Technical capabilities on the horizon for conversational AI

Customer service 3D manager concept. AI assistance headphone call center
(Image credit: Shutterstock/LuckyStep)

Right now, AI is the dumbest it will ever be. I hear that line all the time.

Our favorite adjectives of “unprecedented” or “exponential growth” can’t quite capture the pace at which AI is learning, iterating, and advancing. The models that wow us today might be rudimentary tomorrow.

It’s both a scary and thrilling thought. Scary, when we think of regulations playing catch up, the computational bandwidth it might require, and ensuring tech stack agility. But thrilling when we think of the innovation, scale, and productivity just on the horizon. For now, let’s focus on the latter.

One of the most notable “up-and-coming” sectors of AI is in conversational voice and chat. Right now its market valuation is at $5.8 billion. In three years it’s expected to hit $31.9 billion – a 450% increase.

AI-powered agents offer a level of precision in customer engagement we haven’t seen before. They’re simultaneously analyzing historical and behavioral data while keeping the natural flow of conversation – even understanding the nuance of a well-timed pause on a phone call.

These capabilities seem incredible today, but as we know, the bar keeps getting higher. What advancements can we expect from conversational AI agents in the near future?

Andy O’Dower

VP of Product at Twilio.

Adapting with human sentiment and responding to emotional cues

Sentiment is tricky to pick up on. Something as simple as receiving a one-word text can spark a moment of uncertainty – is this person upset? Is it just their style of communication? Usually, we navigate this by relying on our lived experiences and the context of the conversation.

Can AI do the same?

In the past, sentiment analysis was usually done by categorizing words as either positive, negative or neutral (i.e., a polarity score), which created a framework for AI to learn within. Businesses could then analyze social media comments or product reviews at scale to gauge brand sentiment.

Now, advanced machine learning algorithms are starting to go beyond these preset categories. It can understand how sarcasm changes the meaning of a normally positive word – e.g., the difference between an enthusiastic or jaded “wow” – or sense someone’s urgency in their rapid responses. In text, AI can look at punctuation or even historical data to glean emotional subtext; like the difference between “that’s great!” versus “that’s just great…”

By picking up on these subtle emotional cues, AI could start to adjust its responses in real time. This could range from simply acknowledging a person’s frustration to recognizing replies are becoming gradually more delayed, hinting at a waning interest. By weaving in predictive analytics, AI agents could even start to connect sentiment to things like churn or lifetime values.

Emotionally intelligent AI isn’t a net-new concept – in 1995 Rosalind Picard, an MIT lab professor, published “Affective Computing” on the subject. But this once-futuristic sounding idea is now within reach. It’ll be interesting to see customers' reaction. Emotionally intelligent AI can create better customer experiences, but will an AI agent that’s too perceptive come across as eerie and off-putting? I imagine, much like personalization, its effectiveness will ultimately come down to transparency and customer trust.

AI representatives in the metaverse

When generative AI catapulted into the spotlight it pulled focus from another burgeoning technology and channel: the metaverse. Described as a digital playground and a virtual marketplace, the metaverse is reimagining everything from how we purchase products, to education, and collaboration.

Just look to Roblox, the virtual 3D gaming platform worth ~$49 billion, whose growth has predominantly been organic. Its success has cemented immersive, digital experiences as a channel worth investing in. The NFL, Walmart, and Paramount all saw the potential in Roblox to reach new audiences and blend entertainment and commerce.

Even luxury brands are following suit. Gucci has created several metaverse experiences, like their 2023 Gucci Cosmos Land hosted in The Sandbox. As part of the experience, visitors could explore themed rooms, complete quests, and even purchase digital Gucci clothing. (It was reported that Gucci made over 1 million in revenue by selling digital items through Cosmos Land.)

I expect we’ll continue to see brands build in the metaverse, whether it’s a digital equivalent of a physical storefront or a standalone virtual experience. And AI agents will play a pivotal role. They could be personal shoppers, tutors, or run a 24/7 virtual concierge desk. With advanced language models, AI agents could switch between languages and incorporate accessibility features to make experiences more accessible and inclusive.

Even more meta, consumers could have their own AI agents acting on their behalf, like evaluating the best product or offer from multiple brands at once. It’s a completely new dynamic.

However, AI representatives will only be as effective as the data (and the data architecture) behind them. Real-time personalization hinges on the ability of an AI agent to process multiple data streams in milliseconds – analyzing everything from purchase history to user intent.

There might even be certain scenarios where a customer needs to share sensitive information (e.g., payment details, their home address to coordinate shipping, etc.). Success in the metaverse will depend not just on creating engaging experiences, but on maintaining user trust and compliance through responsible and privacy-conscious data handling.

Closing thoughts

Conversational AI is redefining how and where customers interact with brands. In the coming months or years, I expect conversational AI will become increasingly multimodal – able to process and respond to things like spatial awareness, a person’s tone and speech patterns, and even subtle cues like body language and gestures with greater precision.

The use cases we’re already seeing are brimming with potential, from top-tier customer service to integrating AI voice assistants into cars to help with navigation or even monitor driver fatigue.

When we think of what’s possible, there’s often a far-off feeling attached. But AI is rapidly reducing the time between idea and implementation. That horizon, of what AI will be capable of, is much closer than we think.

We list the best AI chatbot for business.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

VP of Product at Twilio.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.