In 2018, Lexus released what it called the first advertisement scripted by artificial intelligence.
Lexus used IBM Watson to analyze 15 years of “car and luxury brand campaigns that have won Cannes Lions awards for creativity, as well as a range of other external data,” according to Variety.
Watson was able to identify which elements from the dataset would resonate with viewers.
AI-inspired PR stunt? Most definitely.
Yet it also points to a deeper, more important story:
AI is the future of advertising.
AI isn’t just creating ads, though commercially available platforms exist that use AI to create ads without human involvement.
No, AI is transforming what is possible in the world of advertising at every level, from ad creation to audience targeting to ad buying.
Andrew Ng, who built AI at Google and Baidu, highlights online advertising as a major use case for AI in his excellent AI for Everyone course. In fact, machine learning—a core AI technology—is used everywhere in major ad platforms to determine if you’ll click on an ad, and that’s just one use case.
But it goes even further than that.
There are already dozens of use cases for AI in advertising today, with brands using the technology to intelligently identify and segment audiences, build ad creative, test variations, improve performance, and optimize spend—automatically, in real-time, and at scale.
This has profound implications on both a brand’s competitive advantage when it comes to digital advertising and the careers of marketers who plan and run ad campaigns.
In this article, we’ll walk through a working definition of AI as it relates to advertising. We’ll look at some of the top AI use cases for advertising, so you get a better idea of how AI can actually create value for brands. And, we’ll show you some top companies that actually offer AI solutions in various areas of advertising, so you can start demoing tools.
A two-minute definition of AI
The term “artificial intelligence” is an umbrella term that covers a range of machines that learn, with the help of human or entirely on their own. In this way, AI technologies perform certain cognitive tasks as well or better than humans.
The best definition of artificial intelligence that we've heard comes from Demis Hassabis at Google DeepMind, who says that AI is the "science of making machines smart."
That means machines that can read and understand text, see and identify images, physically move around obstacles, hear sounds and understand them, and sense their external environment.
What are some examples of AI?
You use AI dozens or hundreds of times each day.
Gmail and Google Docs use AI to read what you're typing, then understand it enough to recommend what to type next with Smart Compose.
Facebook uses AI to detect who is in your photos, then recommends who to tag.
Self-driving cars use AI to detect obstacles and drive (hopefully) in a safe and effective way.
Siri on your iPhone uses AI to understand your voice commands and create responses that make sense.
Smart home technology, like Ring cameras and Nest thermostats, use AI to sense changes in their observable environments, then take action based on what they sense.
Some AI technologies you might hear about are: machine learning, computer vision, natural language generation, natural language processing, deep learning, neural networks, and speech recognition. There are dozens of others, too.
AI technologies are transforming industries from finance to healthcare to retail. In those industries, AI tools are dramatically transforming how work is done, providing unprecedented revenue opportunities, and significantly cutting costs.
That's because AI technologies have a couple advantages over traditional software.
How accurate is AI, and how does AI improve?
Unlike traditional software, AI has the ability to process huge datasets at scale.
Traditional software certainly has access to large amounts of data (think: all the contacts in your CRM system). The software offers clarity to a marketer, because you now can see all your data in one place and perform tasks more easily. But it doesn't offer any context about the data. Traditional software won't tell you want to do with the data or what it means. It's "dumb."
Artificial intelligence technologies, however, are "smart." They analyze the data at scale, then make predictions about what that data means, after learning from training data.
An AI-powered CRM system, for instance, would contain all the same data as a traditional one. Except, the AI-powered system could also potentially recommend which leads are most likely to close, who you should talk to next, and how to score leads based on their behaviors on your site.
However, just because a system uses AI doesn't mean it's necessarily accurate. AI is only accurate if the data it uses it appropriate for the use case and if the system itself is built to make use of that data in a useful way—at scale. In this way, AI outputs are only as good as the inputs into the system.
Some AI-powered systems have the ability to dramatically improve their accuracy over time, either with human training or by training themselves.
Traditional software does exactly what it is programmed to do. Any useful results you get with the software are possible because programmers built the system to produce those results. If you want the system to get better at what it does, you'll need to rely on a software update, where developers manually make the system better.
Some AI-powered systems, however, can improve their performance over time in response to the data they analyze. Sometimes, this happens because humans manually train AI systems on more and more data, so the system has better information from which to make predictions. Other times, the system can actually learn on its own.
Let's take our traditional vs. AI-powered CRM system again.
A traditional CRM system might be programmed to flag any leads that take high-priority actions on your site, like download an ebook or request a consultation. The system might then assign a lead score to the contact based on those actions. Presumably, their score will go up because they've taken some qualified actions on your site, based on rules that you manually created.
An AI-powered CRM system, on the other hand, could possibly take the lead scoring rules you created, then analyze how well it works over time, based on comparing each lead's score to whether or not that lead converts into a customer. Without your involvement, the AI-powered CRM might then automatically adjust lead scores or create new ones based on what it sees working from the data. Maybe downloading an ebook isn't as strong of a lead score signal as you thought, and leads who download one aren't any more likely to convert. An AI-powered CRM might identify that and improve its lead scoring capabilities accordingly.
