AI can seem scary and very far away. AI may resonate with the fear of a future dominated by robots. Very often people do not even see where AI is applied in their lives.
Here, we are going to give a better understanding about what AI is but also how it is used for Marketing.
First of all, AI, aka Artificial Intelligence is used by Marketing for, at least, the past 10 years already, mainly via Machine Learning. Very often people mix the two so let’s start by understanding the differences:
Artificial intelligence (AI): it is the broad discipline with the goal of creating intelligent machines emulating and then exceeding the full range of human cognition.
Machine learning (ML): it is a subset of AI that often uses statistical techniques to give machines the ability to "learn" from data without being explicitly given the instructions for how to do so.
Reinforcement learning (RL): it is an area of ML. It uses software agents that learn goal-oriented behavior by trial and error in an environment that provides rewards or penalties in response to the agent’s actions (called a “policy”) towards achieving that goal.
Deep learning (DL): it is an area of ML that attempts to mimic the activity in layers of neurons in the brain to learn how to recognize complex patterns in data. The “deep” in deep learning refers to a large number of layers of neurons in contemporary ML models that help to learn rich representations of data to achieve better performance gains.
So, when we were referring to algorithms and recommendations many years ago, we were already into the AI field. These past years, AI has significantly evolved and is now covering multiple technics to make the machines more intelligent and generate better results.
Most of AI challenges are common across the board.
First, it all comes down to input. No AI is good without the right human input. If we do not feed clean, formatted data, the machines will not be able to provide good results.
Then, we need to give time to the algorithms to learn. Each time the system captures more data, it refines its calculations and renders more accurate results.
Finally, humans need to know how and what to refine. Humans can add rules or filters. If not applied correctly, the results can be very off. On one project, where we came in, the team had put so many rules and filters over the years into the systems that first, they did not know anymore which rules were put in and second, all the rules ended up preventing the system from doing what it was supposed to do. We had to clean up all the rules to the bare minimum and let the machine relearn before stirring it again in the right direction by re-injecting the rules that would help get the expected results. Humans also need to be present to readjust the machine when it is bugged (as it was during the Coronavirus crisis where Facebook was censoring lots of posts without reason).
AI is therefore as smart as the human behind it. Indeed, the threat can be that the machine, wrongly managed, achieves unwanted goals. We should not be afraid of the machine but of the human managing the machine. As for most technologies, we can get the best or the worst out of it.
How can AI be applied to Marketing? Let’s organize the applications in four groups.
Improve the user experience
Improve search thanks to image recognition and language processing: The customer can take a picture of a product he is looking for and will get the results. Instead of text, he will use images. Language is what most experience with the use of Siri or Alexa, where thanks to voice, Alexa finds you what you are looking for.
Chatbots & concierge: there are a lot of different ones in the market. For Shopify, we like Tidio. Chatbots are used to provide automated assistance to customers, without the need of a customer care agent. It helps funnel better through customer service, reduce the wait time and provide immediate assistance. Once they are configured, they roll and can provide very detailed answers.
Analyze user experience and help website optimizations: tools like Dynamic Yield allow to test all areas of a site and define which message or design is the most efficient for a specific customer. You can then define multiple interfaces per audience.
Predict churn and create smart customer engagement: Machine Learning algorithms can help gather data about disengaged customers and apply predictive models to find out which accounts are at high risk of churn.
Improve imagery: AI can assist the design process by making it easier. Sensei helps in this way.
Improve marketing relevancy
Improve social media: this is what Twitter and Facebook are doing in targeting your audience with specific messages.
Tag products to be more relevant: organize and categorize product catalogs based on images, creating faster product descriptions. Catchoom provides this functionality.
Demand forecasting for inventory management: automate replenishment, supply chain optimization with tools such as Remi AI.
Predictive analysis: for inventory, customer behavior, buying patterns, churn...
Market analysis and data mining: define virtual panels to test products. Response AI provides test automation for market research.
Optimization of display ads by creating micro-moments: customers are targeted using native ads.
Create dynamic audiences, customer segments, and find the most profitable audiences: Salesforce can do this in their CRM.
Display targeted offers and content: push content and offers based on customers’ interests.
Optimize all elements of an email-based on customers’ reactions (title, colors, buttons, content…) and sentiment analysis: Tools like Persado are giving stunning results.
Help create marketing assets
UGC curation and syndication: identify UGC content or topic to then use it in one’s platform. Reviews can be syndicated by identifying which reviews about some products got written and then compile them to use them in another platform. This is very useful for large retailers selling different brands and products. Upcontent is a curation tool.
Content creation: create blog articles in seconds, using some keywords which can help for SEO. We tried Articoolo that still needs work to be really efficient, or
Help manage marketing operations
Monitor social media: follow all the conversations in social media in your field of activity.
Improve customer service data: this helps agents be more efficient on the phone (by providing them with customer data to handle conversations or by creating scripts/answers based on customers’ specificities).
Fraud detection: detect suspicious behaviors based on location, IP, products which reduce chargebacks.
Robotics for pick packing in warehouses: the robot goes to the right aisle and location in the warehouse to pick the product ordered.
Automatic PPC budgets adjustments and spent.
Besides the automation of some tasks, the biggest value of AI is the relevancy marketers can bring to their customers. Customers are sinking in an ocean of content and offers. They have access to competitors one click away. Customers are more and more demanding and less and less patient with the brands. They cannot stand disruption and do not want to make an effort to find what they are looking for. If you do not chase them and pop up at the exact instant, when they are receptive to a piece of information that brings them value, then you lost them. They can find everything on the internet but do not know what to look for anymore. They cannot tell the right from the wrong either, so marketers' work is more complicated than ever. Marketers need to be accurate and relevant in order to get the highest ROI and efficiency.
Better decision making, more efficiency, better ROI and more revenues. The equation is simple. Solving the equation is way more complicated.
Personalization has shown its efficiency and ROI. In some experiments we ran, we saw results being 5 times higher when using personalization than not using it. Being relevant pays off. Recommendations can generate more than 30% of the total sales of a site. It became a way to help customers search or put in front of their eyes the products that might interest them. Customers require an ever more seamless experience and an ever less disruptive advertisement so we all make sure to serve them quickly what they need to that they can be in and out in few clicks.
But a hyper-personalization can have some drawbacks. By constantly targeting and refining, you are pretty much reducing the funnel of information you present to the end-user. You might have seen on Facebook people complaining that their feed was reduced to only 25 persons and that they were not seeing posts from their other friends. By reducing the users’ window, you control their thinking. This is how you can start molding customers’ minds about a piece of specific information. They will see this information over and over because it is part of their interests.
This technic does not help open people’s minds. People become formatted to a single line of thinking and do not challenge it any longer because it is the only thing they see during their day. You can create a sort of commercial monopoly using information. As a brand, if you are not part of this pool, you can see how you are missing some opportunity.
Such hyper-personalization also prevents ‘’ discovery’’ and it does make it more difficult for a new entrant. This is why, when we do some hyper-segmentation, we need to add some components for discovery in the algorithm so that the customer does not end up with the same recommendations over and over. You need to find a good ratio between relevancy and discovery and expand from there. Customers might want to learn about other products that they would not have known otherwise. In general, a ratio of 80/20 or 90/10 between relevancy and discovery is necessary.
AI does help marketers be more relevant and get more efficient results. Not leveraging AI in a strategy, whether for a small business or a larger corporation, would be a missed opportunity.