Automated Data Entry Unstructured Documents that Can Learn

1
Automated Data Entry Unstructured Documents that
Can Learn
2
Who is Paradatec?

• German company, founded in 1988
• Started automating mail rooms in central Europe
• One product line PROSAIR-AIDA for unstructured
  and freeform document capture
• The fastest OCR in the industry
• Positioned well within healthcare vertical and
  Accounting horizontal
• NO Templates

3
What is PROSAR?

• PROSAR - an OCR/ICR-program  based on artificial
  neural networks
• PROSAR the free-form text recognition system has
  become PROSAR-AIDA, with which data can be
  accessed from formatted and unformatted
  documents.
• PROKEY is software for the correction and manual
  verification of data which could not or only
  partly be recognized by any of the two
  recognition programs.

4
How Does PROSAIR-AIDA Work?
Classification
Extraction
Export
3rd Party Capture
Other App
Quality Control
Quality Control
5

• What are some real-life examples of PROSAR-AIDA
  in action?

6
Follett Higher Education Group

• Paradatec implemented an automated data capture
  system for Accounts Payable using
• Captivas InputAccel document imaging platform
• Paradatecs PROSAR-AIDA OCR ICR technology
• Interfacing to a Lawson payment system
• Up to 30,000 pages per day are scanned into the
  system. 
• Three (3) scanners, connecting to 3 PROSAR-AIDA
  systems via InputAccel. 
• Eleven (11) different document types identified
  by PROSAR-AIDA. 
• More than 30,000 invoice vendors. 
• Credit memos are also processed. 

7
Follett Higher Education Group

• Solution
• Vendor numbers are extracted via database lookup
  that is tolerant to differences in address
  formats between the vendor database and the
  invoice vendor address. 
• Purchase order numbers are extracted and are
  compared to items in Folletts purchase order
  database (three-way match). 
• Canadian stores and vendors are also supported. 
  
• Routing to the appropriate person (for exceptions
  or based on the dollar value of an invoice) is
  automatic, based on the PROSAR-AIDA results file
  and carried out by another tool. 
• PROSAR-AIDA OCR process takes around 1.5 seconds
  per page. 
• The system has been running without interruption
  since production

8
Akron Cardiology

• Challenge
• Extract data from potentially many hundreds of
  different EOB layouts  
• Find a solution with a proven track record  
• Eliminate the need for paper retention  
• Automate data extraction of EOB data  
• Increase the amount of data captured to include
  data needed to recover revenue (e.g. zero amounts
  paid procedures, paid amounts not matching
  contract, inappropriate denial codes)

9
Akron Cardiology

• Solution
• Scan is performed by Ascent Capture.
• Data extraction and validation are performed by
  PROSAR-AIDA and PROKEY respectively.
• PROSAR-AIDA and PROKEY are certified for Ascent
  Capture
• PROSAR-AIDA is used for automatic extraction of
  data while PROKEY is used to allow an operator to
  validate uncertain data.
• Results
• Data automation rates in excess of 75 across all
  insurance company layouts  
• Eliminated the majority of manual data entry
  work  
• Seamless interface into document imaging system
   
• Provision of data needed in order to generate
  exception reports (e.g. procedures where "paid"
  amount is zero ....)  
• Increased accuracy of data  
• More data can be provided for customer service
  databases than could have been possible via
  manual data entry

10
Americas Car-Mart

• America's Car-Mart operates 91 automotive
  dealerships in nine states
• The Problem
• Growing number of suppliers and customers,
  creating two major paper-based problems
  invoices and sales documents.
• Invoices were
• opened at the dealership
• approved
• sent to headquarters for payment
• Invoices were sent by mail/courier - slow and
  expensive and
• Faxing was a faster alternative, image quality
  was a concern
• At the corporate office,
• Six (6) AP clerks sorted the invoices,
• Manually keyed them into the accounting system
  for payment,
• Filed them in cabinets.
• Manual process was slow and prone to lost
  documents.
• Early payment discounts were often not realized.
• Company was not in a position to negotiate
  enterprise level agreements or price breaks

11
Americas Car-Mart

• The Solution
• Dealers equipped with Canon scanners and Kofax
  Capio
• The image files are received by Kofax Ascent
  Collection Server
• Kofax Ascent Capture releases to document
  management system (DMS)
• Dealers access invoices and sales docs via DMS
• To reduce data entry, PROSAR-AIDA extracts
  information from invoice images
• To prevent double paying, Paradatec extracts the
  vendor and invoice numbers and looks for a match
  in the document management system
• If a match is found, that invoice is already in
  the system and an alert is raised.
• If no match is found, metadata is exported to
  Car-Marts accounting system to pay the invoice.

12
Patient billing and eob processing

• Scan paper using any Kofax, Captiva or scan to
  directory
• Prosair-Aida performs work.
• Classify
• Validate
• Extract
• Route exceptions
• Output to SmartSearch

13
Large Hospital Patient Billing

• Automate Explanation of Benefits (EOB)
• Bank lockbox scans all EOBs to TIFF
• Automated Script Manager routines imports TIFF
  files for index into document management (DM)
• If there are less than six (6) accounts per EOB,
  a data entry clerk keys in the index information
  in MVF
• If there are more than ten (10) accounts, no
  index fields are entered and full page OCR with
  content search is enabled. This can take in
  upwards of 2 minutes per EOB to full page OCR
• Currently there are seven people indexing into
  the DM system full time. There is a considerable
  backlog of work. It is estimated another 9 people
  are necessary to catch up with the backlog.
• The EOBs are manually entered in MS4
• Manual quality control is then performed to
  ensure lockbox report matches the documents and
  MS4 entries

14
Hard Cost ROI

• EOB index time of each acct 0.5 hour
• Validation time of each acct 1.5 min
• Batch size 20-30 accounts
• Number of batches per day 40-110
• Percentage of batches OCRd 30
• Conservative Estimate
• Hours spent indexing per day 400
• Number of employees required 50
• Hard cost EOB indexing 8,000 per day or
  160,000 per month
• Estimate based on staff of seven 1,120 per day
  or 22,400 per month
• OCR hours per day 8 hr
• Validation per day 20hr
• !!Paradatec ROI less than 2 months!!

15

• For more information
• Brent Wesler
• VP of Business Development
• Office 1 203-789-8791 ext.119
• Mobile [phone]
• bwesler_at_square-9.com