What you need to know about emotional AI in CX

There’s a difference between great CX and brilliant CX and emotions have a crucial part to play here.
If you take away friction, and help people save time, you enable great CX. You may create positive emotions, but what you are mostly doing is taking away negative emotions like stress or frustration. Of course, nowadays brands often use technology for this time-saving, friction-decreasing part of CX. Think about self-checkout systems or online grocery shopping and delivery. So, that’s great CX.
But if you want to create brilliant CX, then you have to tap into positive emotions like joy and nostalgia and awe or inspiration or gratitude for feeling really understood. Up till now, that has mostly been the domain of humans. Rare were those instances where tech made you feel something deep.
Just think of that restaurant holder driving for 6 hours to deliver the last meal of a dying customer. Or a hotel employee giving his loafers to a customer who had an important meeting but forgot his fancy shoes. Or employees dressing up a horse like a unicorn for the sick little girl of a customer. Or a call centre employee kindly staying on the phone for 4 hours until a customer’s problem was solved. That’s the type of emotions that software cannot enable. For now.
But that may be changing. I have seen so many cases of emotion-analyzing and emotion-driven AI surfacing over the past few months that I wanted to share them with you.
Understanding emotions
The examples that I most frequently ran into were of companies that have invested in technologies that understand human emotions so they can respond appropriately. One of the oldest and most well-known examples is that of MIT Media Lab spinoff Affectiva, which was actually founded 15 years ago already. Their technology analyses human emotions and cognitive states and finds applications in automotive, media analytics, and healthcare.
But for the past months, we’ve seen quite some more news surfacing about companies investing in affective or emotive technology. Amazon, for instance, recently patented a system for “sentiment detection” for audio input in order to improve Alexa to the point where it can read the emotions in your voice, on top of what people are saying. It also had a more controversial project where its software scanned the faces of thousands of people catching trains in the United Kingdom. The image recognition system was used to predict travelers’ age, gender, and potential emotions—with the suggestion that the data could be used in advertising systems in the future.
Roku too launched a patent for “emotional evaluation of contents” to guide ad placements: the tech tracks the emotions displayed in a scene to place ads that match those emotions. And Snap has filed a patent for “sentiment classification” that gleans emotional context from user data, while Microsoft wants to patent a way to offer emotional support via its chatbot.
Emotional CX
Though these are already quite impressive, they are mostly about providing tailored marketing and about selling, rather than providing fantastic CX, which is of course my preferred domain. So it was interesting to see that quite some companies are taking steps here as well.
We’ve had voice assistants like Siri and Alexa for quite some time now, and though generative AI has given their intelligence a big boost, the interaction is often disappointing because they tend to strip all emotion out of the interaction. And so the next frontier here is to develop chatbots an artificial humans that cannot just handle complex questions – or even act on our behalf soon, as “agents” – but that are empathetic and pleasant enough so users will want to keep engaging with them at home and at work. This “emotional AI” will be the holy grail for text bots as well as voice bots and artificial humans. But it will certainly also be highly relevant when humanoid robots will deliver upon their potential. This might be sooner than you think, seeing that so many companies are investing in the area of embodied intelligence.
We’ve already seen examples in the text bot area, if you recall the recent GenAI study for drafting replies to patient messages. The panel preferred ChatGPT’s responses over physicians in 78.6% of the 585 completed evaluations. ChatGPT’s responses were also longer than the doctors’ responses, ranging from 168 to 245 words compared to 17 and 62 words, respectively. The AI responses also rated significantly higher on both quality and empathy.
We also saw Google file a patent application for a voice assistant capable of “emotionally intelligent responses” to questions. Its wants to determine a user’s emotions, and modifies its response based on that. Microsoft Research, then, unveiled VASA-1, a new technology capable of generating lifelike, audio-driven talking faces in real-time. It leverages advanced algorithms to ensure the facial expressions are accurate and convey the emotional tone of the spoken words.
Bland AI released a hyper-realistic AI phone agent earlier this year: The AI answers instantly, understands emotions, and responds in virtually any voice and even any language. Hume AI’s bot, EVI, also processes the emotions the speaker is showing — like excitement, surprise, joy, anger, and awkwardness — and then expresses its responses with ’emotions’ of its own. For instance, if you yell at it, it will get sheepish and try to diffuse the situation. Ultimately, their vision is to build AI into products, allowing an AI assistant to read your speech, emotion, and expressions, and guide you through the experience.
Even one step further, perhaps, going beyond understanding and reacting to emotions, is adding humour to the interactions, which is a very hard thing to do. When Anthropic released Claude 3.5 Sonnet, it specifically added that the model is significantly better at grasping nuance and humour, among other improvements. Elon Musk claims that xAI’s Grok chatbot is supposed to have a rebellious attitude and a sense of humor. Not sure that is the case, but the intention is there. Google DeepMind also did research about humour, asking 20 comics to share their experiences using leading chatbots to help them write jokes. The results were underwhelming, but my point is that quite some companies are investing in humanizing their bots with emotions and other human characteristics.
Ethical concerns
It seems that a lot of this research into emotions for chatbots and virtual humans and friends – like those of Replika or the emotional support bot of Inflection AI – is about differentiation and stickiness. Companies are wondering how to convince users to choose for their chatbot. The informational quality of those if often very similar, but their emotional intelligence could very well be the major USP, here. Old school social media had a lot of tricks to make their product “sticky”, or even addictive. Chatbots will need a very different approach for the same result.
Here too, there is the danger of addiction lurking around the corner. In fact, a safety analysis from OpenAI warned that ChatGPT users could get emotionally attached to GPT 4o’s new human-like voice. One potential negative impact is that GPT-4o may help people feel less lonely but that this would come at the cost of real connections with other humans.
So, even though I believe that there are a lot of really interesting and useful applications when it comes to creating emotional artificial intelligence, we have to be careful and mindful. Because there will be quite some challenges for addiction, privacy and security, to name just a few.
Humans will be ready
There are studies where humans like an exchange with an empathetic AI when they do not know that it’s an AI but then change their mind and feel highly uncomfortable when they learn it’s not a real human. If you think that emotional tech will never catch on because of this, just remember that GenZ doesn’t care either way.
A survey by Sprout Social conducted in February and August of this year uncovered that, when it comes to virtual influencers, only 35% of GenZ respondents said they care about authenticity. Instead, they care more about follower count (47% of them) and posting frequency. Also, 37 percent of GenZ said it would make them more interested in the brand if it works with an AI influencer.
This divide is probably not surprising since GenX, the Boomers and even the early Millennials (partly) grew up in a pre-digital world. I remember a time when social media in itself was not perceived as authentic. You had the physical, the real world and then you had the virtual world, which was not “real”. That perception has completely changed over the years. No one today would count social media as not “real”. Just like that, I think that the concept of “authenticity” and the gap between “real” and “virtual” will change as years tick by, or that we will at least use other words than “authentic” to measure the effectiveness of artificial humans, bots, agents or influencers. And what GenZ showed in this survey is a sign of that.
These are very interesting times to live in, and I wonder how soon – if ever – technology will be able to offer the emotional CX needed to create truly brilliant tech-driven experiences. But above all, even if this ever happens, humans will still keep playing a crucial role in CX. Always ask yourself this question before you replace a human: will this be better for the customer (or just better for our company)? If the answer is “no”, you’ll just have to wait it out, or combine the forces of tech and humans in an augmented way.