AI Can Smile, but it Can’t Laugh (Yet) - Part 1

We’re living in an incredible new age – an age where it feels like we can use generative AI magically to conjure up whatever we’re dreaming in our minds. With a few taps of the keyboard, we’re able to bring ideas, concepts, and even code to life, building content that is almost indistinguishable from reality. With daily announcements of new AI tools and capabilities, it’s truly incredible to see how far we've come in overcoming the original 'uncanny valley'—that unsettling feeling we get from almost-human AI.

As we advance so rapidly and cross one barrier, however, we’re seeing a new one emerge that keeps us from truly bringing to life what’s in our mind’s eye: the 'emotional uncanny valley.' For VPs of Marketing and Sales Enablement leaders, navigating this valley is becoming critical, as the authenticity of AI-driven communication directly impacts customer engagement, brand perception, and ultimately, sales effectiveness.

This valley is an invisible barrier that keeps AI from truly mimicking the full range of the human experience, especially when it comes to conveying human feelings. In a B2B context, where trust and nuanced understanding are paramount, this emotional gap isn't just a technical hurdle—it's a potential roadblock to building genuine connections. My own avatar, for instance, can deliver messages that look, sound, and feel like they're coming directly from me. Yet, when it comes to conveying genuine emotion, that's where the illusion often falls apart quickly.

We've all seen AI 'smile,' but can it truly laugh? Can it feel the sting of anger, the pang of sadness, the rush of surprise, or deliver a punchline with the perfect comedic timing? While AI can process data and generate content with remarkable efficiency, it struggles to replicate the very essence of what makes us human: our feelings.

In our 2-part article, we're diving deep into why this happens and exploring how we can help AI climb out of this emotional valley. Part 1 will map the current landscape, especially for B2B applications, by assessing AI’s current emotional capabilities and introducing a framework for thinking about them. Part 2 will explore how we can bridge this gap and what the future might hold if AI can truly understand and express human emotions.

PART I – Where We’re at Today: Assessing AI's Emotional Output in B2B Sales & Marketing

Understanding and responding to human emotions is crucial in any interaction, but it's the lifeblood of effective sales and marketing. While the range of human feelings is vast, five core emotions serve as fundamental building blocks. For this exploration of the 'emotional uncanny valley,' we'll provide a framework for assessing AI's current emotional capabilities, particularly focusing on audio, avatar, and video—modalities where AI's attempts at human-like emotional expression are most directly visible and audible, and thus where shortcomings can be especially pronounced.

Core Emotions: How Might AI Media Convey Them in B2B?

Let’s briefly set the B2B context for each core emotion before evaluating AI's performance:

• Joy: Imagine a short, AI-generated video clip (distinct from a simple talking avatar) for a social media campaign, designed to visually convey the collective excitement and joy of a new company achievement, perhaps using dynamic visuals and expressive animated elements.

• Anger: Consider an AI-driven training simulation where an avatar portrays a frustrated client, helping sales or service teams practice de-escalation skills by recognizing and responding to vocal tones and facial cues of annoyance.

• Sadness: Envision an AI-generated audio update for an internal team, needing to convey a somber or empathetic tone when delivering disappointing company news or acknowledging a shared challenge.

• Surprise: Picture an AI avatar in a product demonstration video reacting with mild, engaging astonishment when unveiling an unexpected new feature to capture audience interest.

• Humor: Think of an AI video presenter attempting to deliver a light-hearted opening remark or a brief, amusing anecdote in a marketing webinar, aiming to build rapport and make the content more engaging.

AI Emotional Capability Assessment: Examples & Believability Scores

We will now look at examples of AI attempting to convey these five core emotions across audio, avatar, and video modalities, providing a core believability score for each.

Joy

• Audio Example : Play Here

• Avatar Example:

• Video Example:

Anger

• Audio Example: Play Here

• Avatar Example:

• Video Example:

Sadness

• Audio Example: Play Here

• Avatar Example:

• Video Example:

Surprise

• Audio Example: Play Here

• Avatar Example:

• Video Example:

Humor

• Audio Example: Play Here

• Avatar Example:

• Video Example:

In-Depth Comparative Analysis & B2B Implications: Navigating AI's Emotional Landscape

Having assessed the core believability of AI-generated audio, avatars, and video across five key emotions, several critical patterns emerge. These findings offer crucial guidance for VPs of Marketing and Sales Enablement professionals looking to leverage these technologies without falling into the 'emotional uncanny valley.' While 'perfect' emotional replication remains elusive for AI, the degree of believability varies significantly by modality and the specific emotion being conveyed.

1. AI Audio: A Mixed Bag of Clarity and Clumsiness

Across our evaluations, AI-generated audio presented a mixed performance, often scoring in the mid-range for believability (e.g., Joy and Anger at 3/5, but dropping for Sadness and Surprise to 2/5, and Humor to a low 1/5).

• Key Finding: In general, the AI audio files did a better job than the avatars at being believable. This might be attributed to the longer development history of speech synthesis technology. For straightforward emotional tones like basic joy or a serious, firm tone that might be associated with anger, the audio could often convey a general sense of the emotion.

