In both B2C and B2B industries, contact centers are essential to providing outstanding customer experiences (CX). Organizations are using AI more and more to improve customer service and optimize operations as customer demands rise. Businesses typically use AI to alter contact centers, typically in the form of intelligent virtual assistants to gather predictive consumer data and to handle fraud and disputes, according to recent research in the MIT Technology Review.
Furthermore, according to Deloitte, contact centers are under pressure from growing volumes, shifting customer preferences, and unstable economic conditions, which makes investments in automation and artificial intelligence crucial. However, considering that most call centers and many CX rely entirely on audio for communication, the possibility of total efficiency in these business processes has yet to be investigated.
In addition to the massive benefits already brought about by machine learning and more recent generative applications of the technology, the potential to integrate visual CX data into AI-powered contact center solutions has the potential to completely transform customer interactions now that almost all mobile phones have cameras and video capabilities and businesses have the data infrastructure to properly capture visual data.
On the “AI in Business” podcast, recently discussed how incorporating visual data in contact centers will eventually represent a step-level improvement in B2B CX workflows with Jason Aubee, Senior Vice President of Sales and Head of North American Revenue at TechSee.
Three significant takeaways from their discussion are examined in the analysis that follows:
Improving comprehension through visual communication: Including visual components in customer contacts helps speed up problem solving and enhance understanding.
Putting dynamic AI interactions into practice: Making use of massive language models and sophisticated conversational AI to develop flexible, adaptive customer encounters that transcend programmed workflows and quickly meet consumer demands.
Sentiment analysis for customer interaction optimization: Using “episodic memory” to examine sentiment and success metrics from past interactions, the knowledge engine is continuously improved and refined to make sure responses are in line with successful outcomes and changing customer expectations.
Below, you can hear the entire episode:
Improving Comprehension Through Visual Communication
Jason begins by discussing the drawbacks of conventional contact center tactics, especially in regard to channels of communication and sensory modalities. He starts by highlighting the limited scope of existing methods, which only use audio-based communications and may, at most, switch to text-based interactions, making it more difficult to collect data efficiently.
He makes the case that people have multiple senses and stresses the significance of increasing the range of media types that contact centers can use to gather data. According to Jason, the contact center may increase its comprehension and resolve problems more quickly by including visual data and sensors into the discussion. He says that a picture (or at least some type of visual information) might save a lot of time when compared to verbal descriptions in the context of their profession, using the analogy that a picture is worth a thousand words.
“Everyone uses [the phrase, ‘a picture worth 1,000 words,’ which is a bit idealistic.” “Well, in our time, it’s a thousand words in seven minutes,” Jason informs the audience of the executive podcast.
He uses a personal story about resolving a Wi-Fi problem with a client whose son called the business on her behalf to further highlight his thesis, specifically highlighting the influence visual data streams will have on business-to-business and service workflows. Despite the son’s first inquiries, the constraints of verbal communication meant that it took a long time to determine the problem’s underlying source. However, the problem was fixed faster by broadening the viewpoint and taking into account visual indicators, such making sure wires were plugged in properly.
Jason goes into more detail on the idea of modality in relation to contact center omnichannel communication. According to him, omnichannel communication generally uses a variety of text-based channels, such as social media, email, chat, and SMS, but modality goes one step further by incorporating more senses into communication.
He highlights that contact centers may better inform agents and customers by utilizing many senses, including hearing, sight, and reading. For example, agents can more effectively communicate instructions by sending visual content, such as animated graphics, rather than vocally explaining a solution.
Additionally, Jason presents the idea of bi-directional notation, which enables more accessible communication and comprehension by enabling agents and consumers to interchange pictures. Creating such a strategy can entail directing attention using gestures like circling or pointing to particular visual components.
Putting Dynamic AI Interactions Into Practice
Jason then talks on the importance of cutting-edge technology like avatars, metahumans, and more intelligent bots in the future of contact center automation. With the advent of massive language models, he explains, the conventional scripted workflows of conversational AI are changing from strict “if-then” notions to more dynamic and adaptable interactions.
He explains an example scenario in which a client engages with Sophie, an interactive avatar driven by multi-modality and a vast language model. Sophie walks the customer through the process of configuring a new gadget step-by-step while having a discussion with them.
