Artificial Intelligence For Content Management: What Does It Mean?
A Google image search of artificial intelligence (AI) yields tons of semi-opaque heads with brains morphed into circuit boards — mixed in with streams of binary digits reminiscent of The Matrix… and most of these images are unexplainably blue-coloured.
No wonder there’s mixed reports on what AI is. We know how to approximate and define human intelligence — or natural intelligence. When machines begin to do some of the things that humans can do, we call that artificial intelligence. And, to machines to have this intelligence, they need to be taught. This is called machine learning (ML). Of course, there is much more to it than that, but the concepts that follow in this whitepaper deserve a bit of a preamble and basic understanding of AI and ML.
AI is math. There is no sneaky sleight of hand. It has been said that worrying about AI taking over the world is a bit like worrying about the overpopulation of Mars. AI in information management is all about… you. It is about allowing you to focus on what you do best. Better. Faster. Smarter.
Sure, AI is exciting and interesting with all kinds of great tech… But as with any technology, it must come down to this: How can it help you do your job?
What about AI and ML in the context of enterprise information management?
Information management has changed from pure document management and archiving into a real business enabler. Today’s intelligent information management solutions offer ways to automate time-consuming and repetitive document-driven processes within a business.
A key driver in this automation is artificial intelligence and machine learning. AI and machine learning can reason and discover meaning as well as learn from experience. Moreover, artificial intelligence for content management systems can easily churn through lots of information to identify patterns and categories in the data. That ability is put to work to enable new ways to search, find, use, and manage information, and add automated workflows to document management processes.
Jayson deVries, M-Files’ resident artificial intelligence expert, lists two major, emerging themes related to AI and machine learning in information management:
AI Should Not Be Complicated for the User
Some products offer powerful AI capabilities, but they require a deep understanding on how to carefully train, test, tune… and repeat. If a company attempts to implement it without the correct skills, it can turn into an expensive endeavour that doesn’t end well, with the product taking the blame. Then there are products that offer intelligence out-of-the-box by delivering pre-trained “canned” AI. These do have merit but must only be used in their intended contexts – a fact that, if overlooked or misunderstood, will cause disappointing results.
Simplicity of use is a maxim we espouse at M-Files. We believe in providing intelligence services out of the box, so your organization does not need a data scientist to use it in your business.
What differentiates M-Files AI from others is the use of machine learning behind the scenes. Users do not have to change how they work. There are no extra steps to train the system.
We have worked hard to provide a tool with this self-learning capability — all so that people realize the benefit of AI without the need for costly data analytics experts on staff.
AI Needs to Be Tested with Your Data
There has been a recent blitz of “intelligent” products and services on the market, but many are not that intelligent at all. Customers need to lift the curtain and ideally understand how well an AI solution really works with their own data, not just demo data.
To execute testing with real data, the AI solution needs to be easy to implement. If it requires lots of effort to get the artificial intelligence to work, it can be totally impossible to test with real data.
deVries explains, “I’ve recently been using the classic duck analogy. Ducks are swimming around calmly, gliding on the water surface. But a lot of impressive work is happening below the surface to make sure the duck gets to where it needs to go.” He continues, “Extending that analogy further, some information management platform vendors are now selling rubber duckies. They look and quack like a duck – if you squeeze them exactly right – but they’re actually hollow with nothing below the surface.”
AI in information management has a ton of incredible uses. In this whitepaper, we’ll explore five practical examples where artificial intelligence really moves the needle in real-world information management.
Thinks about how you classify documents as a human would. Imagine your boss hands you a piece of paper and says, “Have a look at this.” Instantaneously, your brain cycles through a progression to help you identify the document.
Is it an image or text?
Is it in a language I understand?
Is it immediately familiar — an invoice format I have seen a hundred times before?
Is there information on it I should focus attention on — dates, costs, people?
And your brain does all of this in seconds, all to answer a simple existential question:
Information management AI can apply a similar progression to incoming and existing documents in a company’s information ecosystem. In a document management environment like M-Files, every file in the environment has a class, which defines what type of document it is.
It provides the basis — along with metadata, which we will discuss momentarily — for a Google-like search of information. In M-Files, a search can be done on Class across repositories or folders, so it does not matter if that document is stored in the CRM, SharePoint, network folders or anywhere else.
It is important to get the Class right and AI can automatically detect the class, or type, of a document. The AI is continuously listening and learning from user-uploaded documents, figuring out that these invoices all look like this and these agreements are a standard format.
Benefits Of Using AI For Document Classification
Users tend to become lazy when tagging objects with classes.
AI can do it automatically or suggest the proper Class, encouraging users to be more mindful of document classification.
No more typing errors or using a different term for the same Class. Classification becomes uniform and error-free, ensuring a clean, structured vault.
Easier to find documents.
Since document Class is uniform and automatic in many cases, the ultimate reward is that users can find the information they need faster and easier.
With M-Files Smart Classifier, configuration is simple and organizations start seeing benefits almost immediately as the AI gets to work, continuously learning about common document types unique to the organization.
Before we discuss AI applications in metadata extraction, it is important to understand what metadata is. Metadata are the data fields used to describe and further classify files and documents.
When you download or stream music from the internet, it has metadata attached to it that makes it easily classified, searched, and found. That Beatles track you are listening to is accompanied by a slew of other metadata: song title, artist, songwriter, release year, record company, run time, album, genres. And that metadata adds value to the file. Now it can be found if you search for The Beatles, songs from 1967, classic rock songs or whatever else.
Metadata for enterprise documents and files works the same way. The more metadata points you have, the easier it is to find in the future. There is a dramatic increase in the overall utility it has.
