For two years, there was the huge hype that Artificial Intelligence was going to completely change the world. But just to be honest, how many places have you actually encountered artificial intelligence in software you use every day? Not exactly groundbreaking to speak of. Many people therefore rightly complain that AI is still a hype. M-Files puts an end to this hype and really changes how we humans interact with information and documents.

Yet there is a portion of artificial intelligence that you have to deal with on a daily basis. Take your phone, for example, it can tell you that driving home from work today will take 45 minutes. That's because your phone tracked your daily commute. Your phone knew where your work was because you marked that location on your map and it knew where your home was and also at what time you usually go home. Pretty powerful and intuitive really. M-Files has this same cleverness in it.

Today's AI still needs human help

You see artificial intelligence more and more in information management solutions, and one of its main applications is document processing.

A good example would be employees of a real estate company who regularly file rental agreements. When artificial intelligence receives some input in the form of uploaded rental agreements, it can do the following:

• Recognise the uploaded document as a rental agreement
• Extract data (metadata) from the document such as the tenant, landlord, dates and other important details

Sounds pretty useful, right? It makes classification of documents easier and thus also enables more intuitive searches and queries. That way, people find information they need quickly and easily.

But here is the crux: all AI modules for information management require the manual input of a human.

It is a really laborious process to collect a lot of documents, annotate them in a certain way and tune the algorithms to work the way you want them to. This process looks like this:

1. Whether it is a data scientist or someone in training, a person will have to enter a whole range of applicable documents.
2. A human tries to train the algorithm by manually annotating the documents to create references between data points.
3. Once a human thinks accuracy is high enough, he will perform a blind test on a set of "random" documents.
4. If the test fails, back to step 1. If the test succeeds, then he can implement this model and monitor it manually.

Basically, a human has to go through that iterative process to the point where the AI is properly trained.

AI can read data, but are they the right ones?

Many AI vendors - even the big ones - claim they can extract actionable information from a document. But how useful is that data really? Here's what I mean:

Existing AI technology can, for example, scan a supplier agreement and extract company names, people, dates and other information. But where it often fails is to determine the context of that information. The AI can say, "Here are the five people, seven companies and four objects mentioned in this document." But what it cannot do is say, "This date is the effective date. This person is the owner. This company is the seller and this one is the buyer."

So context is often lacking in current AI.

The M-Files AI processes large volumes of documents with context

So, as you can see, until now a human still had to be involved. With M-Files, that no longer needs to be the case. The M-Files AI is fully self-learning without configuration. Not AI just to apply AI, but actual added value. M-Files is about making users' work easier, so they can do their work without having to worry about the AI or manually configure documents themselves. With M-Files, organisations shorten the time between acquisition and added value to a minimum.

What if your information management solution could just learn from what users do? What if it could just listen and learn? With the Intelligent Metadata Layer the AI module does just that; it automatically dives into a repository full of documents and learns by itself.

Take the following example: users must constantly enter invoices and metadata for things like due date, supplier and amounts. Behind the scenes, AI correlates that metadata with the content of the document. When other similar documents enter the ecosystem, M-Files can automatically extract that information from the document. This without a human having to manually extract the data.

M-Files can teach itself and thus answer the following important question:

What data do users care about?

A practical example: the rental agreement

Let's take the example of the lease agreement. It probably contains some important dates:

• the date you signed it
• the date on which it takes effect
• the moving date
• the end date of the contract

If you just get a bunch of data from a document without any context, it is of little value.

And then the challenge of extracting the landlord and tenant from a document. Less intuitive AI modules may just give you a list of potential matches.

You need an intuitive solution that can accurately find these parties within the content of a document, and M-Files has the intelligence to do just that. M-Files understands who the different parties are in the context of the document.

It does not say this is a date. M-Files says this date is the start date.

It does not say this is a company. M-Files says this is the legal entity that actually signed the lease.

M-Files does that all by itself, without help and without a human hand.

This is actual AI and with it, M-Files starts yet another revolution in information management.