A morning like any other. A customer pops into a coffee shop and is welcomed by the smell of freshly baked goods. Today, there’s an extra hint of cinnamon. “The usual?” asks the woman behind the counter. A quick nod, and a muffin and coffee exchange hands. But the smell in the air has created a new craving. The woman follows her patron’s searching glance and offers a cinnamon bun with raisins, asking, “They’re your favourite, aren’t they?” She’s right. It’s the perfect customer experience and yet another person leaves her shop happy. Predictive analytics helped detect a cross-selling opportunity and convert it into a sale. Although this example is nothing more than a standard transaction at a coffee shop, it still represents a gigantic challenge for artificial intelligence.
In order to present customers with offers that meet a genuine need, Bechtle ist working on using artificial intelligence. To do so, it is currently pursuing two avenues of development. The first is a recommendation engine which analyses a customer’s behaviour on Bechtle’s website in order to suggest suitable products, services and content. Dubbed Ginni, this engine has already gone live.
The purpose of the second project, named Josy, is to semantically analyse texts. By reading about IT, Josy learns to link certain concepts, for instance “firewalls” and “security”. Over time, it will be able to identify trending topics online and determine what might match customer needs.
The goal is for Ginni and Josy to be able to share their respective insights with each other, combining over 70,000 products and diverse services in Bechtle’s portfolio with hot topics filtered from the vast wealth of information available online. When this happens, the result will be a quantum leap in quality, a one-of-a-kind AI able to master a considerable degree of complexity. It will merge the analysis of Bechtle’s offering with an assessment of market demand to give customers tailored, even inspiring, recommendations based on their user behaviour on bechtle.com. And this AI will continue fine-tuning its understanding, engaging in constant self-learning that requires enormous computing and storage capacities, which Bechtle has in its own data centre in Frankfurt.
As promising as it sounds, AI that delivers desirable, meaningful results still requires human intervention. Engineers must continually evaluate its results and optimise its programming. What comes naturally to the coffee shop owner is light-years ahead of what AI can do. The behaviour hard-wired into her human brain has been honed over millennia of evolution, and the capacity of her neuronal network is vastly superior to any AI system today. A simple twitch of her customer’s nose is all it takes for her to know what is on his mind based on the cinnamon smell wafting through the room. Ginni and Josy, even IBM’s Watson and Google DeepMind’s AlphaGo, still have a long way to go—if AI is ever able to develop complex sensory impressions like this at all.
HUMANS STILL HAVE THE ADVANTAGE.
The human cerebral cortex consists of 86 billion neurons whose synapses create 100 billion links. Artificial networks are nowhere near this
level of complexity.
AI is a make, technology a buy. We take advantage of technological expertise whenever it makes sense to do so, thereby automatically integrating AI developments that we cannot produce ourselves. But the AI we build ourselves, that’s ours.
Convinced that human brains and AI will merge sooner or later, the US futurist Ray Kurzweil has put forth the idea of a digital neocortex, in which the cerebral cortex is connected to a universal data cloud, forming a single unit. The resulting intelligence explosion would lead to technological singularity—undoubtedly a point of no return were it ever to occur.
As it stands, no one can say for sure whether we will ever cross this transhuman threshold or what the long-term relationship of humans and AI looks like. Until the end of his life, even Stephen Hawking could not make up his mind if this would be the best or worst thing ever to happen in the history of human civilisation. One of society’s biggest questions currently is how to set the course of AI so that it goes in the right direction.
In the meanwhile, Bechtle is taking a pragmatic approach to meeting customer needs more effectively. It hopes to increasingly calibrate its digital offerings to individual profiles with the help of self-learning programs. Providing regular visitors to Bechtle’s website with their own personalised version of the platform is no longer a far-off fantasy. Such tailored offerings would combine products and services into a modular “service-as-a-service”. They might possibly even be configured in real time with the help of a bot, perhaps using voice control. However, this might not happen for a while, considering that Google’s Duplex voice assistant required months of training before it could make a simple hair appointment.
Nevertheless, the AI being developed by Bechtle is already proving useful to portfolio development and sales efforts. Customer advisors can use machine-learning recommendations for inspiration. New products and services are being developed more purposefully. And company resources are used more effectively. In future, the next-generation system will end up knowing what customers want and when they want it—even before they do. A friendly hologram (perhaps named Lucy?) will appear just in time to present a solution for the problem that the customer is about to encounter.
AlphaZero is a master of strategy games like chess, Shogi and Go. But how does artificial intelligence measure up in areas without clear-cut rules?
Innovation Specialist and Projectmanager KI
CIO-Organisation, Bechtle AG
Get the best from the Bechtle update every two months directly into your mailbox. Click here to register:
Published on Mar 19, 2019.