It’s uncommon today to find a new technology in marketing that does not promote Artificial Intelligence (AI) in some fashion. Some people understand AI as the computer thinking and making business decisions based on gathered customer information and other relevant data. Other people think of AI as a form of intelligent being that could potentially take over the world like Skynet in the Terminator movies.

At the company where I work, we are always working to improve our processes. Since we curate and sell gift baskets online, AI can play a critical role in improving our operations and marketing. But, first, when evaluating what tools to use, it’s best to understand what AI is and how it can be used in marketing.

What is AI?

Defining AI is really more complex that one would think. There is General AI which would cover both the above-mentioned business intelligence and Skynet examples. There is Strong AI which is just a theory but is where people get the idea for computers having the potential to become Skynet and take over human society. There is also Weak or Narrow AI which could be defined as a subset of intelligent activity that a computer can perform like image recognition or winning games. Deep Learning and Machine Learning are both covered in narrow AI.

So, let’s consider some uses for narrow AI as this is the only realistic implementation that would be available at least to the general public at this point in time.

Narrow AI

Narrow AI is a generalized description of activities that can be completed using machines or computers that mimic forms of intelligence but do not have the ability to expand on their potential for learning. Often, the quality of the results for narrow AI computational participation depends greatly on the quality of the data and/or the algorithm. Whereas a strong or more true form of AI would be able to learn outside of the given datasets and/or algorithm playground created, narrow AI is more limited. The two dominant forms of narrow AI currently would be categorized as either a form of Machine Learning or Deep Learning.

Machine Learning

Machine Learning is a subset of narrow AI whereas the “programming” or algorithm can alter themselves. This usually comes in one of several forms such as rules engines, pattern recognition or other forms of virtual model processing. In marketing, you may have experienced this in the form of ‘recommendations’ or perhaps fraud protection.

Machine Learning can start with one set of rules and as data is coupled with reactive data, adjustments are made to the expressed results. So, if you are shopping for a car, and you tend to spend more time viewing imported cars, the machine learning will adjust the decision-making process of what types of cars to show you and imported cars will become more prominent over time.

Deep Learning

Deep Learning is an even more elaborate rabbit hole to go down than the term AI itself. Deep learning is modeled after the idea of taking an assumptive guess at how the brain processes input data (we can’t figure out how the human brain works yet) and trying to mimic how the brain works. These models are called Neural Networks. Typically, the data, processing power and algorithmic complexity of these systems tend to dwarf machine learning systems.

Neural Networks currently show the most promise in terms of delivering on what most people consider helpful and autonomous intelligent computational activity, the type that would be helpful for marketing strategy and tactical participation. It is not common to see a marketing tool described in terms of what its potential version of AI is being utilized so how do we make this assessment?

AI in Marketing

The next time you see “AI” being promoted in marketing tools, some good questions to ask when evaluating whether to use them are:

  1. In what way is AI implemented? (Machine Learning? Deep Learning?) Note: If they say it is a true form of AI they are likely not knowledgeable about the differences and this is a marketing gimmick.
  2. What type of data sets will this system utilize?
  3. What algorithmic considerations and adjustments will the system utilize?
  4. What kind of processing power with the system need?
  5. How will the system need to be maintained as datasets and algorithmic processing requirements increase?

AI can certainly be a good thing, but as always, there is power in knowing what you are getting and why.