After years of learning how to talk to machines, we are beginning to move beyond the need to look at them. Visual interfaces are evolving into voice, gesture and conversation, turning technology from an interface customers see into a presence they experience, while brands invest more than ever to make that presence feel human. The data points to the scale of this shift. The global affective computing market is expected to grow from $88 billion in 2024 to $347 billion in 2029, at a compound annual growth rate of 31.4%. By 2025, 88% of marketing professionals were already using emotional AI in their daily work.
At the same time, another evolution was closely linked to this shift, even if the connection was not immediately clear. The NTT DATA Technology Foresight 2026 report places agentic capabilities and emotions within a single macrotrend. They are two sides of the same shift. Customers no longer want to navigate; they want to delegate. Autonomous agents are taking on tasks that once required their time and attention. We are moving from competing for attention to competing for intention.
These two curves seem to reinforce each other, but they may also be cannibalizing one another. When the interface disappears, so does the clarity around who or what is on the other side.
Who are we designing empathy for?
In its “Agentic capabilities and emotions” macrotrend, NTT DATA Technology Foresight 2026 raises an important question. How can we ensure that emotionally intelligent systems strengthen human well-being and autonomy, rather than influencing emotions in ways we cannot properly supervise? There is another question we need to ask. If the recipient of the conversation is an agent, who are we designing empathy for?
I call this tension the “recipient paradox.” We are investing in humanizing conversations just as the other party is becoming less human. This raises a critical question: are we investing millions to make machines feel something?
Technology Foresight raises a complementary question. Where do we draw the line between empathetic support and manipulation, and who is responsible for that line? This is not a rhetorical question. When brand interactions move through voice, gesture and intermediary agents, the boundary between supporting a customer and influencing that customer becomes thinner and harder to audit.
Transactional vs. relational modes
There are two modes of engagement, and they do not behave in the same way.
- Transactional mode comparing, buying, executing, repeating can be delegated. The customer wants efficiency, not presence. The emotional layer adds friction. The agent filters it out and, if it detects it, may even penalize it.
- Relational mode a complaint, a stressful decision, a failed service experience remains human. This is where synthetic empathy delivers what Technology Foresight describes as trust through presence. It is no coincidence that the report highlights affective service recovery detecting frustration early, adjusting tone and escalating proactively as one of the areas where emotional AI creates real value.
When booking a flight or comparing prices, customers do not want empathy. They want speed. An agent that executes the task without friction creates value. But when customers face an unexpected cancellation or miss a connection, expectations change. Automated responses with an “empathetic tone” fail and can even intensify frustration. The problem is a clear mismatch between the interaction mode and the organization’s response.
The same person, on the same day and with the same brand, may move between both modes. The challenge goes beyond segmenting customers as “rational” or “emotional.” It requires identifying, in real time, the mode of each interaction and responding with the right layer.
Humanizing the seller, dehumanizing the caregiver
The recipient paradox appears when organizations misread the interaction. Organizations invest in humanizing the transactional layer, even though the customer has already delegated the task and the agent will treat the affective layer as noise. At the same time, they automate relational moments with cold efficiency, precisely when the customer needs to be heard by someone who understands. We humanize the seller and dehumanize the caregiver. So far, this has remained an investment mistake that few have clearly identified.
According to Technology Foresight 2026, 31% of emotional AI use is concentrated in customer experience monitoring, while 62% is focused on patient engagement. This offers a clue as to where synthetic empathy finds a human recipient. It creates value when the customer, patient or user is not optimizing a transaction, but navigating a difficult situation.
Designing experiences for this new era is not about building screens. But it is not about adding emotional layers to everything a brand touches, either. Synthetic empathy is not a sales channel. It is relationship infrastructure. Confusing the two means investing in the wrong place.
That is why every interaction should begin with one key question: Is there a human being on the other side? The major challenge organizations will face over the next decade in their use of emotional AI is whether they truly know who they are designing it for. Those that fail to do so will continue refining the conversation in increasingly sophisticated ways without ever finding anyone behind it.