The AI empathy gap: Why AI literacy should include emotional literacy

Recently, I used artificial intelligence (AI) to write a difficult email. A client conversation had gone wrong, and I owed them an apology. I typed the situation into a large language model (LLM) rather than deal with the discomfort of finding the words myself.

AI's reply arrived immediately. It was competent, careful and strangely padded with filler words, inflating the one-line apology into a whole paragraph. This is what I call “empathy bloating” – writing that sounds like it’s coming from the heart, but your heart was never involved.

AI is improving our productivity in astonishing ways, diminishing hours of writing into seconds of work. However, efficiency comes at a cost, and we are now beginning to realise just how dear that price tag is.

According to the National Bureau of Economic Research, people send around 342 million messages a week to AI about relationships, feelings and personal life.

Each of those is a conversation outsourced to a machine, a trend that can also be seen in the workplace: a US survey found that 35% of employees have used AI to draft or edit a sensitive work communication, with nearly one in four using it in daily email – from responding to tricky clients to providing constructive feedback to team members.

People are increasingly attuned to AI writing. When leaders lean heavily on it, the share of employees who see their leaders as sincere falls from 83% to 40%. We are gaining productivity while quietly, almost invisibly, losing our empathy.

Each time we outsource empathy, we weaken that muscle. The next disagreement you have with a colleague or tense client call will need sharp and quick thinking, without time to consult an LLM. Should this consistent outsourcing be replicated across an organization, then you’re left with a skills gap.

AI transformation is sexy right now; it gets a ringfenced budget, board attention and town hall announcements and it mostly runs separately to the people track, covering culture, skills and wellbeing – garnering a fraction of the investment and none of the glamour.

When the two tracks never converge, employees gain the impression that this exciting future being painted does not include them.

Leaders currently have a rosy perspective of AI adoption, which deepens with seniority: in a survey of 1,400 workers and leaders, as reported in Harvard Business Review, 80% of executives believed employees felt enthusiastic about AI adoption, 58% of middle managers agreed and among employees themselves, only 29% did.

People rarely refuse the new tools; they simply leave them unused and the promised productivity never materialises.

The World Economic Forum ranks empathy and active listening among the core skills employers value most. This surprises people, because empathy is confused with sympathy, by people and LLMs alike: my bloated apology was sympathy dressed as empathy.

Sympathy is pity from a distance and it builds parent-child cultures. The empathy I have spent 15 years embedding in companies is closer to hostage negotiation: direct, adult to adult and measured in commercial results.

Machines can take on the work and mimic polite niceties but only humans can build the trust that makes the work possible.

AI behaves unlike any technology before it. A spreadsheet won’t do anything until it receives an instruction; Excel has never said “Hello Belinda.” AI, however, speaks in the first person, drafts messages you did not dictate and performs an empathy it does not feel.

An agent that speaks to you like your colleague, therefore, deserves the same induction and oversight afforded to that colleague.

AI labs certainly bake in some of the AI’s character but an organization can write its own layer on top, i.e. the prompts and rules that tell the agent how to behave with your people and your customers; auditing that layer is part of my job.

The pattern is common across organizations: prompts speak fluently of process, scores and targets, while the language of feeling, listening and reassurance barely appears.

The people writing your prompts are optimizing for efficiency and empathy is rarely in the brief until the 20th iteration. The empathy nudges are smaller than leaders expect: a few tweaks to the language and the agent starts reinforcing the culture you want.

With this in mind, there are three ways to close the empathy gap:

Embed empathy into your AI at the outset, practising it yourself. The next time a difficult email needs writing, the LLM will feel efficient but the minutes saved equate to empathy lost.

Sit with the discomfort, find the words yourself and use AI as your editor, not your writer. Empathy is the most valuable leadership skill and it only counts when it comes from you.