Machine Learning: How to Turn Artificial Intelligence into Relevant Conversations

Blog Post
March 27, 2018

Sure, IBM’s Deep Blue knew how to maneuver a rook for a serious game of chess, but when it comes to anticipating your best customer’s marketing needs, the artificial intelligence program is kept in check by one sweeping figure: 2.5 billion.

That’s roughly the number of gigabytes of consumer data generated every day – a deluge far greater than a human brain can process. It’s even more than some artificial intelligence programs can keep up with, and a sizable dose of that flows into marketing.

Enter the subsets of artificial intelligence (AI), such as machine learning (ML). If AI is an umbrella technology that gives machines the ability to seem like they have human intelligence (like Deep Blue or IBM’s Watson), then consider ML the self-opening mechanism within that umbrella.

When AI Goes ML: Spotting the Difference

ML occurs when computers take large sets of data and learn from them by tracking patterns and then recognizing trends to the point that the software can predict the user’s needs or actions. Well executed, ML can enable an organization to tailor messaging and offers right down to individual customers by enabling the kinds of conversations that would resonate with specific shoppers at particular times.

The oft-used example is Amazon’s “you might like this” product suggestions based on previous purchases, but more advanced uses are emerging—including facial recognition software. The California chain CaliBurger is testing facial recognition in kiosks to identify the order preferences of reward members and make suggestions.

A few examples of AI (and ML) in our world:

  • Voice-recognition. This involves software systems that learn commands based on the user’s voice. Nearly all cell phones and cars now include this as a standard feature, but Google and Amazon have evolved the technology through machine learning for their Echo and Home smart technologies.
  • Smart computers. These devices, such as navigation systems, can combine programmed technology (driving routes) with real-time intelligence (the infamous phrase “recalculating”). When the system stores a user’s previous driving routes and suggests one based on time of day or location, it’s using machine learning.
  • Robotics technology. No longer so futuristic, robotics power everyday tools from self-directed vacuum cleaners to surgical procedures. But when the robot responds to a facial expression (through face-recognition technology), it’s using ML.   
  • Chatbots. These interactive tools, when using limited intelligence, respond to common actions on websites and in social media networking sites. But many brands have amped them up through ML. North Face, for example, will ask online shoppers what specific ways a coat will be used and then make recommendations.

Machine Learning in Marketing

Machine learning, like the spring activation on an umbrella, was designed to enable artificial intelligence to better capture and make sense of the downpour of data.

Simply stated, it takes the data, analyzes it and then creates predictive models that, over time, continue to evolve on their own. Marketers are then able to fish out nuggets of “predictive” knowledge from the waves of data.

Most machine learning requires some guidance or training at the start. After all, machines can’t completely replace humans in their ability to accurately interpret all scenarios. However, after an initial training period, many ML technologies can continue to learn and adapt with impressive accuracy. Marketers could use ML for purposes well beyond purchase suggestions. It can, for example, determine the best phrases to elicit desired behaviors from consumers — the words “blissful results” might get five times more responses than “satisfaction guaranteed,” perhaps.  

However, as effective as ML is, the key to success is combining humans with the technology to refine campaigns. As ML provides insights into content and formats that resonate, marketers should test new approaches until they reach full optimization. Of course, since much of marketing is based on human reaction, campaigns must continue to evolve as consumer preferences change.

Detecting the changes in behavior is where ML comes in but, like any umbrella, it needs a human to push the correct button. Even Deep Blue couldn’t do that.

You might also like: What Is Machine Learning—And How to Use it to Supercharge Your Marketing