• The Interjection Hurdle: However, a significant weakness emerged: AI audio struggled profoundly with pronouncing interjections (like "woohoo" or "aha!") or laughter. This particularly impacted the believability of humor or surprise. These interjections create much of their emotional meaning from nuanced human delivery, an area where AI generation still lacks sophistication.

B2B Implications for Audio: For VPs of Marketing, this means AI audio might be suitable for simpler emotional cues in, for example, an upbeat instructional video or a direct, serious announcement. However, Sales Enablement Professionals should caution teams against relying on AI audio for scenarios requiring nuanced humor, spontaneous-sounding surprise, or deep empathy, as the artificiality can be jarring and undermine credibility.

2. AI Avatars: The Deepest Dive into the Uncanny Valley

Our analysis consistently showed AI avatars as the least believable modality in conveying emotion, with scores frequently at the lower end (e.g., Anger, Surprise at 2/5; Sadness, Humor at 1/5; and even Joy only reaching 3/5).

• Compounded Weaknesses: Avatars often inherit the vocal limitations of their underlying AI audio. If the audio can't convincingly laugh, the avatar's attempt to visually sync with it will also fail.

• The Full-Body Challenge: Furthermore, emotions are very much expressed through the whole body. Current AI avatars often have limited animation rigs, resulting in stiff or repetitive gestures that don't align with the nuanced body language accompanying genuine human emotion.

• The Archetype Factor: The selection of the visual persona used in the avatar was extremely important; certain archetypes fit particular emotions better or worse (an older male figure, for instance, might convey anger more readily than a younger female figure, due to societal pre-conditioning). This adds another layer of complexity to achieving believability.

B2B Implications for Avatars: Extreme caution is advised. While avatars are visually engaging, their current inability to consistently deliver believable emotional nuance makes them a high-risk choice for many B2B interactions. If considering avatars, VPs of Marketing and Sales Enablement must spend time aligning the avatar's appearance with the intended message and emotion and rigorously test the synchronization of visual emotional cues with the audio.

3. Video (Human Examples): The Gold Standard and a Usability Challenge

Unsurprisingly, the video examples featuring humans consistently scored highest for believability across all emotions (typically 5/5, with Surprise at a still strong 4/5).

• Inherent Authenticity: Humans are wired to understand and convey human emotion with a complex interplay of facial expressions, body language, and vocal tonality.

• The Scalability Challenge: However, the strength of human video presents a challenge for scalable, personalized AI-driven campaigns. While AI can manipulate video, seamlessly integrating authentic new emotional performances into human video at scale, driven by AI prompts, is still largely in the future for mainstream B2B applications. Thus, these highly effective human videos are often difficult to utilize for personalized AI messaging beyond general b-roll.

B2B Implications for Video: Human-starring video remains the benchmark for emotionally resonant B2B communication. For VPs of Marketing, this benchmark is crucial when evaluating the trade-offs of using AI-generated content.

Overall Emotional Expression & The Path Forward for B2B Leaders

Looking across the scores, emotions like Humor (1/5 for both AI audio and avatar) and deep Sadness (1/5 or 2/5 for AI) proved exceptionally challenging for AI modalities. This isn't surprising, as these emotions rely heavily on subtle cues, cultural context, timing, and a depth of feeling that current algorithms struggle to replicate.

For Marketing and Sales Enablement Leaders, the path forward involves:

• Strategic Modality Selection: Understand that AI audio, avatars, and video have vastly different capabilities in emotional expression.

• Human Oversight is Non-Negotiable: For any B2B communication where emotional authenticity is key, human review, editing, and often, human creation remain essential.

• Embrace AI for the Right Tasks: AI can be powerful for initial drafts or simple informational content but is risky for nuanced, high-stakes emotional communication.

• Focus on a "Credibility Scorecard": As leaders, begin to evaluate AI tools not just on features, but on their "emotional believability." Consider dimensions like vocal sincerity, facial congruence (avatars), natural handling of interjections, and contextual appropriateness.

By acknowledging these current realities, B2B leaders can harness the efficiencies of AI where appropriate, while safeguarding their brand and customer relationships from the pitfalls of the emotional uncanny valley.

Coming in Part 2: With these limitations and a preliminary framework for assessment established, Part 2 will explore actionable strategies for B2B leaders to work with current AI capabilities while pushing for greater emotional authenticity. We'll delve deeper into the "Emotional Uncanny Valley Scorecard," discuss advanced techniques and ethical considerations, and envision the transformative B2B use cases that truly emotionally intelligent AI could unlock.

Call to Action

We encourage VPs of Marketing, Sales Enablement Professionals, and anyone leveraging AI in B2B interactions to begin observing and informally "scoring" the AI content they encounter. What are you seeing? Where are the biggest emotional gaps impacting your objectives? Join the conversation and share your thoughts as we collectively navigate the future of AI and emotion in the professional world.