I am excited to share this talk with you here. Its a talk I gave on using Artificial Intelligence i.e. AI in Marketing at CSI, Ahmedabad in 2021. You can watch the video below or read the transcript given ahead!
Imagine you just found out that Priyanka Chopra is coming to your house today and that you are supposed to impress her. What would you do? Would you focus only on what to wear, your hairstyle, etc or will you also think of conversation topics to amaze her? You will need to do both, right?
Don’t you think this is similar to the confusion you face everyday at work? About how to impress your valuable customers? Should you use fancy technology and AI or should you focus on engaging and witty marketing and communication?
That confusion is what I hope to clear up today. I will cover 5 ways to woo the customer by combining AI and marketing in the right way.
AI In Marketing Gone Wrong!
In 2012, famous retail brand Target did a shocking thing. It correctly sent maternity coupons to a teenage customer who was secretly pregnant. She had not even told her father of her pregnancy. So how did Target know that she was?
Through Artificial Intelligence!
By analyzing purchase habits of all their women customers, their AI had managed to predict pregnancy, almost accurately.
When this story came out, customers were really angry with Target for digging into such personal health information.
Target was trying to increase long term sales. They wanted to hook the customer right before she became a new parent so that she would keep coming back even after she delivered. But it backfired because they didn’t know where to draw the line.
This is a very tricky balance to maintain- between AI and marketing- how much to use and how to use without angering the customer?
Before I answer these questions, let’s first understand what we mean by AI.
What is Artificial Intelligence i.e. AI?
Most of us think AI means self-driving cars or robots or Iron Man’s Jarvis.
While those too are powered by artificial intelligence, AI is a much larger umbrella term.
AI has 3 core tools: Statistics, Semantics and Logic. Using these 3 tools in different combinations and proportions, we can make machines smarter. So Machine Learning, Deep Learning, Natural Language Processing, etc are just the different AI methods used to make machines smarter.
As you know, the most common and popular of these AI methods is Machine Learning.
What is Machine Learning?
To understand what Machine Learning is, let’s suppose you want to buy some clothes and you decide to visit Myntra for that.
If you login to Myntra for the 1st time ever, you have to set filters according to your preferences because Myntra does not know you. It has zero data about you.
As you start buying a lot of clothes, it starts keeping a track of what kind of clothes you are adding to the wishlist, your frequently used filters, what kind of clothes you normally end up buying, whether you mainly shop before Diwali or before Raksha Bandhan and so much more.
Now when you login to Myntra for the 100th time, it will immediately show you the kind of clothes it knows you would like for sure, suggest the right size for you, send special coupons just before Raksha Bandhan because that’s when you are most active.
This is machine learning. Its main purpose is prediction.
The Myntra algorithm collects data during your past and current purchases, analyzes it and then predicts what you will want in future. And its accuracy on your 100th visit will be far better than its accuracy on your 3rd visit because it is continuously learning on its own.
Steps to start using AI in Marketing
Step 1: Get a marketer to train the AI
If you want to use AI in marketing, then you must have a marketer train it (using training data from marketing), not an IT person. Because if you want to train a junior cricketer, will you ask a footballer to mentor them or will you ask a senior cricketer to do that job?
Step 2: Define the right hypothesis
All of us who wish to bring in AI to help with business, need to start with the right question or hypothesis we want the machine to find answers to. That is the very first and most critical step.
Step 3: Find the right set of data
After we ask the question, we need to find the right data to get quality output. Basically, if we ask the wrong question or feed garbage data, we will get rubbish answers.
Where Target went wrong
This was what caused so much trouble for Target.
Their first question to their statistician was, “If we want to figure out if a customer is pregnant, even if she didn’t want us to know, can you do that?”
The question itself was morally wrong. No wonder the output was also morally wrong.
Responsibility of Marketers and Business Owners
As marketers and business owners, we must realize that no one knows the customer as well as us. It is our job to understand customer behavior, needs, expectations. Hence, the responsibility to find customer insights without violating their privacy lies with us.
AI is only there to help us do our job smartly and quickly. It will only follow our lead.
So if we want to ensure that our use of AI in marketing wins over customers hearts and not their anger, what is the right way to approach it?
5 ways to use AI in Marketing correctly
-
Customers First. Then Marketing. Then AI
Remember when Uber offered free taxi rides to and from vaccination centers for eligible people?
This is the kind of marketing and AI integration that wins the hearts of customers. Because, it lives the primary principle of marketing i.e. ‘Customer comes first. Always’.
