Artificial intelligence (AI) is a form of machine learning that has gained exponential traction in recent years. It refers to intelligent expression by machines to allow cognitive functions, such as learning, in a similar way natural intelligence expresses them.
Every Will Smith or sci-fi fan out there has probably seen movies like iRobot or Terminator (go see them if you haven’t already) where artificial intelligence gets so advanced that it develops sentience and decides humans are a problem that needs to be dealt with.
The AI I’m going to talk about in this article isn’t quite at that stage yet. In fact, it’s not even close.
In marketing, AI is mainly used to collect data, learn from it, and eventually give predictions or automation to tasks based on available data. Most “AI” is actually just, or is very similar to, an algorithm.
I’m going to explore some of the potential uses AI could present for the marketing sector. For some of these sections, I will give an example of something that AI can already do, and give possible applications of how marketers might be able to leverage these factors.
When it comes to automation and machines, the ugly truth is that machines will probably replace most repetitive tasks at some point in the probably not so distant future.
We have already seen this happen in the agriculture sector. From 1991 to 2017 the amount of population percentage that worked in the agriculture sector dropped from around 42% to about 29%.
With advancements in AI, our lives will become increasingly frictionless as many of the more menial tasks around us are automated, making them faster, more efficient, and often safer.
Autonomous cars are an example which has received a lot of funding and attention. Chances are you will probably have heard about the various vehicles that companies like Google and Uber are testing at the moment.
Our grand children’s car?
But while these factors are all exciting and innovative, what can AI do, or what will it be able to do, for the everyday marketer?
The first crazy example I fabricated was the generation of music or sounds for adverts. While I admit that adverts probably won’t be created entirely by AI, the creation of adverts could be significantly changed with Machine learning.
The basis of this theory is Google’s NSynth Super. This is an open source instrument backed by an AI that can generate entirely new, never before heard sounds.
Google’s NSynth Super
Similarly, Amazon has recently released Amazon Polly which creates almost human sounding text-to-speech.
We already know that AI can compose music. An algorithm named Emily Howell (created by musician David Cope) has composed an album of classical music.
It is therefore not entirely unimaginable that the only human input, for an advert would be the video, with the sound (including the speech, if you really want to) being generated or enhanced by AI during production.
As mentioned, the purpose of AI is to make predictions and automation from sets of data. This includes learning from the data the AI collected (referred to as machine learning).
Machine learning has made huge strides in recent years, especially due to artificial neural networks and deep learning which take inspiration from the way the human brain works and applying it to the way a machine learns.
The interesting thing about this is that it enables machines to recognize images and patterns.
Google’s AI perceiving images based on what it was told they are vs what they actually are. Also called neural net “dreams“
In a scenario where this field of AI has matured (which might not be that far in the future), this concept can be applied to marketing.
Specifically, in digital marketing when we use targeted ads, this can have extremely interesting applications.
We already have ‘primitive’ algorithms that can detect our interests and show us ads, but usually, these algorithms don’t really understand the motivation behind our searches.
Let’s go through an example:
Say you’re looking at a book you’re interested in on Amazon.
The targeted ad formats we use today can recognize that you’ve expressed interest in this book and will usually bombard you with ads for the next couple of hours, days or for however long it takes for your interests to shift.
Imagine now that the ad could “learn” about your interests.
If it could understand what the limiting factor was, the ad display would be altered accordingly.
So, if you looked at a book that might be quite expensive (like an academic textbook), you might have checked it out because you were interested, but the high price would have driven you away almost immediately.
AI could pick up on that and in turn inhibit displaying ads about this textbook. The applications of AI that could learn about each individual user’s taste, offer vast opportunities in the advertising sector.
This concept focuses on an AI being able to predict behaviors of customers and respond accordingly.
In fact, predictive marketing is already an active sector and already uses AI to predict consumer behavior.
Traditionally, predictive marketing was done through tedious extraction, transformation, and loading (ETL) from various sources into one database. This data was then used to create consumer profile from which predictions could eventually be drawn from.
This process is tedious, to say the least.
ETL in a nutshell
The advantage of this is that it allows businesses to plan their products/services according to what they already know about their customers, making them much more flexible since it gives them an extended period in which to adjust their strategy.
This aspect is tied in with Relevance (section above), and being able to determine current customer trends and being able to project new ones.
In the future, we can expect the predictive tools, especially the targeted advertisement technologies to grow with the advancement of AI.
