Decoding Artificial Intelligence into Useful Insights for Content Marketers

Posted by

[An Interview with Paul Roetzer, Founder of the Marketing AI Institute]

“Artificial intelligence” sure gets a lot of “ink” these days.

Being the glutton of marketing and technology blogs that I am, I see the phrase multiple times every day. I must admit, for however long the topic’s been seducing writers and readers, I hadn’t paid it much mind.

In my cynical mind I wondered, “Isn’t ‘artificial intelligence’ just another way of saying ‘computing?'” Is it, uh, artificial? I mean, is it science fiction? Is it the real deal? Is it meaningful enough to become yet another area of marketing I need to keep up on?

I decided I’d 86 my boycott and begin making an attempt to understand AI. I suppose what I really wanted to understand was why had AI seeped so heavily into the conversation about marketing and how might it change things? Would little talking cylinders soon put not-so-artificial thinkers like me out of business?

One thing I did know about AI as it applied to marketing was that my friend Paul Roetzer—agency founder, author, speaker—chose to create the Marketing Artificial Intelligence Institute in 2016. I knew if anyone could decode AI into useful insights for you, me and Taboolagins here on the blog, Paul was that person.

Paul was nice enough to allow me to pick his brain for 25-minutes or so. You can check out our entire conversation in the video below—or read the highlights that follow, if you’d rather.

Some of the topics we discussed include:

  • What’s artificial intelligence?
  • Is it changing marketing automation or websites?
  • Is it taking jobs away from marketers?
  • How does AI affect online advertising?
  • Paul’s simple (and slick) “5 Ps” of marketing AI.

Barry: Paul, I’ll ask you to be a dictionary for the moment and define artificial intelligence.

Paul: I always joke that you could go to Google and search, “What is the definition of AI?” and you’ll get ten different definitions from ten different experts.

Demis Hassabis, the founder of DeepMind—which Google acquired in 2014—defined it as the science of making machines smart.

I like that for its simplicity. But for visualization purposes it’s really an umbrella term for all the tools and technologies that are rolled into the idea of being able to make machines smart. Machine learning and deep learning are the two most common types or technologies you hear about within AI.

Barry: So what’s new with AI? It seems to me what’s new is the computers behind the scenes can do things that otherwise humans would have had to do.

Paul: Yeah, that’s a good way to look at it. I always tell people you’re using AI dozens if not hundreds of times a day in your life and you don’t know and you don’t care.

Like if you’re driving in your car and you use Google Maps, there’s no human coder sitting there saying turn left there’s an accident ahead. The machine is learning what’s happening and telling you which way to go by predicting the fastest route.

If you watch a show on Netflix and it recommends something to you, it’s likely a recommender algorithm. Amazon: what to buy next. Your iPhone: when you unlock the phone with your face, that’s facial recognition. Literally it’s just everywhere in our lives and you don’t ever sit there and say, “I bet that’s AI that did that.”

Barry: Are marketing automation platforms fulfilling the promise of artificial intelligence?

Paul: No. I said it before: it’s early, very early. There are a lot of really interesting startups. Some of them have upwards of 50 million or more in funding and even they’re just scratching the surface of what they’ll be capable of doing with AI. Many of the software platforms have been investing in AI. Like if you look at Salesforce, they’ve bought around 10 to 12 AI-powered companies that they’ve rolled up into Einstein and it creates an intelligent layer over their CRM platform.

We just did a feature on Adobe and they’re making massive headway into this space and trying to create something called Master Plan that actually will do marketing strategy using AI. You see a lot of the big players starting to integrate more machine learning into it.

It makes sense because with marketing automation and CRM—whatever area you’re looking at—every decision we make should be based on data. That’s where machine learning excels. It takes historical data and makes predictions about what will happen in the future. So whether it’s email open rates, or the likelihood of a lead converting, or the likelihood of a client leaving, or churning… all of those things are predictable based on past data. But right now, most of the time, it’s humans sitting there trying to figure that stuff out.

Barry: Artificial Intelligence scares people who want to go into marketing, thinking my job can be replaced by robots and machines. What do you think? What’s the implication for the future of marketers?

Paul: You could say that about every profession in the world right now, I guess, it’s not unique to marketing.

I would say in the foreseeable future, more than anything, AI is going to enhance our knowledge and capabilities. It’s going to be an assistant. It’s going to take a lot of the burden of data analysis and trying to make these predictions. It’s going to take that off of the marketer and let them focus on the things they’re uniquely capable of doing, like strategy and creativity.

In the near future there will be career paths that will go away, there will be jobs that will be lost, but the assumption by most people is that the net gain will be greater than the few professions or disciplines within marketing that would go away.

Barry: Is it conceivable that AI redefines what a website or a website experience becomes?

Paul: Yeah. In my 2014 book Marketing Performance Blueprint, I talked about the idea of “Amazonification” of marketing. This idea that brands wanted to create this one-to-one approach, whether it’s the website or email or whatever, that your experience is different than mine. That was before I really dove into AI deeply.

What I would say today is yes, the site should adapt to the individual person. So right now if you’re a marketer you may go in and say if someone downloaded this eBook then show them this next eBook. Or if they visited this page then show them this custom content. Or you may have a chatbot pop in. You’re doing limited things to personalize the experience, but it’s all human driven.

