Co-Founder and VP at Fyusion, Pantelis is responsible for their patented 3D viewer technology and engineering a consumer experience that deliberately, “Does not scream ‘artificial intelligence.'” Rather, Fyusion’s intent is to provide their users and consumers with a life-life photographic shopping experience.
With 10 years of research and engineering experience specializing in optimized, fast, scalable architectures, Pantelis’ view on consumer purchasing behaviour is worth listening to – especially for those of us integrating AI into our commerce businesses.
M.R. Rangaswami: In your opinion, what do you see driving online consumer purchasing behavior?
Pantelis Kalogiros: There are 3 I’s we see driving online purchasing behavior right now:
Immersive: Consumers want better online shopping experiences, better in many ways than shopping in person. For instance, car shoppers want to feel as though they are on the lot, and can virtually touch and feel the vehicles. Not just stitched images or stabilized video, shoppers expect a smooth 360-degree interior and exterior experience that loads at lightning speed.
It doesn’t matter if it’s a big or small item anymore, consumers expect the same immersive experience.
Interactive: The ability to swipe, rotate or virtually walk through an environment is great, but to pull people in as deeply as possible, they have to be more stimulated than that. Highlighting key features of a product as people swipe or “walk” through can help.
Ideally, interactions include the ability to circle around an object at numerous or unlimited angles. This means enabling the shopper to control at what angle they are viewing the object—for large objects this includes angles that might be physically difficult in person. It also entails the ability to get close up shots of features and damage, and provide multimedia such as audio, video and additional images to allow the shopper to drill down to learn more.
Intelligence: Shoppers want to be quickly educated about an item without having to scroll down page after page or click away on an image gallery.
Intelligence can also mean better communication with shoppers. For instance, a customer might visit for one item but need help finding an item with all the features they want. Today’s shoppers expect “conversational commerce” where they can easily continue researching with AI chatbots and/or progress to a live conversation via chat, SMS or phone.
In many respects, the online shopping experience is becoming richer and more nuanced than shopping in person.
M.R. Rangaswami: What are some of the AI functionalities you incorporated into your business that have been “game changers”?
Pantelis: As an extremely research-driven organization, AI tends to permeate all of our products—both in the more overt, user-facing sense, and in crucial behind-the-scenes ways.
Take our automotive product, for example. The output itself—shaky cellphone footage transformed into a crisp, seamless 360—is the result of a confluence of cutting-edge AI techniques: 3D reconstruction and SLAM from the world of robotics, image enhancement and correction from the world of computer vision, viewpoint interpolation from the world of computer graphics, and so on.
With this format, we have also rolled out a number of more overtly AI-powered features. Take the Fyuse social network: We created a “More Like This” button which, for every 360 uploaded, would crawl our database to find similar images and present them to a user. Or, back to the world of automotive: for every vehicle uploaded, we are able to automatically break the car down into its composite parts, and enhance it with visual tags.
It’s the behind-the-scenes work, though, that I think has been the most revolutionary; using AI techniques to subtly enhance the user experience without drawing attention to it. For example, every car uploaded to our system is automatically re-oriented to face in the same direction in order to make the shopping experience more consistent. Every 360 is uniquely compressed, using a variety of deep learning, to keep loading speeds down while keeping the product front and center. It’s these sorts of hidden improvements, which may not scream “artificial intelligence” to the general public but provide a truly lifelike experience, that have set us apart.
AI helps us understand what is depicted in each scene, and provides context that goes beyond pixels and color ranges. Having this contextual understanding can help us improve compression and image clarity, and provide solutions where traditional algorithmic or discrete computing methods fail.
M.R.: What AI features do customers/consumers respond well to—and what have you seen flop?
Pantelis: I’ve found that customers respond best to AI or machine learning techniques when they don’t draw attention to themselves; that is, when they don’t feel artificial.
I’ve seen companies fall into this trap: they implement an algorithm that performed amazingly well on a dataset, release it to the public and brand it as “AI.” While these tend to get attention in the short term, they lack a certain stickiness. Once the novelty wears off, a user doesn’t want something futuristic. They want something that feels natural and that will work every time.
In our own product, we’ve found this most clearly in the realm of 3D imaging.
Features which try to add bells and whistles on top—synthetically modifying elements of the scene, or overtly smoothing out defects—tend to not go over well with consumers. The power of an immersive format is that it enables trust; that it’s more honest, in a sense, than a standard picture. Any feature which violates that trust risks losing the customer. We have learned that whatever we build must look as aesthetically real as a 2D photo.
Customers respond extremely well to AI features which scratch a real itch and improve an experience that already existed. Take our visual tagging approach: It provides real information to a discerning shopper, in a way that is quicker to browse (and more contextually grounded) than a wall of text. It is an enhanced version of the real-world experience of physical shopping.