AI & ML in Retail Promotion

Happy people buy more stuff!

Artificial Intelligence (AI), Machine Learning (ML), and Neural Networks (NN) have allowed us to gather and sort data instantly. You may be familiar with the first two technologies, but the last one will give us the ability to emulate human thought.
At this point, Neural Networks cannot possibly have the vast number of interconnections available in the human brain. If our minds didn’t “cheat” by having states other than Off and On, humans would be incapable of thought and intelligence. NNs are not “efficient thinkers” like us, so we use technology to get around their shortcomings.
Instead, they make up for this inefficiency with incredible speed and what is called parallel processing. A very binary question (yes/no) for humans and the sort that we can answer the fastest is still complicated to an AI… Consider:

Is the light on?

Human: Yes
AI:
a. {Query: Sufficient Power to circuit? Yes}
b. {Query: Switch in “on” position? Yes}
c. {Query: Device capable of producing light present? Yes}
d. {Query: Is there measurable light above ambient level? Yes}
Yes, pending further evaluation.
Despite the added layers (and this is an extremely over-simplified example), the AI can still answer faster than a human. Humans can only handle one thought at a time. However, the cluster of associated information surrounding a simple word like “apple” would probably include trees, dappled sunlight, blue skies, cinnamon, pie, baking, orchards, laptops, phones, as well as memories of a kitchen or a place where someone first experienced “apples” (whether computers or fruit), as well as people who were present at the time.
AIs, with the help of NNs, will learn to do this, too. NNs are still rather primitive at the moment, so if you say “apple,” it will pull up everything it can find on apples, from Biblical references to recipes to computer brands to methods of preparing cider and wine. Everything it can find, without discrimination, and then it will sort through that information looking for relationships and ways to associate it with any apple-based inquiries.
The difference is they look at all the information at once, not one item after another, like a human being. The individual wins by “knowing” what is irrelevant and dismissing a lot of useless data. The machine has to look at every possibility, millions of them, but at billions of operations per second, this is much faster than a human.

So, How do we Use This?

This is where it becomes interesting. Unless you’re running a groceteria, fresh fruit market, or a Genius Bar, “apples” may not be high on your list of priorities…but everything else is.
Let’s consider a Supermarket, or other stores with aisles, as an example. These places lend themselves to serving useful information to customers like no other. Everything is organized predictably, with prices and product placement.

Customers armed with a smartphone that recognizes and can respond to RFID (Radio Frequency Identification) information are well ahead in the bargain-hunting game. Equipped with an in-store App, their phone becomes a walking billboard for you to serve data to them.

The AI of such apps can synchronize with customers’ shopping lists (green arrow) to show where products are as they are encountered in the store. They can point out SALE items (yellow arrow) with name and price displayed. They can highlight favorite items (pink heart), and remind customers who use them regularly, such as hair coloring every 4-6 weeks. Discounts on items regularly purchased (blue arrow) can be displayed, even if not on the shopping list this week.
The Produce or Meat departments might include “floating” recipe-suggestions, or preparation techniques that could be downloaded at a touch. ML could, when buying back-ribs, trigger a swipe-dismiss or timed pop-up that says “Do you need BBQ sauce with that?” because that is what many other customers do.

How does it work?

Beacons are battery-powered devices that transmit RFID information over a very short range. Scattered strategically throughout your retail environment, they are like a private GPS with remarkable accuracy. The customer’s smartphone always knows its exact location, and the AI can line up the images and its data, based on what the phone’s camera is seeing.
Beacons are long-lasting, a year or two between button-cell battery changes, or can be hardwired to a low-voltage supply permanently. Each has a unique identity and can be individually programmed, if desired, with product information to broadcast.

The Ultimate Convenience?

It doesn’t have to be a grocery store, of course. It would work as well in an IKEA, or Big Box Electronics store. A customer goes into a retail location where they already have an associated App. They will generally have a generic (or very specific) product they want to find.
Having a type-or-voice-activated search box, questions such as “Where are the electric drills?” would produce an arrow path superimposed on the camera view. It’s merely a matter of following the arrow to get to the desired product.

Most people are comfortable with this level of tracking because it doesn’t work outside of the store. Alternatively, an overhead map of the store, showing a drawn route could be provided, based on customer or retailer preference.

Robot Assistants

Japan, always on the leading edge with these sorts of innovations, has introduced robot-greeters. One model, named Pepper, is capable of reading expressions and understanding human emotions with a high degree of accuracy. You can ask Pepper for product information, or the location of desired items, and “she” will direct you as required.
Pepper speaks (several languages), is made from smooth, white plastic, stands about 4 feet tall on wheels, has two fully articulate hands capable of communicating using International Sign Language, and a tablet-like display built in her chest to interact with customers.
This robot looks like a vacuum cleaner that grew a head and arms, but it is incredibly effective at getting attention. Customers love it. Placement of the robot has foot traffic doubling, even in the North American venues where Pepper has been installed, with a distinct and significant increase in sales.
ChatBots

Retail Direct Clients

Nowadays, clients can connect with a Digital Assistant for a text chat. In some cases, this can be a voice chat, almost indistinguishable from a human. A customer can say “I saw the blue sweater on page 341 of your catalog” or they can say “I was looking at item number 108-1234” and then ask for specifics in colloquial English such as “Can that be ordered in red?” That’s only the beginning…

Commercial Retail Clients

B2B clients can order parts and equipment, too, but with increased levels of efficiency and accuracy. The ChatBot knows precisely what is in stock, knows the delivery schedule, and knows when something can be on the customer’s receiving dock.

Protecting Yourself

Equally important, AI/ML protects your company by knowing if the client has sufficient account credit to place the order, or whether Accounting needs to speak to the client and assure payment. Organization-wide information transfers keep financially over-burdened clients from putting you at risk by ordering product they cannot pay for.

The Uncanny Valley

One thing we must be aware of it is a statistical anomaly called the “uncanny valley.” Acceptance of robots for customer service, or any other purpose, is quite easy provided they don’t look like “almost-people.” The closer they get to human-looking (70%-90%), the higher the sub-conscious fear. At least until they reach a level that looks remarkably human-like, then the fear magically disappears. Everyone loves C-3PO then it falls apart…
The high percentage of elderly patients in Japanese hospitals calls for Aide Robots to prevent injuries to workers. The robots are needed to lift patients from the floor (it is common to sleep on floors in Japanese culture) to help them stand, get into a bed, a bath, or sit in a wheelchair. These powerful machines look very much like a big smiling Panda Bear.
The un-human look is reassuring to dementia patients as well as regular folks. For this reason, until we have very human-looking robots, it is best to use “cute” ones like Pepper (mentioned above).
These service robots can attend airline ticket sales counters, for example, with the advantage that they can read the emotional state of a customer and respond appropriately to anguish, desperation, happiness, and much more. Better yet, whereas a human counter attendant may not know that there is a way to get a passenger to a destination, the AI knows everything about available flights and can find solutions that humans would otherwise miss.

The Takeaway

The Retail Service Experience for customers is going to improve dramatically with the implementation of AI, supplemented with ML and NN. No matter how good your level of service is, everybody knows it can always get better.
The people we currently employ for customer service won’t be replaced. They will have to be retrained (at company expense) to program the AI to be as effective as possible.

It’s just not worth it to hire brand new people that don’t understand your company culture or the needs of your business when you already have those people in your employ. Make it a point to reassure your staff that they are not being replaced, but rather up-skilled and given increased responsibilities.
Resistance to AI is useless because it is inevitable in all aspects of our lives. It’s better to embrace it—so strap in, hang on, and let’s enjoy the ride!