Procurement and Accounts Payable

Invoice Parsing

Affinda uses AI to extract key fields from invoices:

  • Invoicing company
  • Invoice date
  • Due date
  • Invoice number
  • PO number
  • Delivery address
  • Account number
  • Payment instructions (e.g. account number, IBAN, etc.)
  • Total amount
  • Currency
  • Subtotals
  • Tax amount
  • Line items
  • Quantities

Extract key data into buckets

  • Invoicing company
  • Invoice date
  • Due date
  • Invoice number
  • PO number
  • Delivery address
  • Account number
  • Payment instructions (e.g. account number, IBAN, etc.)
  • Total amount
  • Currency
  • Subtotals
  • Tax amount
  • Line items
  • Quantities

Business context and challenge

Unpredictable invoice formats make consistent information extraction difficult.

Every reviewer must scan through each invoice received, identifying and recording each line item such as the invoicing company, payment instructions, total due, and more.

Although many procurement departments rely on manual data entry staff to perform this processing, this approach can cause a wide range of problems:

  1. The necessary man-hours are often prohibitively expensive.
  2. Inconsistencies in documents create bottlenecks.
  3. The resulting data is highly prone to human error.

As a result, some procurement departments have turned to rules-based processing. However, because invoice formats vary significantly by supplier, any rules-based process is bound to generate errors — resulting in further bottlenecks as problematic results are referred upstream for a second layer of human review.

Affinda's solution

Capture data from 40+ customizable fields across a diverse array of document formats

Affinda’s inbuilt intelligent OCR is capable of reading scanned invoices, and even photos of invoices. Our invoice reader can understand a range of formats, including PDF, JPG, PNG, word, and more. The vast proportion of fields can be automatically extracted with better-than-human level accuracy.

Reduce costs

Improve productivity

Better auditability

Continuous improvement

Want to double-check results? Simple! Affinda’s deep-learning invoice parser makes it easy for you to examine and verify subsets of results in just  seconds — giving you a built-in tool for making sure all extracted data meets your standards. You can even amend and make notes on sample output, and feed those notes directly back into the AI.

As your in-house experts check and verify more result subsets, our invoice scanner automatically learns and improves its accuracy over time, resulting in an ever-decreasing number of results that require human validation — and ever-increasing ROI, captured straight from day one.

This ROI will continue to increase through ongoing usage, as the model gets better at recognising the invoices that are specific to your company.

And because the underlying machine-learning system applies not only to invoices, but also to contracts and other logistics documents, the machine-learning model only needs to be trained once before being rolled out across other workflows and departments.

The above process is faster than manually extracting the information, and also offers the added benefit of improving the model’s performance each time you verify or correct the model’s output.

And best of all, we can deploy our solution in whatever way best suits your needs — whether that’s by generating Excel or CSV output, an invoice extraction API, or a user-friendly web interface. 

About Affinda