Artificial Intelligence and Machine Learning in Marketing: What marketers (even those who don’t care about tech) should know about AI and ML

This article was published in the MarketingSherpa email newsletter.

After sharing an experiment pitting an experienced marketer against artificial intelligence and machine learning tools in Marketer Vs Machine: We need to train the marketer to train the machine, Flint McGlaughlin went on to say, “There is no way a marketer can process or even comprehend the tidal wave of data and tools.”

So true. Scott Brinker of ChiefMartec tracks marketing technology. His first Marketing Technology Landscape Infographic had about 150 solutions on it.  

Creative Sample #1: Marketing Technology Landscape, August 2011 (published with permission from Scott Brinker)

And that in itself is a lot to keep up with.

Now, 11 short years later, the chart looks like this…

Creative Sample #2: Marketing Technology Landscape, May 2022 (published with permission from Scott Brinker)

Where’s Waldo?

The 2022 Marketing Technology Landscape has 9,932 solutions on it.

I got into marketing because I’m a creative. When I started as an advertising copywriter, things like Quark XPress and the PDF were the main cutting-edge technology. You may laugh about it now, but the ability to put ad concepts together on the computer and send those concepts to clients without having to FedEX them was groundbreaking at the time.

If you’re in the same boat as me – not necessarily old, just less focused on tech in your career – this article is for you.

This isn’t a definitive guide by any means. Rather, we’re trying to give you just enough insight so you can understand where one of the hottest areas of the martech landscape is right now, allowing you to focus on what you really need to do – optimize that funnel, help potential customers, build some pipeline, whatever it may be.

And spur some ideas in you for how you can leverage artificial intelligence (AI) and machine learning (ML) to help you meet those goals.

Takeaway #1: AI can help you scale

If you break down what we’d want from any new technology, it would likely come down to two things – allow me to do more, and allow me to do it better.

Think of the airplane. This technology has allowed us to do more – essentially, we could travel to many more destinations than previous technologies. (Is it better though? For anyone who has spent time in security lines and coach lately, perhaps a cross-continental horse-and-buggy ride would be better).

How about a technology as simple as the electric toothbrush. I can’t necessarily brush more of my teeth, but I can brush them better.

We’ll get to “better” in the next takeaway, but first a quick example of “more.”

“We spend $500,000 annually on Google Ads. It would not have been possible to scale our digital marketing without AI and ML,” said Apurv Sibal, Vice President, Best Views Reviews.

For example, the team uses AI to generate keywords for search engine marketing campaigns. And it uses Google’s Dynamic Search Ads to automatically generate ads based on the website’s content.

Takeaway #2: The machine is only as good as the signals you give it

I worked in the software industry for many years, and I used to joke that the way a lot of enterprise software was sold was akin to “software is magic.” It really only took three steps:

1. Buy our software product
2. The software does its special magic
3. And… abracadabra… your company gets great results

Anyone involved in a failed enterprise software implementation can tell you what a joke this is.

When AI and ML came along, it seemed to me like the next generation of this sales message – AI is magic on steroids.

But the machine doesn’t work unless you do. You need to do what humans do better than the machine (determine the right strategy for your business) so it can do what machines do better than humans (process MASSIVE amounts of data).

Kevin Daly, Director of First Party Data, Making Science, provided an example from working with Bark, a services marketplace.

Individual customers post requests for a professional on Bark, and professionals can purchase credits they then use to acquire these leads.

“Bark’s objective was to target its campaigns to attract professionals who use credits on a recurring basis, those who are valuable. The main pain they had was that the volume of conversions, professionals who clicked on the Search campaign and ended up buying credits within seven days, was so low that Google’s algorithms could not learn and therefore could not optimize for these users,” Daly said.

So, the team proposed creating a new funnel – adding an additional type of signal to the ones the company already had. These new signals are generated when professionals subscribe for free (a professional can sign up for a free subscription but cannot respond to leads without paying a one-time lead cost or applying credits through an ongoing paid subscription).

This funnel strategy also ensures that priority is given to new users, users who have never subscribed before. A machine learning model assigns this user a value that coincides with the expected revenue in 90 days, in terms of credit consumption plus monthly fee contracting.

This project was activated in US campaigns and multiplied the volume of signals (new professionals with potential recurring credit consumption) by a factor of 20 and tagged them according to their value. This allowed Google Search Smart Bidding campaigns, with tROAS (target return on ad spend) bidding strategies, to have enough data to learn and optimize towards the professionals that Bark had as KPIs (higher-value prospects).

The result – the team increased high-value sign-ups by 54%, increased sales revenue 25% since sign up in a 30-day period, and decreased the cost per top sign up (professionals of greater value for the company) by 12%.