The fact that setting up predictive marketing revolves solely around compiling data means it is the perfect candidate for a task that can be taken over by an AI that will automate it for you.
Especially when marketers need to determine which leads are likely to generate the highest ROI, AI can calculate this, based on quality and quantity of metrics, to save businesses vast amounts of time.
Banks already do something similar when determining if a customer is eligible for a loan for instance.
Chatbots have been around for a while now. For the consumer, the most common place is probably Facebook Messenger. In a previous article, I outlined exactly what Chatbots are and how businesses use them, so I won’t go into detail in this article.
However, one area about Chatbots that I do want to focus on is the concept of the conversational user interface (CUI, not really an official abbreviation, but I will use it in this article to make it easier on the eyes) that they offer.
CUI is closely interlinked with the idea of intuition, which many businesses use as a buzzword to sell their products. Intuitive products mean that a user should be able to use them based on, well… their intuition.
This means that you shouldn’t need an instruction manual or a google search query to find out how to do something.
Think about it this way:
You want to change the font size on your phone, but you don’t remember exactly where in the settings the relevant option is located.
With CUI there are no longer menus to click through. You simply tell the AI “Hey, my eyes hurt. Could you increase the font size a little?”
The AI will do it for you.
In some devices, you can already do this (think of Siri for the iPhone) but there are a lot of limitations to do with this, namely that Siri is not actually an AI and cannot learn from the way you speak, your habits etc.
Chatbots used on platforms like Slack and Messenger can learn from their users. With advances in AI and the advancement of Chatbots to become even more sophisticated will no doubt expand the possibilities that marketers can explore.
This is especially exciting when exploring territories that Chatbots haven’t really been deployed in before.
Navigating a website, for example, will become much more intuitive if you can just ask a Chabot. Dharmesh Shah explained this very well in his presentation.
Additionally, a business could potentially find out a lot more about their consumers if they connect various social platforms to their Chatbot, making adverts both more relevant and more predictive.
This idea also extends to the concept of Dark Social.
To clarify, Dark Social stands for parts of social media that are not publicly accessible, such as the content we share with our peers on social messaging platforms such as WhatsApp, Messenger (and WeChat).
There is a trend that consumers are increasingly opting to abandon the ‘public’ domain of social media platforms in favor of the privacy of social messaging.
For marketers, this is a problem since there are no ways to collect the information consumers are sharing via social messaging.
AI could provide a solution for this.
IBM has developed an AI integrated platform called ‘Watson’. The vision behind this platform is to implement CUI for the whole customer journey and analyse the consumer along the journey, all completely anonymised.
The only information the consumer has to give is purely contextual, i.e. his/her name, address, and payment information if he/she wants to buy something.
But Watson doesn’t need anything else since it can extract the other relevant information whilst the user is on the platform.
Being an AI, Watson can also detect abnormal behavior along the customer journey e.g. a customer being in check out for five minutes trying to get a voucher code to work etc.
Applications of AI like this, that allow data to be collected in an anonymised way, might, eventually, prove to be more useful than traditional metrics (especially with the recent crisis Facebook is facing), and is something marketers might want to keep an eye on.
I want to mention search since it is an important aspect of digital marketing (where would we be without PPC?).
Google uses an AI called RankBrain, as part of their algorithm (called Hummingbird) to help process the vast amounts of data collected.
Google also uses various other forms of AI, such as neural processing that can be used for things like image recognition and analysis.
The constant evolution of algorithms (and AI) that companies like Google and Facebook use to determine how websites and posts (respectively) rank on their platforms will likely continue to keep marketers busy for the foreseeable future, much in the same way they do now.
AI applied to search will likely get more sophisticated to understand our queries better, but could also advance data collection for use, similarly to how we can gain a CUI from Chatbots, intelligent search engines could compile and display data for us based on the queries we enter. But this is now pure speculation.
I found it only appropriate to mention some of the limitations of AI, considering I spend most of this article talking about possible uses of AI.
As mentioned, for many people increasing automation could mean that they will be out of a job.
But AI, at least in the way that it can be predicted to develop, will be limited to data-centric tasks.
This means work involving creativity and active content creation (like writing a blog) will still have to be done by humans.
Some reports such as the scores for a football game or superficial financial reports could be done by AI, however, anything that wants an opinion/a specific perspective, requires critical thinking, or is based in a creative background will still be done best by humans.