By Andrew Mort
Posted on Apr 2, 2019
Today’s contact centers face a daunting challenge. The explosion in connected devices and the prevalence of online shopping mean that millions of products are now shipped daily,
With online commerce projected to pass the $3 trillion mark in 2019. And with 51% of Americans now preferring to shop online, the pressure on customer support operations is about to increase exponentially.
It’s not only about rapidly rising call volumes, but also more complex and varied technical issues. As a result, countless high-tech solutions have emerged over the past decade, creating a buyer’s dilemma for decisions makers. Bombarded with buzzwords, and ever-conscious of meeting their KPIs, customer experience managers must choose between a dizzying range of automated solutions that all promise to reduce average handling time, motivate agents, improve first time resolution rates and enhance customer satisfaction.
However, most of these innovations revolve around language. From speech recognition and voice analytics to neuro-linguistic programming, they seek to solve customer’s problems by telling them what to do. The missing link has always been the visual dimension: the ability to show the customer exactly how to proceed. And that gap was the genesis of the computer vision revolution in customer assistance.
What is Computer Vision AI?
Computer vision is the technology that enables computers to see, recognize and process images in the same way as humans — and then some.
Automatically extracting and analyzing data from both stills and video, computer vision gives a machine the power to derive meaning by interpreting the visual data according to sophisticated algorithms built around previous experiences.
As one of the most exciting forms of artificial intelligence, computer vision is already being implemented across a wide range of business sectors, enabling enterprise systems to achieve high-level understanding from digital images and then suggest — or even perform next best actions.
Computer Vision AI Comes of Age
A watershed in the field occurred in 2015, when computer vision overtook humans in the ability to recognize objects, a turning point analogous to the day in 1997 when IBM’s Deep Blue chess computer defeated the legendary grandmaster Garry Kasparov.
That moment opened the world’s eyes to the potential of artificial intelligence, and since then the technology has, of course, come on in leaps and bounds. Its full value, however, while easy for the corporate world to comprehend, was difficult to explain to the layman. The visual dimension — the means by which ordinary people could relate to AI — was always lacking.
Over the past few years, computer vision AI has emerged as a key customer-facing technology in both the B2C and B2B realms. From small helpdesks operated by specialist manufacturers to vast contact centers run by leading telecoms and consumer electronics providers, end users can now receive faster, more effective service and support from both live human agents and virtual assistants.
Computer Vision AI — Reinventing the Contact Center
Live video calls through which contact center agents assist customers with their issues is the perfect environment for computer vision to demonstrate its worth.
Object Recognition
Even before a customer connects with a representative, the technology’s capacity for lightning-fast object recognition enables it to classify the nature of the call, assess its urgency and carry out a series of basic checks.
Computer vision-powered systems can zero in on objects within images or videos, isolating them from the background. Backed up by sufficiently large data sets, they then recognize the category of a given device, even identifying the manufacturer and the exact model number.
It doesn’t end there. The most advanced computer vision systems can now recognize a device’s components and their operational statuses — seeing exactly what the problem is in real time, achieving accuracy levels of over 95%.
Armed with all this data, a computer vision-enabled platform then identifies the correct department and agent, providing a full visual analysis and a suggested fix, based on previous successful resolutions. This information is then relayed to the customer, usually in the form of augmented reality instructions, overlaid on the user’s mobile screen.
When it comes to gadgets and gizmos, it’s often a case of simple troubleshooting, and showing the customer which buttons to press with simple arrows and boxes is proven to provide more first-time fixes, while boosting customer satisfaction scores.
Image Restoration
Real life is seldom neat and tidy. That’s why contact centers need computer vision to clean many of the images provided by their customers, eliminating blurring, reflections and shadows. This technique is performed by imaging a point source and using the Point Spread Function (PSF) to restore the lost image information. When a stressed-out customer is trying to show his router to a remote agent, the video image he’s providing isn’t going to win an Oscar. But with real-time image restoration, the contact center agent can literally make the best of a bad situation.
Facial recognition
We’re all more protective of our data these days, with identity theft still a hot-button issue. That’s why the companies we entrust with our most private information — like banks, medical providers and insurance companies — rely on computer vision as a foolproof means of verifying their customer’s identities through facial recognition — even in challenging lighting conditions and from unusual angles.
Image to text
A modern contact center relies on its knowledge base to streamline its operations. When a new type of issue has been successfully resolved, the challenge is to make the relevant information readily available across the organization. By automatically generating textual descriptions of objects or issues identified within images, computer vision platforms make it easier for everyone to search the company system and find exactly the solution they need. For example, an insurance agent can simply type “fender bender” or “cracked windshield” into the search bar and instantly find relevant images of similar incidents, enabling them to estimate the cost of the damage in no time flat.
Image similarity
Visual search engine technology from the likes of Pinterest, Bing, Target, and ASOS now enable us to search for images similar to the ones we already have. It’s hardly surprising that companies which receive millions of images from customers every day have embraced the possibilities. By dropping an image of an unusual error message on a device into a search bar, agents can find the right fix even faster than with text search. When it comes to those common troubleshooting issues — like when a customer has plugged the wrong cable into his new router — it’s a technology that can slash minutes off a call.
Motion estimation
For leading contact centers, it’s not enough to provide a customer with instructions — the agent must also verify that they’ve been carried out successfully. Computer vision algorithms can now track the changes from one 2D image to another — usually between adjacent frames in a video sequence — to track a customer’s movements and make sure that they’ve completed the fix correctly. Remote visual assistance gives a remote agent eyes on everything, but motion estimation essentially gives them the ability to guide a customer’s every move — giving the agent hands-on problem-solving abilities.
Crystal-Clear: The Future of Computer Vision
In the here and now, computer vision has become a must-have technology for the contact centers of the world’s top brands, routing customer enquiries and assisting agents with numerous decision support tools.
But with customer experience advancing rapidly towards full self-service as standard, we’ll soon be visually interacting with virtual assistants that can visually guide us towards self-resolution of all our tech issues.
This article was first published on the TechSee blog.