You’re writing the branching logic that decides all of it in the future. And even some technology today enables this. The machine should make all those decisions. That is absolutely something a machine can do better than a human, because what you’re doing is you’re trying to predict what will be of value to them next based on past behavior. So if they looked at one eBook, that’s no different than Netflix trying to recommend the next show to you.

On a website it’s like if we know who you are and what you’ve done we should be able to predict with relatively high accuracy what it is you might do next. That’s where machines should absolutely take the burden off of humans and replace the vast majority of those roles.

Barry: At what point does AI become content intelligence or does AI crash into content marketing and make us better at it?

Paul: We’ve profiled 40 AI-powered companies so far on the Institute. We’ve tracked about 750 or so. A lot of the vendors we’ve profiled so far that we’re looking at are trying to solve pieces of the content puzzle.

There are the obvious ones like programmatic advertising, like digital media advertising and buying, that came very early in the AI space because there’s a ton of money there.

As you start going down, what we’re seeing is content strategy is ripe for disruption. Again trying to find what keywords to build, trying to figure what topic clusters, trying to figure out what content to write about… All of those things should be based on predictions of how content will perform. Again, we go back to the whole idea that machine learning is prediction.

Barry: How do you perceive that AI will affect native advertising?

Paul: I think much in the same manner we were just talking about. With native advertising you’re wanting to show ads that you know will be relevant. You’re trying to create and present content that drives people through, gets them to click, gets them to the next step. So the more informed that content can be and the less it’s humans trying to say, okay, if they didn’t click this, show that, or if they chose this, show this one.

It goes back to the idea of training Spotify or Apple Music. Like when you’re saying, “I like this, I don’t like that,” the same kind of thing can be done with content and with native advertising. If you’re saying, “This was helpful to me, this wasn’t,” that’s training the algorithm.

Another misnomer about AI is people sometimes think you can just flip a switch and put AI on it. That’s not how it works. AI requires a lot of data and it often requires a ton of training.

Barry: Chad Pollitt (in his eBook, Everything You Need to Know About Marketing Analytics and AI) says the use of machine learning is already creating cost-per-engagement models, CPE, as opposed to what we’re used to, cost-per-click or cost-per-impressions, cost-per-acquisition, and so forth. What do think?

Paul: I think it makes a lot of sense. I did read Chad’s eBook and I think the more we move towards the engagement model, the better off everybody is in the industry. Impressions have their issues. Everything else has an issue, but when you look at engagement it’s a viable way to start measuring the performance of content better.

Barry: You said there are five Ps for which marketing artificial intelligence will go to work. Let’s look at them. Planning.

Paul: The way we look at it is when we look at these 700 companies or so, we have a really hard time figuring out how they fit in together. And there wasn’t a logical, “this is a content marketing… this is SEO… this is an advertising.” And so the five Ps became the framework.

In the planning model, we’re looking at resource allocation strategy development primarily, but we would also lump in things like figuring out what to write about, figuring out which leads to pursue. Anything in that planning phase, and it can be across multiple disciplines.

Barry: Production.

Paul: Creation and curation of content largely. That’s mainly what we look at there, but it can also apply to websites. There are AI tools to build and design websites. Logo designs. So it’s not just written content, it’s really any content that can be used to drive the growth for a company. Creation of written content is the most logical, but it goes far beyond that.

Barry: P number three, which we’ve touched on already: Personalization.

Paul: The vast majority of what’s happening for AI in marketing is the attempt to personalize every experience because it creates the greatest opportunity to drive growth.

Paul: That’s a good question. I think people are getting used to it, almost immune to it. So there’s obviously a little backlash in GDPR, sort of threw a wrinkle in it for some people, some brands. But generally speaking what’s going to end up happening, the convenience that comes from the personalization will override the concerns people have for the convenience factor.

Barry: Back to our list, P number four: Promotion.

Paul: Everything that would traditionally fit under promotion within marketing. It’s actually taking the content, taking everything you’re creating and getting it out to the audiences, but being able to do it across channels, your own site, third-party sites, and doing that in a personalized way and being able to target people like you haven’t done before using AI.

Barry: P five: Performance.

Paul: It is critical, but this this is probably the earliest of all of the categories. There aren’t that many players doing this part well. It’s basically taking the data and turning it into intelligence and the intelligence into action.

It’s a really hard problem to solve. There’s lots of false positives, so when you’re looking at something and you think you’ve found that cause-effect relationship and then you realize it’s actually not.

If you take Google Analytics for example and you try to apply AI to everything happening on the site to make predictions about what will happen next or even to prescribe what you should do, there’s a whole bunch of stuff in there that might just be noise and may actually not help you get to a smarter strategy. So it doesn’t mean a lot of people aren’t trying to do it, but it’s a really difficult one.

A really practical example of performance would be if you go into analytics they have a thing called “Ask Analytics” now in Google Analytics and you can just ask it questions, like “How was our traffic in New York last month?” There are actually all kinds of AI being applied to be able to understand that question and respond to that question.

Just as a small example, that kind of capability could be carried over into every marketing platform. Whatever your platform is, you should just be able to ask questions of your platform. You shouldn’t have to dig through reports. You should just ask questions. It’s hard though. To enable the software to do that is not easy. But that’s the kind of thing we look at in the performance realm.