Uber did not create the technology first and then think ‘what shall we do with this technology?’
They first started with the question ‘How can we make our customer’s life easier in these crucial Covid times’?
This is marketing at its core. They created the entire marketing strategy in terms of what to offer, to whom, at what price, where. Then figured out how to promote it.
Only after this, they brought in AI to help execute these strategies.
AI helped them identify eligible customers and offered a free ride only to those eligible customers and that too only if they chose a vaccine center as the destination. This level of granularity and this speed would not have been possible without AI.
This Uber experience made their customers happy and if customers are happy, business will be happy.
Focus on creating positive, memorable customer experience.
Products or services that are loved by customers today, become outdated and forgotten tomorrow. Only the experience we offer, stays with the customer and thus becomes our differentiator.
And the only way for us to give a memorable experience to every single customer, meeting their every need and expectation, is through AI.
Uber started with the question- ‘What will give my customers a memorable experience?’ And built their AI Marketing solution on that base.
Whereas Target started with ‘What will make my company more revenue and for a long time?’ Customers did not feature in their equation at all. Is it any surprise they failed to make the customer happy?
Bottomline? To woo customers, first build a customer centric marketing foundation and then bring in AI later, to augment and support it.
-
Personalize the experiences
I love this quote from Brian Solis “Welcome to a new era of AI marketing, where machines help make marketing more personal and human.”
What can be more personal than designing your shoes such that you are the only one with shoes of that design?
Nike Makers Experience
Source: Nike
In 2017, Nike launched a system called Nike Makers Experience that allowed customers to design their own sneakers in-store.
Customers had to put on blank Nike sneakers and choose their own graphics and colors. Using augmented reality and projection systems, the system then displayed the design on the blank shoes. The designs could be printed on the sneakers and be available to the customer in about 90 minutes.
This customer engaging feature not only drove sales but more importantly, it allowed Nike to collect data about customer preferences.
Nike then used this data with machine learning algorithms to design future products and deliver personalized product recommendations and marketing messages.
Who wouldn’t be excited by this speed and level of personalization? And it was all possible because of the right use of AI.
Personalization makes the customers feel special.
Just ask my husband. When Swiggy was launched, he was an early adopter. We started ordering from his Swiggy account.
Although I too have a Swiggy account, because of his longer history of orders, Swiggy’s AI knows him much better than it knows me. It knows what type of cuisine he likes, his favorite foods, restaurants, etc. That’s why it sends him marketing offers and coupons that are far more tailored and far more frequent than it sends me.
He keeps getting superb offers absolutely suited to his tastes and hence we keep ordering from his account, not mine. Since we order more from his account, he gets even more superb offers. And the cycle continues; not letting Swiggy know me well enough and my preferences.
A brand that makes its customers feel special and showers them with attention, earns a rightful place in their hearts and wallets.
But personalization at the cost of privacy can be disastrous.
-
Privacy. Privacy. Privacy.
Like Spiderman said “With great power, comes great responsibility”.
There is a very, very thin line between collecting enough data to personalize communication and continuing to collect beyond that.
The most difficult part about using AI’s power in marketing is to know when to stop. If we cross that line, we may become a brand people hate.
Try this question as a litmus test of sorts. “Will using AI like this create a positive experience for my customer?” If the answer is yes, go ahead. If no, can it.
Suppose you have just created an app that tells you the ingredients in a dish simply by clicking on an image of it.
Now obviously, you will have to get access to the user’s phone camera and gallery to identify that photo. But you don’t need to access to their calendar or their contact list.
Or suppose you have created an app that syncs with the smart lights in my home.
When I switch on the TV, the light turns soft and contains more red because that is good for my eyes. When I sit to read a book, through the webcam of the room, your app can see me reading, so it automatically turns the lights bright and white.
Now to do all this, your app will have to get access to my webcam feed and other smart appliances. That is someone else’s sensitive information in your hands.
You need to ensure right from the time you start building the app, that you collect only as much data as absolutely needed to give the customer a delightful experience, not one iota more. And also that the data is kept completely safe and not sold to other interested parties.
That is respectful of customer privacy.
Apple and Privacy
Apple does this very well. That is one of the reasons Apple continues to be a beloved brand of customers simply by making customer privacy their first priority.
Its latest iOS update forces apps to ask for permission upfront so that customers have total control over their information- who can use and how much of it can they use.
-
Oversee and course correct
Teaching and training AI is not a one time thing.