Takeaway #3: Humans must build business strategies

The previous example shows how tweaking a funnel strategy to make better use of machine learning and AI improved results. But sometimes the best funnel strategy is an entirely different strategy altogether.

Again, this is where the human must put the AI to best use and can’t simply apply the right AI to the wrong strategy.

“My biggest lesson from using AI in marketing is that you can't blindly rely on recommendations from AI tools,” said Karthik Manoharan, Co-founder, WeCodee. “We spent around $10,000 on ad-spend over a period of three months, with $300 for the tool. The result – crickets. Why? Because the channel was inappropriate. Our target customers were C-level executives who won't trust an ad.”

“We pivoted, focused on content marketing efforts and utilized Hubspot for automation, and got fantastic results. Why? Because the tool is catered to inbound marketing, something that works for our industry and niche. We continue to generate around 5,000 monthly visitors for our blog and increased our DA (Domain Authority) by 10 points.”

Takeaway #4: AI can write for you…kinda, sorta…well at least in certain situations

Of all of the things AI can do in marketing, I think the most [everyone's got a different opinion of AI, so reader, feel free to insert your preferred skeptical or hopeful adjective here] one it can do is actually write.

After all, writing is fundamental to the human experience. “Let me live, love, and say it well in good sentences,” Sylvia Plath said. A machine can’t live or love, so can it craft good sentences?

More to the point for my readers, writing is also fundamental to marketing success. And good writers are expensive, deservedly so. So, a quick and cheap machine is very enticing.

When I looked for sources for this article, writing was a hot topic.

The AI failed some marketers.

“We tried using numerous AI-text generation tools to speed up and decrease the price of our articles,” said Matic Broz, Founder & Head of Content, Photutorial.

“While these tools helped in a few cases, they were not useful for our affiliate content. The main reason is that you need to write reviews based on your opinions, but the AI text is, for now, too mainstream. And it's not very good at writing reviews. We occasionally use AI to generate definitions; however, even these need a ton of editing.”

Others found success but weren’t ready to turn every aspect of writing over to AI (see Takeaway #5 for an example of that). They discovered AI could help with writing in certain use cases. For example, reducing the time it takes to write.

“The truth is that we are not able yet to copy and paste an entire article and have a different one in one minute. What is usually done is to mix sections of different articles modified with AI and make slight manual modifications,” said Marco Genaro Palma, SEO consultant, GenaroPalma.com.

“In terms of [search] positioning I haven’t seen any difference, but I have seen a big difference in the time we spent on content creation. My copywriters used to take 10 hours to write an article of 1,500 words and today they are doing it in two,” he said.

Or using the machine to figure out the machine – that is, to help understand the search algorithms.

“In the past few months, we've been trying to crack the best ways to rank our content on Google. After many failed attempts that get us between four to 10 clicks per month, we implemented two AI-assisting tools to help us create content that ranks better,” said Ofir Kruvi, Product Marketing Manager, Jika.io.

“The results were outstanding – as a rather small company we were able to rank among some of the most gigantic sites. Our best AI-assisted article was about Nancy Pelosi, it got to 10,000 impressions with a CTR of 2% in less than three months. To put it into comparison, all of our non-assisted content (around 20 articles) reached 200 impressions all together in six months,” he said.

Takeaway #4.5: Use AI ethically

Added on October 10, 2022

After publishing this article, a keen-eyed reader shared on Twitter his concern that some marketers could use artificial intelligence to rip off others’ content, a practice known as content spinning or article spinning.

I’m glad he brought it up, because I didn’t even think to address it in this article. But c’mon… don’t be shady.

While ripping off others is certainly nothing new, technology can magnify the problem. As a writer myself, I’ve often seen my articles stolen. While content scraping is frustrating enough, black hat AI-written content is even more infuriating because they mangle your writing. I’ve stumbled across my writing butchered on LinkedIn by AI mixing it with other articles. I’ve even had poorly AI-spun descriptions of the How I Made It In Marketing podcast emailed back to me in pitches.

What prompted the concern about AI spinning of others’ content was the above quotes from Marco Genoa Palma. I didn’t go too deeply into any of the examples in this article. Unlike our case study articles, these sources were just meant to provide you many quick snippets illustrating how AI is being used in marketing.

But the Twitter comment raised a very valid point. So, I asked Palma if we could go deeper into his process, so you can see more clearly how his team uses AI tools. He obliged…

“In most cases, my clients have enough money to pay copywriters and they create unique and high-quality content according to SEO guidelines. Everything runs smooth, copywriters do their task and just with some SEO optimization, pieces of content are ready to be published,” he said. He calls this the classic method, and it usually cost around 500€.

