Artificial Intelligence & Machine Learning 101: A Primer for Marketing and PR Professionals

9 min readNov 29, 2017

Artificial Intelligence (AI) and Machine Learning (ML) are transforming how marketing professionals work and operate. And although many of these technologies are decades old, much of its potential is only now being realized. So, if your only knowledge of Artificial Intelligence came from one article in an in-flight magazine, I’m here to help dig your head out of the sand. Keep reading to discover what AI is all about, how it’s currently being utilized in the marketing world, pitfalls you may encounter along the way and how it can improve your organization.

Let’s define some terms
Artificial Intelligence and Machine Learning are often used interchangeably in the popular press. Yes, there is some overlap, but they are two distinct things.

Artificial Intelligence isn’t, technically speaking, a real thing. It’s an overarching term for a variety of technologies. Artificial Intelligence is the general label for the field of study devoted to making machines intelligent. You could also say it gives software super powers.

All the other stuff you may hear about related to AI — Data Science, Deep Learning, Neural Networks and Machine Learning are more precise subdomains for scientific and engineering methods under the big AI umbrella.

Machine Learning is one of the more talked about and used subdomains of AI. It uses (large amounts) data and algorithms (a set of instructions telling a computer what to do) to discover useful patterns between the pieces of data.

What you should know about Machine Learning is that in the past, if you wanted a machine to do something, you had to specifically program it to do that one thing. The computer would get a series of inputs and then output a particular solution — almost like how a recipe works. But now, if you have enough data, Machine Learning can help computers learn without being explicitly programmed.

The machine essentially takes patterns discovered in the data and make inferences about the behavior of future cases. Humans do this sort of thinking every day. Let’s say you run into a guy with slicked back hair, large sideburns and wearing a jumpsuit with flashy gold sunglasses. Your mind instantly takes these characteristics into consideration, ponders the relevance of each, discovers which are the most predictive for your situation and makes a decision. Since you’re in Memphis, you determine this guy is an Elvis fan. Machine Learning does the same thing. It can even grow and change it’s opinion with each iteration and as information evolves.

Here’s a more concrete (and overly simplified) example. Let’s say you have a huge customer database and you want to find out which customers are most likely to leave for a competitor in the next quarter. The machine will go and figure out the attributes of each customer that best solve for this based on previous experience. Machine Learning would funnel all your data down (Figure 1) to find that the stuff that matters to solve this problem is just a small set of criteria (Figure 2). Keep in mind that Machine Learning doesn’t care if the correlation is logical. It only matters that it is actionable.

An overly simplified example of machine learning in action

So, if there’s one thing you take away from this entire piece, make it this. Artificial Intelligence is a buzzword for an intelligent machine and Machine Learning is the ability for a machine to learn without be explicitly programmed.

You’ve used AI and Machine Learning before (but may not have realized it)
Unless you’ve been living in a bunker somewhere in North Dakota, there’s a good chance you’ve had at least one interaction with the big four — Apple, Amazon, Facebook or Google. Here are just a few ways you use and help train their AI systems everyday.

By using Siri to send a message or asking Alexa to play a song you’re using a subdomain of AI called Natural Language Processing. It helps computers understand human language as it is spoken. Siri listens for her command, takes to your words, translates them into something actionable (for her) and then executes on it.

If you’ve had an ad (from Amazon or any number of online retailers) follow you to each and every site you visit, you’re dealing with programmatic advertising. Programmatic advertising is a series of algorithms that allow marketers to deliver the right ad at the right moment, based on hundreds of factors like your demographic information, past online behavior and the content you’re looking at when the ad appears.

Examples of AI training with captchas and facial recognition software

To confirm you’re not a robot, websites will often use captchas that make you pick all the images that have road signs or cats in it. Here you’re helping to train the accuracy of Google’s AI. It could be to improve their image search, Google maps or even to train their self-driving cars.

If you’ve swapped faces with your bestie or turned yourself into a half human, half dog then you’re using computer vision and object recognition. These systems take their knowledge to identify everyday objects and learn that certain objects have certain patterns and features (in this example, they know that most people have two eyes, a nose and a mouth in the same general place). Beyond just vomiting rainbows, this technology is used in marketing research to measure audience reaction when exposed to different ads or creative.

One final example is customer analysis. Brands will use large scale data to analyze and sort their customers into different groups based on their demographic information, past purchases, offline behavior, online browsing history, etc, etc. Through predictive analytics, marketers can identify when consumers are going through major life events — the time periods during which they are most likely to switch up their shopping habits. In one famous example, Target used a customers’ past activity to send mailers featuring baby products to a woman it predicted to be pregnant, tipping off her father before his daughter had told him the news.