Artificial Intelligence can be used to automatically extract and generate metadata – different types of metadata for different types of media.
An example of this might be where users at a property management company frequently store Lease Agreements. When the AI module receives some inputs in the form of uploaded Lease Agreements, AI might come in handy. If a user uploads such an agreement, AI can extract data and suggest context in the form of metadata entries for lessee, lessor, lease dates and other important information.
Lots of AI vendors — even the big ones — claim they can extract usable information from a document. But how usable is that data really? Here is what I mean:
Sure, their AI can scan a vendor agreement, and extract company names, proper names, dates and other bits of information. But where it fails is in knowing what those bits of data are. It can say, “Here are the five people, seven companies and four dates mentioned in this document.” But what it can’t do is say: “This date is the Effective Date. This person is the vendor account representative. This company is the vendor and this one is the purchaser.”
Ultimately, current information management AI often fails to apply context to those data points. Not so with M-Files.
When you just get a bunch of dates in the metadata card without any context, it is of very little value to you.
And then take the challenge of finding the lessor and the lessee. Less-intuitive AI modules can just give you a list of potential matches (with Facebook, LinkedIn, and Google on that list since they were added for marketing in the footer of the document). All pop up as potential lessees and lessors.
You need an intuitive solution that can find these parties accurately within a document’s content and M-Files has the intelligence to make that happen. We are contextually understanding the various parties in a document as they relate to the document.
We are not just saying this is a date. We are saying this is the start date.
We are not just saying this is a company. We are saying this is the legal entity that signed the lease.
And we are saying these things without the help of a human to manually batch a bunch of documents, annotate them and hope it works.
Non-Compliance with data privacy regulations can be costly — both in terms of fines and reputation. Companies with information management AI are much better positioned to comply with privacy regulations like California’s new California Consumer Privacy Act (CCPA) and the EU General Data Protection Regulation (GDPR), just to name a couple.
The most sensitive information often lies embedded within documents. AI can be used to automatically detect and act to protect information that requires special care.
Personally Identifiable Information (PII) like social security numbers and driver’s license numbers.
Intellectual property like methods, blueprints, schematics and more.
Financial or legal restrictions like credit card numbers and bank account numbers.
AI like M-Files Discovery helps customers find sensitive data within the contents of large document archives. In addition to finding that data, M-Files Discovery can automatically set metadata, classify, and categorize documents, as well as update document permissions and initiate workflows.
It can automatically examine the contents of M-Files Vaults and connected repositories — analysing a one-time batch, only incremental changes, or periodically reprocessing everything. Once sensitive data is found, a workflow can be initiated so those documents are always properly handled — thus, reducing non-compliance risk significantly.
How long would it take a human to scan through 1,000 documents on the hunt for social security numbers? Weeks, if not months. AI can do it in a matter of minutes.
Search and Artificial Intelligence
Google has set a high bar for user expectation for a search function. We are all familiar with search. We have learned to use just the right keywords to get what we are looking for. But search is increasingly getting smarter – getting into our heads. How often have you said, “How did it know what I meant?”
But search has become more than just keywords. Artificial Intelligence can help you get access to the information you need when you need it. Let us say you wanted to find a couple of slides you saw in your colleague Maggie’s product presentation from last week. With traditional search, you would have to remember the title of the presentation or even the filename — and you probably have no idea what Maggie named her file.
When AI automatically extracts context from documents, it makes searching for those slides much easier. Now, for example, you can find what you need with a search like “recent product slides from Maggie” and in Google-like fashion, you would find the most likely results.
With M-Files, users also have federated search capabilities, meaning searches are performed across repositories, so it does not matter if documents are in the CRM, SharePoint or network folders. How incredibly powerful would it be to search across your entire information ecosystem and find exactly what you need, much like you can in Google?
Amazon has a knack for showing you products you might want to purchase. If you use streaming media services like Netflix, Pandora or Spotify, you know that they have an uncanny knack for recommending things you’ll find interesting. That is AI at work.
With Netflix, for example, their recommendations system helps you find a show or movie with minimal effort. Their AI-powered algorithms estimate the likelihood that you will watch a title, based on several factors including:
Interactions with the service — such as your viewing history and how you rated other titles.
Other members with similar tastes and preferences.
Information about the titles – such as their genre, categories, actors, release year.
In addition to knowing what you have watched on Netflix, to best personalize the recommendations, the algorithm also looks at things like:
The time of day you watch.
The devices you are watching Netflix on.
How long you watch.
In the same way that Netflix suggests content, AI can help you discover information that you never knew even existed in your enterprise information environment. It can suggest documents you might want to check out by analysing a combination of:
Your usage patterns.
Usage patterns of colleagues.
It can predict the answer to the question: What might be helpful to you right now?
Imagine a scenario where you’re working on a sales proposal for an existing client and your information management system suggests that you check out a previous proposal for that client, a press release they recently issued and specs on their last purchase. How impactful would that be?
There has been massive hype for years about how AI is going to change the planet. Just pick up your phone and it seems to know it is going to take 45 minutes to drive home today. You did not do anything other than your normal commute for a few months and it watched your movements. It knew where work was because you tagged that location on your map. It knew where home was and when you usually leave to go home. Powerfully intuitive.
AI applications in information management are getting just as smart and contextually relevant. With so many functions in an enterprise information ecosystem, it’s not hard to see how AI can have such a positive impact on a company.
Gone are the days where files were hidden away on individuals’ laptops. With M-Files and its embedded AI functions, a brave, new world of information classification and findability is here. This is the revolution. Are you going to take advantage of it or swim against the tide?
For more information on how we can help your organisation please email: firstname.lastname@example.org or visit www.documentmanagementsoftware.com.au