This is why AI is starting to gain traction in marketing and advertising.
Thanks to the digital marketing revolution, we have tons of data at our disposal from CRM systems, marketing automation software, ad platforms, etc.
But we lack the time, energy, or cognitive capacity to process all of this data effectively, even though it probably holds insights that can dramatically improve our campaigns. Our marketing and advertising performance suffers as a result, costing huge amounts of time and money for brands.
As a result, entrepreneurs and forward-thinking marketing leaders are turning to AI for its ability to increase revenue, reduce costs, and build massive competitive advantage.
And, there are already plenty of use cases and tools for AI in advertising that anyone can understand, pilot, and scale.
Advertising platforms give us tons of data to work with, including measurable impressions, click-through rates, bid levels, demographics, and more.
Humans certainly have the ability to produce good advertising, measure that advertising, and improve ads based on what they learn.
But, digital advertising across search, content, and social media channels, gives us an almost unlimited ability to generate data on what works and what doesn't.
That's what makes advertising at scale tricky (read: impossible) for humans. And it's what makes AI a natural fit for advertising.
With the right data, AI-powered ad tools can detect patterns at scale in your advertising data, then predict what changes to campaigns will improve performance against a specific KPI. This can all happen in seconds, rather than the hours, days, or weeks it might take a human to analyze, test, and iterate across campaigns.
Advertising costs a ton of money, especially if you're selling a product or service that doesn't produce an immediate return.
AI for advertising has the ability to increase your return on ad spend (revenue) and reduce the amount of money you spend on staff time and ineffective ad budget.
But, AI can actually go one step further.
In one high profile example we covered, an AI-powered advertising system actually helped a brand discover and convert new customers they didn't even know existed.
Entrepreneur Naomi Simson, a host on Shark Tank Australia, owns a company called RedBalloon, which sells gifts and experiences online (think: an experience-focused Groupon).
She was spending $45,000 per month on ad agencies alone to run digital advertising for the brand. She was paying over $50 to acquire a single customer at the time.
“It was just not sustainable,” she told us in 2018. “We were being held to ransom.”
Desperation drove her to investigate every possibility. She found an AI tool for advertising called Albert, which we've profiled in the past. The tool uses sophisticated AI to analyze ad campaigns, then manage targeting, testing, and budgets.
The tool was able to do things humans couldn't. In one day alone, it tested 6,500 variations of a Google text ad and learned from the experiment.
Over time, the tool was so effective at learning from data to improve performance that it skyrocketed RedBalloon's return on ad spend. The company was averaging a whopping 1,100% return on ad spend using the tool when we spoke with them. They had also cut marketing costs by 25% thanks to improved efficiency, all while improving results.
The tool also identified hungry potential customers that Simson didn't even know she had. The system identified from its experiments and the data generated by them that Australian expats were highly motivated to buy.
Normally, this wouldn't make sense. RedBalloon sells experiences in Australia, not the other countries where these Australians lived.
But, it turns out, the expats were highly motivated to buy gifts and experiences whenever they returned home, either for themselves or themselves.
Also, the system identified people traveling to Australia from other countries as prime customers.
“I found markets in the US and UK of people traveling to Australia that I didn’t even know I had,” Simson told us.
(She was so impressed, she started a company to add to her portfolio that acts as an exclusive distributor of the technology in Australia.)
The story sums up the promise of AI for advertising in a nutshell:
- Increase revenue by analyzing and acting on data at scale.
- Reduce costs by acting on that data faster and automatically.
- Build a massive competitive advantage with both superior insights and superior speed.
AI is critical to the infrastructure that underlies advertising products on many platforms, though you may not always see it. Modern programmatic platforms often use AI to manage real-time ad buying, selling, and placement.
Digital advertising exchanges and platforms all use artificial intelligence to regulate the purchase and sale of advertising in real-time. That includes programmatic exchanges, third-party networks, and advertising on platforms like Facebook, Instagram, and Snapchat.
You won’t find these exchanges, services, and platforms revealing how their AI works anytime soon. But that’s the point: Even behind the scenes, artificial intelligence dictates how your ad spend gets used, who sees your ads, and how effective your overall campaigns are.
That means if you run paid advertising, you need to understand the terminology around artificial intelligence and ask the right questions about how the AI used by ad platforms may be affecting your spend.
A very basic example of this is:
Facebook advertising, specifically ad frequency and relevance score. These two numbers are key pieces of data that Facebook’s algorithms use—without human involvement—to dictate how much you pay and how your ads are displayed.
You might think showing your ad more frequently is good. But it’s not.
As Social Media Examiner puts it:
Traditional advertising research has shown that optimal ad frequency is at least three exposures within a brand purchase cycle. Traditional advertising schools say that you need to “hit” your audience with the same ad as many times as possible. However, repeat exposure on Facebook might actually hurt your campaign.
