The state of AI in document management
Virtual and traditional filing cabinets — jam-packed with contracts, vendor agreements, onboarding materials, and other documents — are still stifling productivity, even in 2019. Studies find that 46% of workers waste time every day on paper-intensive workflows, that workers spend more than 11 hours per week dealing with document management issues, and that six of those hours are considered to be wasted time.
Beyond lost productivity, document management issues present a real challenge in terms of organization and customer service.
A law firm knows they have a document, but are unable to locate it. An insurance company might struggle to connect the dots between the paperwork on hand and related documents on file — especially when account, contact, or candidate history is lacking. This leaves businesses with two options:
- Continue to dig for the proverbial needle in the haystack, typically using a manual workflow, or
- Speed up the process by integrating artificial intelligence (AI) into their document management system (DMS).
Instead of duplicating traditional document workflows in a digital format, AI is becoming a force multiplier in document management. It has the ability to make every step of the workflow better, smarter, and faster — from the processing of documents to their storage to the extraction of the information they contain. It’s already eliminating wasted time, enhancing and improving collaboration and engagement, and quickening turnaround times on common workflows.
The following six areas are where AI is revolutionizing management, creation, and usage to expand the benefits of document management.
1. Automating classification and processing
AI has made enormous strides in what is commonly called computer vision, the ability to recognize what it sees and make decisions. In the world of document management this ability is being applied to optical character recognition (OCR) technology that enables a Document Management System (DMS) to read the contents of a document and automatically classify and process it, without human intervention.
The more documents AI reads, the more it can observe how employees interact with the documents, and the smarter it is at identifying and processing information.
In one example of this, the newly acquired Nuance Document Imaging, part of Kofax’s Intelligent Automation Platform, uses OCR for fast, easy scanning of documents and transforming paper documents into actionable digital information. This technology promises to save many document-choked employees — for instance, a hiring manager facing a stack of resumes — by reading the entire stack and identifying those documents with words and phrases that indicate a high level of qualification for the open position.
2. Data extraction
An AI-powered DMS can accurately and more quickly extract information currently hidden in individual documents.
Google’s Document Understanding AI, for instance, allows enterprises to ingest data from forms, documents, and contracts and extract key-value pairs and entities. In addition, businesses can add their own schema where Google provides knowledge capabilities through a new alpha capability called knowledge service.
This technology is already being put to work by Google partner Taulia to accelerate invoice processing by automatically recognizing and extracting critical information, like an invoice number or line item, from each invoice to speed up invoice processing and better manage company cash flow.
3. Clustering documents
The concept of using some programming to perform cluster analysis on a body of documents has been behind search results of web documents for quite some time. However, the application of AI to this task brings with it a far higher level of sophistication and precision.
An AI-powered DMS can more accurately assign a business’s vast library of documents to different topics or hierarchies (which is especially helpful when topics and hierarchies are not previously known), understand relationships between documents within a broader context, form inferences and hypotheses, and discover similarities between documents. The result is easier categorization, organization, and search of company documents when a deep dive is required.
Tasks like these are especially critical for law offices and other business services that rely on their seemingly endless document stores. It’s not uncommon for lawyers to cull through countless documents, looking for that proverbial needle in a haystack.
iManage RAVN’s software classifies and extracts key documents. By understanding and organizing data and information, the program identifies and clusters based on similarities and relevance. From there, RAVN can even extract meaningful data from contracts, financial statements, and other paper records, making it easy to work through years or even decades of work of files to find exactly what’s needed.
4. Bringing order to unstructured data
For all the structured data that is gathered by the modern enterprise, 80% of enterprise data is unstructured and 70% is free-form text — think emails, specifications, memoranda, and comments, for example.
This is not surprising, as this is pretty much how humans communicate. We tend not to communicate in a structured way, and, when someone tries to impose structure on our communications, we tend to chafe and rebel against it — which might explain why Slack has become so popular! Still, no enterprise can be satisfied with having up to 80% of their data locked out of view.
But thanks to an AI-powered DMS, companies are already uncovering this hidden treasure to stunning effect. One example of this can be found in the partnership between Dropbox and GrayMeta AI, which allows enterprises to automatically analyze files, extract technical metadata, and then tag files with relevant information, including identified people, objects, logos, landmarks, speech-to-text conversion, and more.
In a more dramatic example, some companies are using AI and machine learning to scour emails, texts, and other customer communications to understand words, semantics, and sentiments, and connect that data with billing and service history to predict who will buy what products and services. Most remarkably, these models regularly outperform models that use structured data only.
5. Supporting content and document development
The AI-powered DMS has incredible potential to streamline the content and document development workflow. While no frontrunners have materialized around this particular challenge, apps like Grammarly are demonstrating how AI can be applied to “pre-edit” documents without extra human intervention. And this is just the beginning, I believe.
Consider a group of various stakeholders participating in the drafting of a legal document. Even with current digital collaboration tools, reviewer feedback tends to arrive unstructured, forcing one human agent to push everyone to comment, gather, and collate their comments, and then negotiate with the group on what changes should be made. Adobe’s own PDF share and review tools in Acrobat are great examples of new ways to digitally collaborate, as are recent offerings from Microsoft.
This is one place where an AI-powered DMS’s ability to read, understand, and bring structure to unstructured comments could really streamline the process.
6. Securing documents and data
Security is more of a hot button than ever, especially when it comes to sensitive documents in financial services or healthcare industries. An AI-powered DMS is uniquely positioned to provide document security at scale.
AI, for example, can be taught to detect sensitive and personally identifiable information (PII) in documents and then flag those documents for special handling. Automatic classification and processing can ensure that no documents are left in unsecured locations before they are actioned. Anomaly detection can also be deployed to identify and flag potentially fraudulent documents.
Revolutionizing document management
The potential for the AI-powered DMS is immeasurable at this point. Certainly, solution-makers are racing toward the point where any company — regardless of its budget or the number of data scientists it has on staff — will be able to take advantage of these tools. In document-heavy industries especially, the possibilities are endless. A law firm could off-load to AI much of the arduous task of data discovery that requires sifting through a mountain of documents. Mortgage companies could accelerate the processing of closing documents by allowing AI to streamline their workflows.
Whatever the application, the benefits are obvious — higher productivity, lower costs, more agile operations, and more data-driven decisions. And those are benefits AI is uniquely positioned to deliver to anyone who interacts with documents — in other words, pretty much all of us.
This article was originally published on Information Management.