Incorporating AI into Marketing (and 6 Standout Tools)
At their core, sales and marketing disciplines are based on understanding the psychology of why people spend money.
This behavior doesn’t always seem easily predictable, and although there are patterns in people’s spending behavior, many purchase decisions often change from day-to-day, seemingly sporadically.
However, by applying artificial intelligence technology to the massive amounts of available consumer data, marketers have new and ever-improving tools in their advertising and sales arsenals.
How More Data Makes “Artificial” Feel Personalized
Although there’s still a ways to go before personal home assistants are buying products at someone’s beck and call (Amazon’s dream come true), the decision to easily make a purchase is a few taps away on your mobile device.
The ease and connectivity of a streamlined consumer environment isn’t without its downsides. Friction is now developing from information overload when making a purchase, as well as shorter attention spans that distract us.
This is where AI tools like natural language processing, machine learning, and deep learning are making a difference, which is impacting where our attention, and our money, goes.
Before diving into AI applications for marketing, it’s important to provide some background on the subsets of AI technology that are driving the current “cognitive revolution.”
Algorithms process and analyze data, and build models that will predict or improve performance for that data set in the future.
Within machine learning, there are three categories:
- Supervised learning — which aims to optimize to a specific outcome
- Unsupervised learning — which aims to identify normal behavior
- Reinforcement learning — which aims to optimize some reward scenario
Natural Language Processing (NLP)
NLP is the ability for a computer (or other machine) to understand human language. This is most commonly seen with a person dictating commands to a computer (or mobile device) and having information fed back to them.
That is person-to-machine NLP, but it can also work the other way, going from machine-to-person.
This is accomplished by the machine analyzing large amounts of related text to identify similarities in available information and provide a summarized, easy-to-understand response to human queries.
Within the human brain, there are billions of neurons through which information flows to make decisions and carry out our everyday lives. For example, to identify a particular type of fruit, some groups of neurons might identify what color it is, while other groups would analyze the shape.
In the context of a machine, there is a lower number of neurons than you might find in the human brain, and each “neuron” is responsible for a particular decision or action. Like in your brain, these decisions pass through one another in what is called a neural network.
Sometimes used synonymously alongside machine learning in marketing material, deep learning is simply the concept of using more complex neural networks.
Instead of having a simple neural network with a few “neurons,” a deep neural network (or DNN) will contain many layers of neurons.
Marketing professionals pride themselves on staying ahead of the curve, whether it be anticipating trends, understanding buying behaviors, or keeping up with the evolving nature of communication.
As available information continues to grow, consumers have come to expect a more personalized buying experience where a basic understanding of their needs is already known and relevant content is served to them, all within a timely manner. So how can something with “artificial” in its name deliver a more natural, customized experience to users?
Learning Your Behaviors — The End of A/B Testing
As more actions and signals are monitored during the purchasing process, and the sharing of this information is becoming increasingly accessible, machine learning is able to utilize this data and unleash its potential as marketing intelligence.
Being able to better understand the websites that potential customers visit, their behavior on those websites, customer actions on social media, purchases they’ve recently made, emails they’ve engaged with, and so on, help marketers more broadly understand what improvements should be made.
Recent advancements are allowing marketers to anonymously connect the data from these platforms so as to better market to you specifically, either through your email, cookies, or possibly IP address.
Many online users have Facebook connected to a variety of services, thanks to single sign-on and verification tools. For example, music streaming services like Spotify or Pandora are able to suggest music based on your likes, and Facebook can use that information to promote events and music.
If I am a music agency trying to promote my artists or a venue hoping to sell out upcoming shows, I can look to these services and see what your friends follow to determine advertising targets.
Even simple recommendation engines that suggest what you should listen to next or “people who bought this also bought” recommendations are examples of machine learning algorithms.
When a suggested song or movie gets downvoted or when an item is bought after being recommended, the machine learning model is reinforced and further optimized.
With more data and information available to analyze, marketers can look at models generated from machine learning to make more educated decisions and create a more contextual buying experience for consumers.
True Buyer Personas
When someone begins looking for a particular product or service, or identifies a problem that they are trying to solve, there is often a particular process they follow to assess their options.
Each individual has their own way of conducting this search. For years, marketers have been analyzing purchase behaviors with questions like, “What was the source that led someone to our product? What was their behavior once they went on the site? What engagement tactics were effective to advance someone through a sales cycle?”
With this information, marketers would build profiles around a few select samples of successful conversions. These profiles were then turned into “buyer personas,” and the next time someone turned up with a matching title, position, company type, department, or other such criteria, marketers would try to replicate strategies that had been successful with that persona in the past.
Although this thinking has worked for the past few years, the reality is that even though someone might be an engineer, that doesn’t mean their personal system to evaluate products is the same as other engineers, or that two medium-size companies in similar industries have a similar buying process.