Most crucially, Jason emphasizes that this connection is not limited by pre-scripted routines; rather, they dynamically adjust to the demands and inquiries of the consumer, even going so far as to smoothly upsell extra goods or services. He provides an example of this concept by explaining the complete process where a consumer, referred to as a “home user,” gets a brand-new gadget from the TechSee website and contacts their contact center to troubleshoot how to use it:
As the chat begins, he essentially states, “I opened this, and I don’t know how to use it.” After demonstrating how to remove it from the box, apply filters, connect it, and get it going, [the Sophie avatar] smoothly moves on to say, “By the way, now that is working.” On my Google Home network, I can’t see it. In today’s conversational AI, that is not a scripted workflow; it isn’t even a scripted workflow that you give an agent because it initiates a fresh dialogue that goes, “Hey, we got you working.” You’re asking how to tie it to something that we don’t even provide now?
“Excellent!” says [Sophie]. I am proficient with Google Home. Show me what you’re examining. [She] looks at the screen, determines which button to press, how to link it, how to attach it, etc. Then, in Google’s world—not even the world of the final product manufacturer—the user simply clicks through the process, and we pass them off to an upsell at the conclusion.
— Jason Aubee, Head of North American Revenue and Senior Vice President of Sales at TechSee
Jason highlights that AI-driven systems like Sophie can manage upselling encounters naturally and without hesitation, in contrast to human workers who would be reluctant to do so out of concern for upsetting the consumer.
According to him, implementing AI technologies may make the customer experience more seamless and allow for the presentation of upselling opportunities in a more approachable, conversational, and less invasive way, which will eventually increase customer happiness and revenue.
Enhancing Customer Engagement Through Sentiment Analysis
Jason continues by outlining the fundamental ideas underlying the cognitive engine they created for their service provision. He distinguishes between two categories of memory: episodic memory and semantic memory.
Recalling facts, ideas, and numbers is referred to as semantic memory. Semantic memory in a call center would comprise knowledge from training materials, FAQs, manuals, and other documented knowledge required to effectively handle consumer inquiries.
Conversely, episodic memory entails acquiring, retaining, and retrieving knowledge from experience. This stands for tribal knowledge and applied content in the context of a call center—the understandings derived from prior encounters and experiences.
Jason demonstrates the distinction between episodic and semantic memory with an example. Along with precise information about the kind of battery required, semantic memory would determine that a given alarm indicates a battery replacement. Based on prior successful fixes of related problems, episodic memory—which is powered by machine learning from previous interactions—may recommend extra measures, including giving the client the precise battery and replacement instructions.
He goes on to explain how they use sentiment analysis and success measures like survey results and NPS (Net Promoter Score) to improve customer interactions and their knowledge engine.
They can find trends of productive conversations and incorporate this data into their knowledge engine by examining the sentiment and success metrics from prior calls. They can improve their reactions and more successfully adjust to various consumer situations thanks to this procedure.
To say, “Hey, we’ve done this 100,000 times,” we leverage all of your intrinsic interaction data, which is a veritable goldmine that many customers don’t use. What were the greatest and worst results, and how can we determine the reasons behind them? With analytics, we have been doing it for ages. In other words, “Just give us the best.” We’ll train our system to only produce the greatest results and to do it in a way that allows the response to change when the best outcomes do. You are under no obligation to recode it or do anything. It gains knowledge dynamically by interaction.
— Jason Aubee, Head of North American Revenue and Senior Vice President of Sales at TechSee
Jason concludes by talking about their approach to customer service and the changing expectations of their clients. He clarifies that they don’t restrict their clients’ ability to think about what they can do. Although many companies in the call center and services sectors first concentrated on using AI to standardize customer interactions, they have discovered that their clients’ needs and wants are constantly growing.
Keeping up with their customers’ changing demands and preferences is now their biggest issue. Customers want businesses like TechSee to become brand ambassadors who can handle a variety of problems, including ones that aren’t immediately related to their product or service, rather than only offering solutions. Jason shares a story about a customer service agent at Zappos who went above and beyond by placing an order for pizza for a customer who was in need while they were discussing how to get shoes. Only unassisted humans have historically been able to provide this degree of proactive and individualized help.
He maintains that TechSee’s objective is to use their digital skills to offer a comparable degree of human-like connection. In order to keep control over the brand message and provide outstanding customer service, they hope to provide conversational AI that can provide dynamic responses that can change course depending on the conversation’s evolving nature. Customers still need a degree of service that feels both human and customized, even when they are aware that they are engaging with digital systems.