Welcome to part three of our six-part series that will take readers on a journey through the latest concepts in multichannel document capture and intelligent OCR, with an emphasis on how AI has transformed what’s possible in making your documents and data work for you – and not against you.
In part one, we took a look at how RPA marked a revolution in empowering businesses to solve problems associated with manual, data-centric tasks, yet was historically ineffective in automating document processing. In part two, we looked at the emergence of Cognitive Document Automation (CDA), which does the “head work” of understanding what the document or email is about, what information it contains and what to do with it. And we examined exactly how CDA works, particularly in tandem with RPA, to streamline business processes.
In part three of our series, we are doing a deeper dive into what you should look for in a CDA solution. (Hint: You’ll need much more than just OCR functionality)
To help you when evaluating CDA solutions, we’ve compiled a list of “must-have” features. Use this handy checklist to ensure your CDA solution has all the necessary capabilities to empower you to turn documents and data into savings and simplicity.
CDA Must-Haves Checklist:
Distributed capture
Support for centralized back-office document capture using production scanners, as well as distributed use cases in field and branch offices; also central administration, licensing, reporting and scanner profile management to minimize TCO.
Multichannel capture
Comprehensive multichannel capture to accommodate all customer preferences, including mobile, email, web, fax, desktop scanner, folder and MFP front-panel integration. Mobile SDK should allow developers to integrate a full suite of mobile capture capabilities including image capture, compression, perfection, classification, recognition, extraction and data validation into their own websites and apps.
Mailroom capture (multiple departments)
Benefits delivered across multiple departments (and their associated documents); this includes support for the acquisition and understanding of any document type (forms, invoices, shipping documents, mortgage documents, onboarding documents, medical documents, emails, letters, contracts, etc.) across any department.
Document classification
Automated document classification capabilities for any document type; human workers should not be needed to apply barcode stickers or insert cover sheets. Documents should be automatically classifiable without human intervention using multiple methods (i.e., document layout, document content and regular expression-based rules), and these methods should “machine learn” via document samples and ongoing user behavior.
Document separation
Automatic document separation capabilities within a batch (stack of documents) or document package (customer-specific) without human intervention to insert separator sheets. Document separation should “machine learn” based on samples given to it (manual configuration should not be required), and ongoing user behavior.
Data extraction
Support for extraction of data from any document, in any language and in any format (i.e., structured forms, semi-structured and unstructured documents); CDA solution must extract all types of fields, including machine print (any font), handprint, cursive, barcodes, bubbles and checkboxes. For maximum automation, solution should be able to vote between multiple OCR engine results. Merely obtaining the raw OCR data is insufficient - CDA solutions locate, format and interpret that OCR data to make it business-actionable.
Data validation/validation rules/database matching
Intuitive and keyboard-friendly data validation interface, enabling humans to quickly locate and correct any unconfidently extracted characters and fields. Validation rules should be supported (e.g., Field1 + Field2 = Field3), as should database lookup shortcuts. “Fuzzy” database matching for extraction and validation (for vendor and PO lookups, for example) is also important and should scale to very large databases (>1M records).
Machine learning of documents and data
Machine learning is especially critical: applying AI to samples to train the system and continuing to learn from production user input increases the system’s document classification and data extraction intelligence over time—without the cost of maintaining rules.
Natural language processing of unstructured content (emails, etc.)
Natural language processing is another key AI algorithm; it helps drive better understanding of the content and sentiment of unstructured documents (like emails, letters and contracts) so humans don’t have to intervene. CDA solution must either include this type of technology natively, or call out to third-party cloud providers like Microsoft and Google via REST services.
Document and data exports and integration to systems of record
Support for the export of documents and data to common ECM and ERP systems—without the need to write and maintain integration code. Ideally, CDA solution includes pre-built export connectors to common destination systems and no-code ways to integrate with unsupported systems or systems lacking exposed APIs. (Note that RPA is ideal for integrating with hard-to-reach systems that lack exposed APIs).
Process intelligence
Process discovery and analytics to help the business identify automation opportunities and track performance; should include document sources, classification and extraction automation rates, user productivity and costs per document and per channel, at a minimum.
Project customization
No two projects are the same: a CDA solution should make it easy to perform common functions and also enable (via scripting) more unique, application-specific projects. The ability to add script to CDA projects and easily debug scripts is paramount in making the system do exactly what is required by the business.
Integration with RPA and BPM/DCM
RPA applications and requirements are expanding to include CDA and process automation (BPM and dynamic case management). Process automation is necessary to handle business rules, user forms and exception handling capabilities, at the very least. Any CDA solution should be a part of a broader platform that delivers these robotic and process capabilities to minimize complexities in procurement, licensing, operation and maintenance, and to ensure consistent strategic direction from the vendor offering these components and unifying platforms.