Thanks to the scale of processing power and cognitive computing, more data can be analyzed in a more isolated environment and then continually optimized to better understand potentially interested buyers.
Individuals in different roles from different types of companies might do their preliminary research on similar content sites, have active Twitter profiles, and look for company information outside of the vendor companies’ actual websites.
With machine learning, these variables can be identified and analyzed to market to these potential customers directly on Twitter, and engagement efforts can be focused on leading them to third-party sites.
The focus on inbound marketing over the past five years has led to significant improvements in the quality of content being produced. Marketers are now creating content with consumers in mind to help increase awareness and provide educational information on topics that might lead a buyer to a specific line of products.
Now, with the ability to track a user’s behavior and target similar buyers, marketers can push this content to outbound channels based on their interests.
This is mainly done through marketing automation, where a particular trigger like viewing a page on a website or clicking a link in an email would lead to what content is served next.
Machine learning can optimize for the best-performing content in real time and alter which content is served based on how well that content is performing, thus augmenting a marketer’s ability to analyze their content and its performance.
When it comes to producing content, there are several natural language processing and machine learning tools to help marketers identify the best topics and keywords to write about.
These tools can analyze content to provide predictions on how successful it might be, and make suggestions to improve the content. In the not-so-distant future, AI-based writing products will be able to start writing content based on topical inputs and data from previous posts.
In the case of social media, machine learning is able to build a model based on engagement from past posts and analyze content from related popular topics to make recommendations for post topics.
The Washington Post (owned by Amazon) is already using this natural language technology to analyze financial reports and then write articles, which are published without any further human review.
Improving Speed of Interaction
Improvements in technology have increased expectations from customers, whether consciously or subconsciously. This is particularly true in regards to time, or more appropriately, customer patience and attention span.
New mediums for customers to connect with companies have created a huge challenge for marketers to keep up with. This has led to the recent rise in “bots” that are able to intelligently alert you to important customer engagements and even go as far as responding on behalf of the company so as not to keep customers waiting.
Despite the challenge of keeping up with all of these new outlets, marketers want to encourage customer engagement. By introducing additional lines of communication, marketers put themselves in a better position to capture customer interest and/or create a delightful service experience.
More companies are adding a “live chat” feature to their sites. Even Facebook now offers business users the ability to communicate with customers through Messenger.
In most cases, these communications begin with a chatbot. This allows the company to quickly respond to customer questions or provide timely updates on the status of an order.
When combining these technologies with the other enhancements mentioned above, you can imagine a future where relevant content is served based on previous behavior as well as behaviors of similar profiles, all in a manner that will lead to a positive buying experience.
Maybe someone prefers tutorial videos while another would prefer to receive a call to walk them through a process. Marketers will be more capable of determining which customer is a good fit, who might be a problematic customer, and who is just browsing — tire kickers.
This information will then be used to guide the customer’s journey. As more information is tracked throughout the buying process, marketers will be able to create a better buying experience. This will continue to drive the slow rise of customer expectations.
Data is driving advancements in our society, not just in marketing, but across all disciplines. If you can measure something (or even define it), you can optimize it.
Marketers aren’t interested in selling data they collect — at least, not the smart ones. For them, the value is in using the data to learn and deliver better experiences to their customers.
AI gives marketers the ability to identify hidden connections and patterns that might have never been realized to create better content and marketing strategies. It allows consumers to be reached in the manner they most prefer and gives them as much (or as little) information as they might need to make a buying decision.
These improvements are subtle and gradual, but are creating noticeable differences in the purchasing process, largely due to advancements in AI.
As with any department, the most important thing to remember with AI for marketers is to not get stuck in what AI can do, but instead look at the problems you are facing and ask yourself: what can AI do for you?
Cool Marketing and Sales AI Tools
One of my favorite tools, Crystal analyzes a person’s publicly available information and writing style, then runs it against a personality profile to show the best way to communicate with that person.
These tools analyze your database and build a model to identify the characteristics and qualities of your best prospects.
If you’re considering machine learning-based lead scoring, it’s important to be patient. Give the model time to optimize against your ideal customer behaviors.
It also never hurts to have more data.
Digital advertising has changed dramatically over the past couple of years as the different mediums and scale of online advertising space has increased.
This has led to a rise in what is known as programmatic advertising. That’s where machines are buying (and selling) ad space instead of humans.
Acuity has a unique algorithm that is able to optimize the sites, mediums and ads that are served based on a desired marketing outcome (such as clicks, content downloads, demo requests, and more).
Although this product is still in early development, it’s the tool I have dreamed of since I discovered natural language processing.
Kylie analyzes the contents of your emails and automatically drafts responses for your most common emails. Their applications are strongest for customer service but soon will automatically draft templates for sales teams.
The way to be most effective in today’s sales environment is to be relevant.
Nudge helps take out the lengthy research process on your prospects by providing you with relevant news stories about individuals and their companies while you’re writing emails.