You can automate warmth. You cannot automate presence
- trinimaturana
- Tendencias y Futuro
Índice
There is a specific kind of message most people have received at least once.
It arrives at the right moment. The tone is warm. The words acknowledge something real. It feels considered, almost personal. You read it and think: someone understood what this moment was like.
And then, almost immediately, something else happens.
A small doubt. Difficult to name. Not about the words themselves, but about what sits behind them. Did someone actually write this? Did anyone actually think about me?
That doubt, even when it stays quiet, changes something.
The thing AI has learned to do
For years, the gap between what organisations wanted to communicate and what they were capable of producing at scale was a practical problem. There were not enough hours, not enough writers, not enough emotionally intelligent managers to cover every moment that mattered.
AI has closed that gap with remarkable speed. Not only by increasing output, but by becoming unexpectedly good at reproducing emotional texture.
The language of warmth, acknowledgment, genuine-sounding vulnerability, is no longer difficult to produce. A system trained on millions of human interactions can generate something that reads like care in seconds.
That capability is genuinely useful in many contexts.
It is also, in specific ones, a problem organisations have not yet fully reckoned with.
What trust is actually made of
Organisational trust has always depended less on what is said and more on what people believe sits behind the message.
When a leader recognises someone’s work, the words matter less than the sense that someone actually noticed. When an organisation communicates during a difficult moment, the tone matters less than whether employees believe anyone genuinely understands the human weight of what is happening.
That belief is built, or destroyed, through accumulated small experiences over time.
A recognition message that felt real. A leadership note that seemed to come from an actual person. A manager who said something that could only have come from paying attention.
Or the opposite. Messages that were technically correct but somehow hollow. Communications that hit every emotional note but left people feeling, inexplicably, more alone.
The second category has always existed. Bad communication is not new.
What is new is that organisations can now produce the second category while believing they are producing the first. Consistently. At scale. Without anyone noticing the difference until the damage is already done.
The trap is not obvious
That is what makes this genuinely difficult.
The pressure to use AI in emotionally sensitive communication is not coming from bad intentions. It is coming from real organisational constraints.
Managers who are overwhelmed and need support to communicate with their teams during a restructure. Leaders who cannot personally reach thousands of employees. CI teams trying to maintain consistency across geographies. Organisations facing moments, redundancy announcements, performance conversations, wellbeing check-ins, recognition after difficult cycles, where the volume of human contact required simply exceeds what is possible manually.
AI offers something real in all of those situations. The risk is not in reaching for it.
The risk emerges when efficiency becomes the primary measure of success. When the question shifts from “does this communication build genuine connection?” to “does this communication sound like it does?”
Those questions look similar. They produce very different organisations.
Performing humanity versus practising it
Empathy is not something organisations can template.
It is not how a message sounds. It is the accumulated experience of feeling seen, understood, taken seriously by someone who actually chose to pay attention.
That cannot be automated. Not because the words are too complex, but because what employees are interpreting is not just the message. They are reading the signal behind it: did this organisation consider me as a person, or did it produce something that would work for anyone in my situation?
When that signal weakens, something subtle but consequential shifts. People do not necessarily stop reading the communications. They stop believing them. And once that happens, even genuine human effort starts being received with suspicion. The organisation has trained its people to distrust closeness.
That is a harder problem to fix than bad writing.
The question internal communication needs to sit with
None of this means AI should stay out of emotionally sensitive communication. That framing is already too simple.
The more honest conversation is about judgement.
Which moments require genuine human presence, not because AI cannot handle them technically, but because the relationship depends on someone actually showing up? Where does AI genuinely support better communication, and where does it quietly replace the relational effort that was the whole point?
And perhaps the most uncomfortable question: how would we know if we had already crossed that line?
Because organisations that start performing humanity instead of practising it rarely announce the transition. It happens gradually, through small decisions that each seem reasonable. Until one day the communications are impeccable and the trust is gone.