But for some SMB and lower-budget clients, they cannot afford this, so he has developed a less-expensive, although admittedly lower-quality, option.

Since he’s from Argentina, he can easily find non-native-English copywriters that will write 1,000 words about a subject inexpensively. He gives them articles as inspiration for how he would like the article to look.

“Since the articles I get are lower quality (grammar mistakes, incoherences, etc.) I usually order two to three per subject. I choose the parts I like from each, I put it on a Google Drive document, and I use a tool called WordTune,” Palma said. “Once in the Drive, WordTune will make me (or my English native copywriters) choose how to rewrite each sentence.”

“Once I have the article done, I keep all this information for me, and I feel free to reuse it when I need it. It might be one section, a full article. It really depends,” he said. This method costs about 80€ for an article.

Palma also tried with cheaper countries than Argentina, but the quality-price ratio was not so good. And he tried an AI-writing tool that writes articles from one title, but results were very poor.

Takeaway #5: You’re not the only one with AI

As mentioned above, Google has been using AI and machine learning to try to improve ad performance. Makes sense, that’s how they make their money.

And marketers are using are using machine learning to help assist them in figuring out Google’s search algorithm to improve rankings.

However, assist is a key word. It can be taken too far as well.

Every indication would suggest that Google is also using machine learning and artificial intelligence to sniff out AI-created content meant to attract search traffic.

If I was a bigger sci-fi fan, I’m sure I could drop in the perfect movie or book analogy here of when the machines just started fighting each other and bypassed the humans entirely.

To give you an example, here is a cautionary tale of being overly reliant on AI for your content.

“I have been using AI to write blog posts since August 2021. I've gotten millions of impressions on one of my pet websites that I created entirely based on AI-based content,” said Matt James, Founder, Visitingly.

Creative Sample #3: Original traffic using AI-written content

However, in the recent Google search algorithm updates, James noticed a drop on this site as well as some of the other AI-content based websites he built, including a 42% drop in traffic to the aforementioned pet website after Google's Helpful Content Update.

 Creative Sample #4: Traffic drop to AI-written website after Google algorithm update

“I have AI-generated content websites in many niches, and I'm seeing similar drops, especially just after Google's Helpful Content Update had finished rolling out,” James said.

Takeaway #6: AI can help make your website interactive

Remember when ecommerce was just starting? One of the promises was that now every store could be like the corner bodega – open 24/7.

But that corner bodega had an actual human you could interact with. Ecommerce, not so much.

Enter, the AI-powered chatbot.

Sometimes referred to as customer service (it can probably be best used on the front end of a customer experience much like a phone tree, instead of as a true customer service replacement that might frustrate more than it truly helps), in my opinion a better way to think of it is interactive marketing. Even though all online marketing is sometimes referred to as interactive, if you look at most websites and landing pages, there really isn’t that much interaction going on.

AI-powered chatbots are a way to interact. I think of it as a choose-your-own-adventure guide to all the content that could be littered across your website and not so easy for the customer to unearth.

It’s a tool that can work for B2B websites.

“We utilize a chatbot on all of the Schlesinger Group sites. We monitor users’ behavior on the site and utilize the chatbot to speak to where we think they are in their buying journey. We are able to filter and redirect the non-B2B audience to other related Schlesinger Group sites for panels and better qualify B2B prospects before passing them on to a sales rep by utilizing a few questions that are delivered to the prospect by the chatbot,” said Ellie Ahmadi, CMO, Schlesinger Group.

She continued, “Our chatbot focused on our digital qual products, for example, has a 22% conversion rate for emails provided and a 10% conversion rate for meetings booked. In total, 30% of opportunities through the chatbot are new logos resulting in 32% of our chatbot-sourced opportunity value this year.”

And it can also work for B2C websites.

“Another great use case of how conversational AI can help marketers is by reinforcing branding. For example, the bot we created for our client, Medieval Times, speaks like one of their knights, queens, or squires. Many companies will try to get their agents to consistently talk in theme, but it can be difficult to get humans to oblige, especially as the midday fatigue kicks in. Fortunately, our assistants will not roll their eyes when asked to perform in character all day,” said Marco Medugno, Data & Insights Coordinator, Satisfi Labs.

Creative Sample #5: Medieval Times chatbot

Related resources

Benefits of AI in Marketing: How do the views about artificial intelligence in marketing differ between leaders and practitioners? [chart]

Startup Marketing: How to build a marketing program from 0 to acquisition by Intel, be BOLD, learn as you go (podcase #28) – Interview with Maya Perry, Marketing Director, cnvrg.io (an Intel company), a machine learning operating system

Marketer Vs Machine: We need to train the marketer to train the machine