Rapport in chat is not the same as rapport in person
Agents who came from phone or retail arrive with a set of warmth habits that do not transfer
cleanly. In person, small talk fills a gap that exists anyway while you look something up. In chat,
the same small talk arrives as a message the visitor has to read, consider and possibly answer, in a
window where they are waiting for an answer.
That is the core difference. Warmth in person is often free; warmth in chat costs the visitor
attention and time. It still pays, but only when it is bought deliberately.
Warmth early, efficiency in the middle
The pattern that works in most conversations is front-loaded. The opening line is where warmth does
its work, because the visitor is deciding whether this is going to be a good experience and whether
there is a person on the other end.
Once the actual problem is on the table, the visitor's definition of a good experience switches to
progress. Chatty asides in the middle of troubleshooting read as delay, however friendly the
intention. Then at the close there is room for warmth again, because the pressure is off.
Agents who get complaints about being cold are usually skipping the first phase. Agents who get
complaints about being slow are often extending it into the second.
Small talk that is worth the message
Most small talk in chat fails because it is generic. “How is your day going?” asks the
visitor to do work for no return, and often gets a one-word answer that helps nobody.
What does earn its place is small talk with information in it — a remark that shows you read
what they wrote, or that acknowledges their situation specifically. Noticing that someone is clearly
mid-migration, or that they have been patient through a long thread, lands as attention rather than
filler. It is warmer than a greeting question and costs the visitor nothing.
Using their name once, early, does more than most people expect, and using it in every message does
the opposite. Mirroring their register works the same way: if they write in clipped fragments, long
friendly paragraphs feel like a mismatch, and if they are chatty, pure efficiency reads as cold.
Humour, and the rule that keeps it safe
Humour is the highest-variance tool an agent has. When it lands it produces the conversations people
screenshot and share. When it misses it can turn a mild problem into a complaint about being mocked.
The rule that holds up: humour is safe when the visitor started it, and risky when you did. Matching
someone's joke about their own typing, or their own chaotic week, is joining in. Introducing a joke
into a conversation the visitor is treating seriously is a gamble with their problem.
Never be funny about the thing that is broken, the amount someone was charged, or how long they
waited. Self-directed humour is the safest register, and even that should stop the moment the tone
turns. Sarcasm does not survive the loss of tone of voice, and irony arrives as rudeness often enough
that it is not worth the attempt.
Rapport is mostly reliability
It is easy to treat rapport as a personality trait and miss that most of what makes visitors warm
to an agent is behavioural. Saying what you are about to do and then doing it. Narrating a silence
before it becomes worrying. Remembering what they said at the top instead of asking again. Following
through on something after the chat ends.
The agents people ask for by name are rarely the funniest. They are the ones whose conversations
feel predictable in a good way, and that is a set of habits any team can be trained in, unlike charm.
Keep the saved replies human
A library of canned responses is what makes speed and warmth
compatible — the fast path is fast, so there is time to be human elsewhere. But a library written
in careful corporate English will flatten the voice of every agent using it.
Write them the way a good agent actually types, keep a gap where the specific detail goes, and
expect agents to edit rather than send them raw. A saved reply that arrives with no personalisation is
recognisable as a template, and being handed a template is the opposite of rapport.
What to measure
Rapport resists direct measurement, so use the indirect signals. Ratings and comments are the
obvious one, but read the comments rather than the score — people describe the interaction in
words that tell you whether they felt attended to.
Watch the gap between how fast a conversation was and how well it was rated. Chats that were quick
and rated poorly usually mean the visitor felt processed, which is a rapport problem wearing an
efficiency costume.
In transcript review, look at first and last messages specifically. Those two are where warmth is
won and lost, they take seconds to scan, and they are the most coachable part of any conversation.