Why is AI such a big deal now?
AI needs three main components to be successful — money, processing power and tons of data. There have been iterations of AI since the 1940’s and 1950’s with on and off success, but nothing like we’re experiencing today. The reason AI is catching on now is because we’ve hit a sweet spot. Today -

we create tons more data than in the past
processing power is exponentially faster and cheaper
we’ve improved our algorithms
there is broad investment in AI from governments, venture capitalists and universities

How can AI improve your marketing and work?
Machine learning and other forms of Artificial Intelligence are becoming a popular replacement for manual decision making because (1) they are often more accurate than human experts, (2) operate faster — often automating decisions on millions of records in a matter of seconds, (3) they’re unbiased (when designed correctly), and (4) are cheaper to deploy than humans. Here are a few ways that teams are currently employing various new technologies to improve their marketing and PR.

Artificial Intelligence is used to discover new markets and niches. Machine Learning will slice and dice customer data to identify customers who are undermonetized relative to their peers. For example, it might identify a customer behavior in a particular region of the country that requires a specific product assortment and pricing mix. It may even show that this particular niche does a lot of late-night shopping and has a high propensity to abandon their cart.

Media teams are already optimizing their marketing mix and media spend with the help of AI. I don’t believe it’s too far off to think of an Echo-like device on your desk with access to research papers from the American Marketing Association, general business strategy and previous quantitative results.

Me: ‘Hey dingus, I need to write a media plan for a shampoo’
Dingus: ‘I’ve analyzed the social media feeds of everyone who cares about shampoo, and here is the best media mix to reach them.’

AI can improve speed and accountability in the creative process. In the short run this will accelerate production tasks like layouts and versioning, but it can also help you develop more relevant and faster content. Automated Insights offers a product called Wordsmith that will take a spreadsheet of data and instantly write relevant text. This could be articles on quarterly earnings reports, sports box scores, hotel descriptions or even earthquakes. Similar technology has also spread to video. Wibbitz can create accompanying videos automatically based on the story inputs. See it in action at USA Today.

Like you may have seen with programmatic advertising example earlier, corporations are getting better at personalizing advertising and recommendations to target consumers based on multi-modal data (mobile, social media, location, etc). Red Roof Inn previously used an algorithm that determined the opportune time to message customers on nearby hotel availability and rates. They used information based on weather conditions, flight cancellations and customers’ locations to offer deals to stranded travelers. Smart creative, smart targeting and smart bidding

Machine Learning can increase savings through better contract negotiations (hooray, less lawyers!). By feeding a machine your history of contracts, it can be trained to highlight opportunities for savings by matching suppliers that better align with your needs. Through this procurement method, ML can also help capture revenue that would have otherwise been missed.

Though marketing organizations and agencies all use data and analytics to some extent, most decisions today still come down to subjective hunches. In the not too distant future, AI will become critical to decision making and results tracking by delivering better understanding of our target audiences and where we should deploy resources.

Things you should be aware of
As with anything, all that glitters is not gold. AI and Machine Learning can be amazing, but these technologies are still being trained. Here are a few things you should be aware of.

AI systems are only as good as the data you feed them. When Machine Learning systems fail, it’s rarely because of problems with the algorithm. Human bias can poison the results if the initial data or systems are based on unfounded preconceptions or stereotypes. A great example of this is “fake news” on social media sites. Often these types of systems are optimized based on engagement, but don’t heavily factor in (or have the brains to figure out) if the information is real or not.

Be aware of potential legal ramifications if you work in certain industries. If constraints aren’t implemented, these AI systems may overstep the law. For example, Machine Learning at a brewery may find that there’s an untapped market of 12–14 year old boys and then proceed to target ads to them. Of course, marketing alcohol sales to underage kids is off limits.

How do you know how to breathe? That’s a tough question, right? The same can be said with many AI systems. The more information that gets fed into them, their creators may not really know how advanced algorithms do what they do. This could leave many individuals or corporations accountable for the system’s actions.

And finally, regardless of how cutting edge AI may seem, it can just be a waste of time and money if it doesn’t align with your organization’s objectives, working practices, current technologies and culture.

Gaining more AI knowledge as a marketer

“Artificial Intelligence, deep learning, machine learning — whatever you’re doing if you don’t understand it — learn it. Because otherwise you’re going to be a dinosaur within 3 years.”
Mark Cuban at the Upfront Summit, February 2017

Yes, the Cuban quote above can be a little scary, but I believe you don’t need to moonlight as a data scientist to be an effective marketer working with Artificial Intelligence. I believe smart marketers and PR professionals can and will find a balance. They’ll tackle the creative and more emotional intelligence parts of the job while the machines will support them with strong decision making data.

That being said, you should go beyond having a basic understanding of these technologies. The better you can communicate with data scientists and computer engineers, the easier it will be to apply these tools to your marketing problems.

Like the video from Silicon Valley example above, this is still a very new and emerging field. You are not behind, but I believe it’s in your best interest to start understanding these technologies and how they’re used sooner rather than later.

Below I’ve added some more beginner level articles and books that I’ve found really helpful in my learning this year.

Artificial Intelligence & Machine Learning 101: A Primer for Marketing and PR Professionals

Written by Bob Hazlett