With so much data being created every moment, on top of what already exists, it’s beyond a human’s ability to sort through and gain insights from all of the data made by even a single organization.
“Companies employ people to generate, disseminate, and use data,” stated Bill Tolson, Vice President of Global Compliance & eDiscovery for Archive360, during ARMA 2020, “for whatever the business’s goal is.”
I agree with Bill, and I’d go one step further to say that data and information by itself is (to put it bluntly) useless to people.
Most leaders agree that the value of information is only as good as the insights that you can glean from it to guide your decision-making.
According to most of the clients I talk to, the biggest struggle they have is understanding what information they have to work with, never mind what use it is.
Gartner predicted that as much as 80% of content will be unstructured this year. This means that these records, whether physical or digital, are sitting in a box or server somewhere and people haven’t the faintest idea what’s in it or what to do with it.
Enter the robots (or more accurately, Artificial Intelligence).
This post will cover an overview of AI, its current application in the world of intelligent information management, and what the future might hold.
Before we get started, let’s define some of our key terms so we’re all on the same page.
When someone says AI, we all immediately think of HAL 3000 from 2001 a Space Odyssey or Skynet from the Terminator films. But what does it mean in the real world?
AI as we understand it today was driven by the work of the brilliant mathematician, Alan Turing, and was eventually coined by John McCarthy who defined Artificial Intelligence as “the science and engineering of making intelligent machines”.
There are myriad types of artificial intelligence in information systems research. The main two are “strong” and “weak” AI. What it comes down to is how well it follows the rules: A weak AI won’t break the rules no matter what, while a strong AI adjusts the rules as it sees fit. To put it another way a “weak” AI is a very complex algorithm that can’t step outside its bounds while a “strong” AI is closer to human cognition.
Machine Learning can be considered the younger sibling of artificial intelligence. Machine Learning allows an algorithm to “learn” and adjust based on past data.
The two major types of machine learning are supervised and unsupervised.
The difference is easily explained through a story I heard recently that essentially said, if you ask a Machine Learning algorithm, “if all your friends jumped off a bridge, would you?”, an unsupervised Machine Learning system would likely say ‘yes’. It makes decisions based on prior decisions and actions and nothing else.
With supervised machine learning, someone (hopefully) would be able to step in to say, “bridge jumping is a poor life decision”.
Right now, supervised is what’s working and where we’re at. Unsupervised is what’s next.
Intelligent information management (IIM) takes the knowledge and best practices of records and information management professionals and empowers them with the computing power of artificial intelligence.
What does that look like though? Let’s run through what a perfect future state of IIM would look like in a three-step process best described by GCN:
That’s the ideal, Garden-of-Eden-type world but are we there yet?
Not quite, but that’s not to say it’s impossible and all hope is lost. There have been major strides in this arena even in the last few years.
There’s one question that always comes up whenever I talk about using AI to power a records and information management program:
“Does this mean records managers will be obsolete?” In a word, “no”.
In a few more words, “AI has the capacity to make records managers even better at their jobs”.
According to Tech Jury, the current estimate of data created per day stands at 1.145 trillion MB. To hold that much data, it would take 795 TRILLION floppy disks (remember those?). Then, if you had the free time, you could lay them end to end from Earth to Mars with tens of thousands of kilometers to spare.
This is all to say that the only way to solve the inherent Big Data problem is to use bigger and smarter Artificial Intelligence as time goes on to power your information management program.
Synthesizing data from multiple systems of record is another huge benefit of IIM. Instead of relying on a human to make a square peg fit into a round hole, a technology-enabled information management program would do all the heavy lifting and analyzing of the data while pointing out what is important so the humans involved can then decide if and how to act on that information.
Not too long ago, an entity undergoing the discovery process during legal proceedings would hire a team of contract attorneys to dig through all the information they had, understand which were actual records and which weren’t, and what was relevant to the proceedings and what wasn’t. The problem is obvious: it’s slow, costly, inaccurate, and inconsistent.
The solution is leveraging AI and machine learning (ML) for predictive coding on discovery.
Even today, ML algorithms have a provable accuracy rate and have been shown to be able to quickly sort through identifying a record versus garbage.
ML and AI have been buzzwords in information management for decades. What has changed is the pandemic’s effect on it. Like most other digital transformation trends, the pandemic has accelerated changes that were already underway.
Here are a few specifics in how artificial intelligence is affecting intelligent information management today:
The one area used the most is in classification – sorting through massive amounts of data. Today, AI and ML are already being used to sort and categorize digital records quicker and more accurately than humans ever could. Coupled with the improvements in Optical Character Recognition (OCR), physical records are also able to be sorted much quicker, although there is still room for improvement.
While the word “record” brings to mind “text”, let us not forget the other visual types of records like photographs and images. “The recognition and classification of images is what enables many of the most impressive accomplishments of artificial intelligence,” writes Daniel Nelson. This is one area that has seen a massive improvement over the last few decades. When it comes to facial recognition technology (FRT), according to RecFaces, “the leading FRT algorithms nowadays have almost reached perfection in human identification with an error rate of 0.45%,” though most don’t reach nearly this high of a standard and need human assistance.
Similar to images, audio is another area where AI has made big strides. Think of any time you ask Google, Alexa or Cortana to do a search for you. It is, in effect, turning your speech into text and searching the internet for a match. Whether through automated closed captioning for the hearing impaired or the creation of a transcript of a court case without the need for a court reporter, AI and ML are making a big impact in the transmediation of audio.
In the end, we need to leverage technology like AI and ML to make us smarter and better equipped to do what we’re asked or required to do in the course of being information management professionals.
While there are great solutions available now, we’re not yet at the point where we can ask Alexa to index and classify a pack of HR files from 1984.
But, at least she will play The Weeknd any time I want.
So that’s something.
To learn more about technology innovation and how to start your digital transformation journey, visit our website.