“A thing that knows no grief has no right to poetry.”
Samuel Hurley - Poet
No, you’re not about to read yet another lament about AI’s lack of emotion. We want to take a different road and ask a different question. What if the thing that knows no grief can still comfort you? What if it can still understand your problem, identify what you want and give you exactly what you need in that moment?
What you’re about to read is a rallying cry to design empathy into machines. Not as a watered-down expression of some non-existent sentiment, but as foundational AI infrastructure that will help you better serve your clients.
Time to evolve your definition of empathy
To understand how AI can be designed to be empathetic, we need to move beyond the traditional definition of empathy.
In a business context, empathy isn’t just about sharing feelings. It’s about recognizing, understanding and responding to emotional states. This distinction is crucial when designing AI systems. The way I see it, there are two ways to classify empathy in a business context:
- Emotional empathy: the ability to feel what someone else feels in a certain moment, by pulling from a shared experience. This is deeply human, difficult,and impossible for AI to replicate authentically.
- Functional empathy: the ability to recognize emotional cues and respond appropriately. AI doesn’t feel, but it can be trained to detect tone, sentiment, and context, and adapt its responses accordingly.
Functional empathy is what makes AI useful in business contexts like communication, marketing, and customer service. Not mimicking emotion but rather meeting emotional needs. And in many business contexts, functional empathy is exactly what can make the difference for your clients.
Think of a customer support chatbot that de-escalates frustration with calm, personalized responses. Or a shop assistant that adapts its tone based on a shopper’s level of disappointment with certain previous products. Or a recommendation engine that understands when you’re not just shopping but seeking comfort.
These cases don’t describe emotional machines, but they do describe emotionally aware systems. A powerful distinction, don’t you think?
And as a former copywriter, I’d like to note that voice isn’t a nice little detail in these cases. On the contrary, voice is often the first thing users respond to emotionally. According to my colleague Matthew Bradbeer: “A consistent and well-crafted tone of voice is essential for making AI interactions feel more human and aligned with your brand.”
In his article on designing AI with empathy (coming soon), he argues that tone, accent, speech patterns and vocabulary of your AI agent shape how human it feels. Voice is central to how empathy is conveyed. When it is tuned to context and culture, even the age of the customer, it creates trust, diffuses tension, and makes the interaction feel human.
How to build an emotionally aware system
Just like accessibility or usability, empathy can be:
- Engineered: Through natural language processing, adaptive tone, etc.
- Trained: On emotional user data like sentiment analysis, preferred brands, etc.
- Tested: Through user feedback, emotional response tracking, A/B testing, etc.
- Optimized: Based on data about context, history, persona, intent, etc.
If this is the case, if it is possible to create synthetic empathy, then it is no longer enough to wait around for it to occur organically. It shouldn’t be left to a content designer to sprinkle some empathy at the end of the prompting process. On the contrary, empathy should be part of how you design your AI foundation.
“This means empathy is no longer a soft skill, it’s a design principle.”
Sarah CR Claeys, Director of Content Design, Empathy Lab by EPAM
At Empathy Lab, we think of empathy as part of the experience architecture you are building and automating with AI – read more in our first article, A business case for empathy. It’s not a performance. It’s a protocol. And it’s becoming a core part of how we think about inclusive design with AI:
- Interfaces that adapt to neurodivergent users
- Chatbots that adjust tone based on emotional state
- Synthetic personas that evolve emotionally over time in gamified environments
How empathetic should your AI be?
The answer depends on your context, your users, and your brand’s values. But one thing is clear: empathy in AI isn’t binary. Like all our human emotions, empathy exists on a spectrum. Let’s call it the Empathy Spectrum of AI (ESA). It ranges from basic scripted responses to deeply immersive, emotionally adaptive systems. Understanding where your AI sits (where it should sit to benefit your business) is key to designing meaningful, responsible interactions. And to allocating your budget and efforts, but that’s a whole other article.
Here is a short description of the stages on the Empathy Spectrum of AI:
- Scripted: At this level, AI follows predefined responses. It’s useful for FAQs or compliance-driven environments, but it lacks nuance. If your brand voice is warm and human, this may feel too robotic.
- Contextual: Here, AI can recognize the situation and adjust tone or content accordingly. It knows if a user is frustrated or confused and can respond with more care. This is a solid baseline for most customer-facing applications.
- Adaptive: This level allows AI to learn from past interactions and tailor its responses over time. It can mirror user preferences, adjust tone dynamically, and even shift strategies based on emotional cues. It’s ideal for long-term engagement and loyalty-building.
- Anticipatory: Now we’re entering proactive territory. AI at this level doesn’t just react, it predicts. It can sense when a user might churn, when a conversation needs escalation, or when a moment of delight could make all the difference.
- Holistic: The most advanced form, where AI creates the illusion of a persistent, emotionally aware companion. Think virtual coaches or assistants. This level demands the highest ethical standards and transparency.
So, how empathetic should your AI be? As empathetic as your users need it to be and no more. Overstepping can feel manipulative; underdelivering can feel cold. The goal isn’t to mimic human emotion, but to meet human needs. Empathy is a design choice. And like all good design, it should be intentional, contextual, and aligned with your brand’s promise. More about designing for AI in our third article (coming soon). If you’re not currently designing your AI with empathy as a design principle, then you’re designing for failure. Because the future isn’t just intelligent, it’s emotionally fluent.
A word of caution
Before you run off to brief your design and content teams to create an empathetic AI system, please pause and ask yourself some of the hard questions:
- Should our AI’s empathy be transparent or seamless? Should users know they’re interacting with a machine? Or should it feel natural and human-like?
- What happens when our synthetic empathy gets it wrong? When a chatbot misreads grief as sarcasm? When a virtual assistant offers cheerfulness in a moment of crisis?
- Are we already doing everything we can to grow empathy in our humans, not just our machines? Are we looking at AI to enhance human connection, or to avoid it?
- And importantly: can we simulate care without being misleading? Can we design emotional fluency without emotional deception? What happens when machines comfort without understanding? If we’re building emotional systems, can we be very clear about what’s real, what’s synthetic, and what’s safe?
At Empathy Lab, we’re not looking to replace your people with machines. We believe in a future filled with possibility, where machines amplify what makes us human. Our work may be continuously powered by AI, but we still lead the charge. If you’d like to have a conversation on how you can build empathy into your AI foundation, we are up for a chat.
Contributor in this article