Of course, when we are just creating AI, we have to spend far more time and effort with it. And as it starts learning, we can go back to our strategy and ideating tasks.
But it is still our responsibility to supervise the AI algorithms, to know that only right, healthy data is being fed. Otherwise, garbage in. Garbage out.
Because, it’s a machine after all. It will not have the moral compass to know right from wrong. If we leave AI unsupervised, the consequences can be immense.
What we mean by training and supervising AI?
Before we get to the consequences, let me first explain what I mean by “Training” and “supervising” AI and machine “learning”.
To understand that, we need to first understand how we humans make decisions.
We have a situation, we get Input Data, marry this Input Data with our Experience to arrive at a Decision.
Example:
Situation: What outfit to wear for the day
Input Data: Is there some occasion today or what is the weather like, reason for stepping out today, place of meeting, etc
Experience: When we meet our friends, we are more comfortable in casual clothes. But we have also learnt that if it is rainy outside, we better not wear jeans but some other casual clothes that will dry quickly.
Input Data X Experience = Decision (Wear a knee length cotton dress)
This is also the same way machines learn or arrive at decisions.
They are fed Input Data by us. But since machines do not have any experience of their own, we have to feed them a data set called Training Data i.e. sample data that will tell them that ‘if this happens, then that will be the outcome‘.
If good, accurate Training Data is fed along with correct Input Data, machines will learn quickly, accurately and start arriving at right Decisions.
Now, let’s go back to the consequences of not overseeing and supervising the Training and Training Data
An infamous example is Microsoft’s Artificial Intelligence chatbot ‘Tay’, who was modelled to speak ‘like a teenage girl’.
Tay was supposed to become more and more intelligent and human-like by learning from the conversations she was having with real humans on Twitter.
We know how people love trolling others online. Interacting with such trolls and nasty people, sweet and innocent Tay went from this …
To this…
In less than 24 hours…
I think you can see for yourself the issue here. She was deactivated. Obviously.
Let me share another reason why supervision and course correction is needed for AI and why it should not be left alone to do whatever it wants to.
Google’s blunder
Sometime back, Google Photos tagged the photo of a girl’s African friend as ‘Gorilla’. And this created outrage.
That is why overseeing the training of the Machine Learning algorithm or AI is so critical. But you don’t have to do it all by yourself.
You can get training help from users too.
See how Google uses our help to train its Gmail Algorithm how to identify what is good and what is bad. This is how it does this ‘supervised learning’ everyday.
Supervising AI and guiding it in the right direction are good but not enough.
-
Use marketing and AI to give back to society
The ultimate aim of any business should be to use AI and marketing to give back to society.
Marketing is all about communicating in a way that influences habits and perceptions. If used correctly and for the right reasons, it has the power to influence the habits of 7 billion people.
In 2020, every brand, business, government communicated the importance of hand washing. And almost overnight, 7 billion people started doing that multiple times a day.
That is the power of marketing.
AI is super human and insanely powerful and quick.
Combine the two and you have got a power packed option to do some good for society.
How a brand used an app and marketing insight to help scores of Indians book vaccination slots
In the first week of May 2021, 18+ adults in India were playing a mad game of fastest finger first by trying to book a vaccination slot. Slots were getting filled within 1 minute of them opening.
In such a scenario, I received a Whatsapp (same one across multiple groups) asking me to register on a website called vaccinateme.in. That website would send me an instant notification every time a slot opened up.
I, like thousands of others, registered, mentally blessing whoever had created such a convenient app.
Only when I received a welcome message, did I realize that this act of service was done by the pharma app HealthifyMe.
It used its marketing skills to figure out the biggest health related pain point of its customers at the moment. It decided to lend a helping hand, not expecting anything in return. To bring this social wish to life, it turned to AI to create a solution and help millions of young Indians secure their health.
Cadbury too amplifies marketing insight with AI for the greater good
During the 2020 lockdown, Cadbury realized that local, neighborhood stores did not have enough visibility.
So it manually collected, sorted and mapped data of stores against pincodes and then used AI to figure out the pincode of the viewer and serve an ad to that viewer containing names of neighborhood stores from that area.
This massive personalized marketing and AI exercise, made not only the customers happy but also the retail business partners.
And as a quick recap:
So the way to a customer’s wallet is through their heart.
And the way to their heart is through safe, memorable, personalized experiences.
And the way to those experiences is through AI in Marketing.
And the way to AI is through data.
So, start collecting data from this minute onwards so that going forward, you can bring in AI to help you grow.
Thank you!