That’s because Facebook’s algorithms take into account user feedback. If you show your ad too often, and it’s rated poorly by users, your relevance score may go down. “In most cases,” says Social Media Examiner, “the higher the frequency, the lower the relevance score.”
A high relevance score means your ad is more likely to be shown to a target audience than the other ads you’re competing with. That translates into better performance and lower costs.
Optimize budget and targeting
Performance optimization is one of the key use cases for AI in advertising. Machine learning algorithms are used by commercially available solutions to analyze how your ads perform across specific platforms, then offer recommendations on how to improve performance.
In some cases, these platforms may use AI to intelligently automate actions that you know you should be taking based on best practices, saving you significant time. In other cases, they may highlight performance issues you didn’t even know you had.
In the most advanced cases, AI can automatically manage ad performance and spend optimization, making decisions entirely on its own about how best to reach your advertising KPIs and recommending a fully optimized budget.
In another case, there exists at least one platform that allocates ad dollars automatically across all channels and audiences, so human beings can focus on higher-value strategic tasks, rather than manual guesswork about what works and what doesn’t.
Your ad targeting matters just as much as, if not more than, your ad copy and creative. Thanks to platforms like Facebook, LinkedIn, Amazon, and Google, you have a seriously robust set of consumer data with which to target audiences, both through desktop and mobile advertising. But manually doing so isn’t always efficient.
AI can help here. We know of at least one AI system that looks at your past audiences and ad performance, weighs this against your KPIs and real-time performance data coming in, then identifies new audiences likely to buy from you.
As mentioned, Albert is a player in the AI-powered advertising space. The company’s AI platform analyzes data across your ad accounts and customer databases, then uses sophisticated machine learning to target, run, and optimize your ad campaign.
Another AI-powered tool is GumGum, which uses computer vision technology to learn from images and videos across the web, then help you place ads in the exact spots consumers will see them.
A third vendor to explore here is WordStream. WordStream's AI platform analyzes your advertising campaigns across Facebook and Google Ads, then helps you quickly make changes to campaigns.
OneScreen also uses AI for ad delivery, targeting, and measurement. The company's machine learning algorithm automatically optimizes which content and ads get shown to audiences, taking the guesswork out of advertising.
Ad creation and management
AI-powered systems exist that will actually partially or fully create ads for you, based on what works best for your goals. This functionality is already present in some of the social media ad platforms, which use some intelligent automation to suggest ads you should run based on the links you’re promoting.
But it also exists in some third-party tools, which actually use smart algorithms to write ad copy for you. These systems leverage natural language processing (NLP) and natural language generation (NLG), two AI-powered technologies, to write ad copy that performs as well or better than human-written copy—in a fraction of the time and at scale.
For instance, tools like Phrasee use AI to tackle ad creation. One of the tool’s main capabilities is that it automatically writes email subject lines better than humans—but that same AI-powered functionality has now been adapted to automatically write Facebook ads and push notifications.
The platform’s AI is tailored to your brand’s marketing language and tone, so it sounds exactly like what your human copywriters would write. Then, it produces a huge number of ad copy variations at scale, thanks to sophisticated algorithms designed to increase clicks and engagements.
The result may sound like it puts ad copywriters out of business, but it actually frees up copywriters to work on longer-form and higher-value projects—and drive even better performance when promoting copy and content—instead of spending weeks working to get ad creative created, approved, and launched.
Another tool, Pattern89, uses AI to predict winning Facebook and Instagram ad creative, even before you launch a campaign.
The company’s AI-powered solution uses proprietary algorithms that have access to an unprecedented amount of data across Facebook and Instagram. Pattern89 analyzes this data daily, in addition to analyzing over 2,900 dimensions of platform users’ ads, to predict winning creative, including imagery, copy parameters, colors, targeting and more.
You know what will work for your campaigns, before a dollar is spent running them.
Another vendor, Pathmatics, uses AI to bring transparency and insight to advertising. The tool shows you exactly how your ads perform across channels and gives you competitive intelligence about how your competitors' ads perform, fueling ideas for effective creative and placement.
Beam.City helps clients affordably and quickly acquire customers with AI-powered ads on millions of sites. The company uses AI to do automatic ads setup, monitoring, learning, and optimization pattern recognition. They also use AI to create predictive datasets for targeting personas based on where they live and work in Canada and the US.
How to Get Started with AI in Advertising and Marketing
To get started with AI in advertising and digital marketing, you're going to first want to fully understand the opportunity presented by the technology.
Reading this article is a good start. The next step is to go deeper and download our Beginner's Guide to AI in Marketing.
The guide offers 100+ resources on how marketers can understand and apply AI.
After going through the guide, you'll be prepared to begin evaluating use cases for AI-powered technology in your organization.
What are you waiting for? AI can be a competitive advantage in your company and career—and the Beginner's Guide to AI in Marketing will save you weeks and months as you get up to speed.
Click below to get it for free now.
About Mike Kaput
Mike Kaput is the Director of Marketing AI Institute and a senior consultant at PR 20/20. He writes and speaks about how marketers can understand, adopt, and pilot artificial intelligence to increase revenue and reduce costs. Full bio.