCR tool.

1. Data Extraction From Any Invoice Layout Without Invoice Templates

The tool does not require any invoice template whatsoever. Its deep learning models are already trained to understand thousands of real-world invoices of the most common invoice layouts out there. Additionally, it creates customer-specific pipelines adapted to your specific invoices and data requirements.

2. Supports a Variety of Invoice Formats

The invoice OCR tool supports a huge number of formats:

• Digital formats: It supports digital formats like PDF, EDI, XML, and images (including noisy or low-resolution photos). It supports fetching files from network storage and cloud storage.
• Digital invoices from popular invoice managers: It has built-in support for fetching invoices from FreshBooks, QuickBooks, Zoho Books, Xero, and Pilot and processing them without any configuration.
• Digital invoices from accounting systems: It supports invoices from SAP FICO.
• Emailed vendor invoices: It monitors specified mailboxes for invoices received through email.
• Paper invoices: It can handle paper invoices received by mail or fax and digitized using smartphone cameras or scanners. Paperless accounting reduces your real estate costs and staff costs.
• Handwritten invoices: It’s capable of understanding handwritten text to accurately guess what’s written just like people do. Handwritten additions or corrections are automatically identified and added to the relevant invoice field.

3. Capable of Exporting to Other Systems

The invoice OCR tool supports end-to-end AP automation by sending the extracted invoice data to an ERP system like SAP FICO or to another accounting software, CRM, database, email, or some management workflow.

4. Supports Many Export Formats

The tool can export extracted data to many formats like PDF, CSV, Javascript Object Notation (JSON), Excel, and more.

5. High Accuracy and Reliability

The accuracy of extracted data is very high. The results can be reliably used for other tasks like financial report generation and invoice matching.

6. Scalable and Fast

Taking just three seconds to accurately process an invoice, the invoice OCR software can handle millions of invoices per day. If your organization has historical invoices (even typewritten or handwritten), we can create extra cloud resources to process them fast. Historical invoices enable your business intelligence teams to detect long-term business trends.

7. Supports Custom Fields

The tool comes with built-in support for over 50 common invoice fields. But it also lets you add the custom fields you want. Some of the custom fields we have seen include:

• Mapping line items to relevant general ledger codes
• Due dates
• Various postal and email addresses
• Other key data

8. Captures Additional Information

In addition to the common fields, it can identify important details like special payment terms using deep language models. For example, it uses natural language processing to identify a sentence with payment terms and file it against the payment field.

Start Using Width.ai’s Invoice OCR Tool in Just 6 Steps

Width.ai’s enterprise-grade automated invoice processing software can be quickly integrated into your accounting practices in just six steps.

Step 1: Integrate the Invoice OCR Tool Into Your Business

For a smooth transition to our invoice OCR tool, we start by understanding your existing invoicing workflows:

• Where do you receive your invoices from?
• What formats are they in?
• Do you get paper invoices by fax or mail?
• Are you using invoice managers like QuickBooks?
• Which invoice fields are important to you?
• Should invoice details be exported to an ERP, CRM, or database?

These details help us plan the deployment and infrastructure necessary for your workload.

Step 2: Configure the Invoice OCR Tool to Read Your Invoices

With a deployment plan in place, configure the tool to fetch your invoices:

• Digital invoices: Configure the API and authentication credentials to let the tool connect to your invoicing or accounting system like FreshBooks, QuickBooks, Zoho Books, Xero, Pilot, or SAP FI.
• Invoices in Amazon S3: Supply bucket details and OAuth details to fetch invoice images from your S3 cloud storage.
• Emailed invoices: Set up the tool with email credentials to the mailboxes that get invoices as email attachments. The tool fetches emails regularly and processes any attached invoices in supported formats.
• Paper and handwritten invoices: Set up invoice scanning using scanners or smartphone cameras to digitize paper invoices to PDF or image formats.

Step 3: Adapt the Invoice OCR Tool to Your Invoices

Although the tool can handle a wide variety of invoice layouts right out of the box, your organization may have invoices with unique layouts and custom fields. So, to boost accuracy and correctness, we fine-tune the tool’s deep learning models on your invoice and receipt samples to create a customer-specific invoicing pipeline just for your business.

Step 4: Set Up Invoice Data Export

The invoice OCR tool can export the extracted data to different formats, cloud storage services, local storage systems, third-party software, or workflows. You can configure it to select invoices using some conditions (like a date range) and export them to specified storage destinations. Or send invoices that go over some threshold amount to a management approval workflow.

Step 5: Start Capturing Invoices

Following configuration and fine-tuning, the tool’s ready for bulk invoice capture with little operator intervention.

Unlike traditional OCR which relies only on pixels and contours in images and suffers from lower accuracy, our fully automated invoice OCR tool identifies invoice data just like people do by combining visual, linguistic, and spatial characteristics. These details help accurately identify an element as a purchase order or an invoice number, and so on.

The outcomes of this step are the extracted fields and their values for each invoice. This data is exported to a structured format like JSON or XML or another software like ERP.

Step 6: Measure and Monitor Regularly

The tool measures precision, recall, and F1 scores to help you judge the effectiveness of your fine-tuning. It also generates confidence scores for each extraction to help you find problematic invoices and further fine-tune the tool on them.

The tool also lets you monitor its progress in real-time and alerts you to any processing problems over Slack, email, Jira, PagerDuty, or your CRM.

Start Processing Your Invoices With Width.ai’s OCR Invoice Processing Software

‍

You’ve seen all the features and benefits of Width.ai’s invoice OCR tool powered by the latest deep learning technologies. At Width, we have years of experience developing data capture software for business processes in multiple industries. Contact us to see a demo of our touchless invoice processing.

‍