You are here

You are here

Predictive analytics vs. AI: Why the difference matters

public://pictures/rutty.jpeg
Michael Rutty Senior Evangelist, IM&G, Micro Focus
 

There are few movie scenes I can recall from my childhood, but I vividly remember seeing the 1968 Stanley Kubrick sci-fi movie 2001: A Space Odyssey in 1970 with my older cousin. What stays with me to this day is the scene where astronaut Dave asks HAL, the homicidal computer based on artificial intelligence (AI), to open the pod bay doors. HAL's eerie reply: "I'm sorry, Dave. I'm afraid I can't do that." 

In that moment, the concept of man vs. machine was created, predicated on the idea that machines created by man and using AI could (eventually) defy orders, position themselves in the vanguard, and overthrow humankind. 

Fast forward to today. Within the information governance space, there are two terms that have been used quite frequently in recent years: analytics and AI. Often they are used interchangeably and are practically synonymous.

Organizations—​as well as the software vendors that supply their needs—​have largely tapped analytics to provide deeper information beyond basic indexed searching, which typically involves applying Boolean logic to keywords, date ranges, and data types. 

Search concepts have expanded to filter out application-specific metadata (e.g., parsing mail distribution lists, application login time, login/logout/idle times in chat and collaborative rooms, etc.). Today's search also includes advanced capabilities such as stemming and lemmatization—methods for matching queries with different forms of words—​and proximity search, allowing searchers to find the elusive needle in the haystack.

The latest whiz-bang features that are all the buzz within the information governance space are analytics (or predictive analytics) and AI (or artificial intelligence/machine learning). These are here to stay, and we are just beginning to scratch the surface of their many uses.

Here's what you need to know about predictive analytics vs. AI, and why the differences matter.

Analytics, defined

Analytics (or predictive analytics) uses historical data to predict future events. Typically, historical data is used to build a mathematical model that captures important trends. That predictive model is then used on current data to project what will happen next, or to suggest actions to take for optimal outcomes. 

Analytics as we know it has deep roots in data science. Combined with the ability to view archived data in a more 3D-type analysis, analytics can provide deeper insight beyond basic Boolean search. 

Based on prior history and outcomes, organizations can gain deeper insight into trends and patterns regarding employees, customers, and competitors. You can also mitigate risk, and predict success and security. This is a result of capturing and analyzing current data from multiple channels, including emails, files, instant messages, CRM applications, relational databases, collaboration tools, and social media.

“With increased competition, businesses seek an edge in bringing products and services to crowded markets. Data-driven predictive models can help companies solve long-standing problems in new ways,” notes Mathworks.

How artificial intelligence differs

AI has existed for a long time. But machine learning is actually being developed. 

Machine learning, an AI technique, is a continuation of the concepts around predictive analytics, with one key difference: The AI system can make assumptions, test, and learn autonomously. AI is a combination of technologies, and machine learning is one of the most prominent techniques utilized in information governance to yield deeper insights about data.

In machine learning, algorithms are "fed" data and are asked to process it without a predetermined set of rules and regulations. Predictive analytics is the analysis of historical data as well as existing external data to find patterns and behaviors. 

Machine learning typically works by combining large amounts of data with fast, iterative processing and intelligent algorithms, allowing the software to learn automatically from patterns or features in the data.

Real-world use cases

A good example of analytics versus AI would be online retailers using search and buying habits to predict a customer's next likely purchase and then targeting their ads and advertising emails based on that prediction. My parents bought an acoustic guitar for my 10-year-old daughter after a few successful music lessons. It was a surprise, even to me; as it turned out, I had purchased a mini electric guitar for her the same week.  

What fascinated my parents after the fact was that both online retailers from which we purchased the guitars were obviously using either analytics or AI software. My parents' inbox was deluged with offers for folk, rock-and-roll, and heavy-metal sheet music, grunge T-shirts, guitar accessories, and concert events. 

My targeted advertising however, was clearly using AI; the ads and spam were more geared toward iTunes, Google Play, Spotify, and Bluetooth headphones. 

Predictive analytic algorithms (in their infancy at the time) assumed a guitar buyer would likely purchase accessories, was over the age of 18, and played specific genres of music—clearly not my daughter.

What happened was that my daughter wanted more guitar-based music in her library to listen to and later requested headphones to listen in privacy. In this instance, the AI software more accurately predicted my next likely purchase. 

Behavioral analytics

Another example of predictive analytics is in behavioral analytics. Let's say Bob is the lead sales rep for a profitable manufacturing firm. Bob is knowledgeable about his industry and often creates thought-leader content for his firm, which is met with delight by the marketing department. 

Bob is about to close the largest account in company history. His typical usage of the almost 1,000 files in the company's records management system and/or SharePoint is to access files two to four times each week and download an average of eight documents. 

Behind the scenes, analytics are tracking Bob's access to both the applications and files and are using basic machine learning to log Bob’s behavior. By the end of the fiscal quarter, Bob has accessed over 275 documents, and he has downloaded over 100 documents in just the past 36 hours. 

But all may not be as it appears. Predictive analytics can determine that Bob did not receive his bonus or promotion for the quarter, and is downloading files to resign and take "his" intellectual property to a competitor or to start his own firm. 

By identifying Bob's unusual behavioral pattern, predictive (behavioral) analytics is in this instance acting as an early resignation-detection system. This is particularly helpful to organizations looking to mitigate risk. 

Archiving with analytics and/or AI

Many vendors in the archive space are touting product features that indicate a layer of analytics and/or AI. Many are still in their infancy. In some cases, the policies, search terms, patterns, and algorithms are not fully developed as yet, while new innovations are being developed. 

The needs are specific to each vertical and are further defined by each business's internal policies. Further, software vendor A's analytics are engineered quite differently from those of software vendor B's. 

There are no universal standards, and an out-of-the-box schema will not likely succeed. Building business policies around historical events, while tied to business goals, takes time and subject-matter expertise on both the part of the company and the software vendor. 

We are far off from the time when an AI system thinks for itself and can decide to not open the pod bay doors, as Hal did in Stanley Kubrick's classic movie. The hope is that we will have taught these systems to discern, and we will always retain the ability to manually override. 

As this industry continues to advance, we are, figuratively, building the AI / machine-learning spaceship while flying it into orbit.  

Keep learning