In text, the words are all you have
On a phone call, a warm voice can carry an apology. In chat, there's no voice — the exact words on the screen do all the work of calming someone down. That makes phrasing unusually important when a customer is angry. The right sentence can defuse; a slightly wrong one, read in the customer's own frustrated tone, can detonate.
Phrases that calm
- “You're right, and I'm sorry this happened.” — a genuine, specific acknowledgment beats a defensive explanation every time.
- “Let me take care of this for you.” — ownership tells a frustrated customer the buck stops with you.
- “Here's exactly what I'm going to do next…” — a concrete plan replaces anxiety with a path forward.
- “I completely understand why that's frustrating.” — naming the feeling, sincerely, makes the person feel heard.
Phrases that pour fuel on it
Some lines reliably escalate. “Calm down” tells someone their feelings are the problem. “As I already said” scolds. “That's our policy” with no empathy reads as a wall. “Unfortunately there's nothing I can do” surrenders. Passive constructions like “your issue has been noted” sound like a machine shrugging. Each subtly blames or dismisses the customer.
The principle behind the phrases
De-escalation isn't a magic word list — it's a stance the words express: acknowledge the feeling, take ownership, and show a concrete next step, without defensiveness. Any phrase that does those three things helps; any phrase that blames, dismisses, or stonewalls hurts. Learn the principle and you can write your own lines for situations no script anticipated.
How MyLiveChat fits
MyLiveChat lets you keep well-crafted de-escalation phrases as canned responses an agent can reach for under pressure — then personalize before sending, since a pasted apology that ignores the specifics backfires. Transcripts also make de-escalation a teachable skill: review the chats that turned around and the ones that didn't, and learn from both.
Why phrasing carries more weight in chat
On a phone call, tone does most of the de-escalation work. A warm voice can carry a clumsy
sentence. In chat none of that exists — the words are the entire signal, and an upset
reader supplies the missing tone themselves, usually the worst available one. A neutral
“As I said, the policy is 30 days” reads as patient in the agent’s head and
as condescending on the customer’s screen.
This is why scripted phrasing matters more here than in any other channel. Not to make agents
robotic, but to give them sentences that survive being read in the worst possible tone.
The four moves, and a phrase for each
Almost all de-escalation is four moves in sequence. Skipping straight to the fourth is the
single most common mistake.
- Acknowledge the specific thing. “You have been waiting nine days for a replacement that should have taken three. That is genuinely frustrating.” Specificity is what makes this land — a generic “sorry for the inconvenience” signals a script, not attention.
- Take ownership of the next step. “I am going to sort this out with you now.” Not the company, not the department — you. Upset customers are usually reacting to feeling passed around.
- Say what you can do, before what you cannot. “What I can do right now is X and Y.” Leading with the limitation invites a fight about the limitation.
- Commit to something concrete. A named action with a time attached. Vague reassurance re-escalates the moment it expires.
Phrases that reliably make it worse
- “Calm down.” Never once worked. It tells someone their reaction is the problem rather than the thing they are reacting to.
- “Unfortunately, our policy is…” Hides a person behind a rule. Say what you can do, then the constraint, in that order.
- “As I mentioned earlier…” Scores a point about who was listening. Just repeat the information more clearly.
- “There is nothing I can do.” Almost never literally true — you can escalate, log, explain, or follow up. Say the true version.
- “You should have…” Assigns blame mid-complaint. Even when accurate, it costs more than it settles.
- Bare “No.” True answers still need a sentence of reasoning and an alternative.
Apologise once, properly
Repeated apologies read as insincere and as stalling. One clear apology that names the actual
failure — “We told you Tuesday and did not deliver, and nobody updated you.
I am sorry” — does more than five rounds of “so sorry about that.”
After the apology, move to action. Customers forgive the error far more readily than they
forgive the sense that nothing is happening.
Two mechanical habits change outcomes as much as wording does.
First, do not go quiet. A silent gap while you research reads as abandonment to someone
already annoyed. A short “still with you — checking the order history now”
every so often costs nothing and prevents the “hello??” that hardens the mood.
Second, slow down slightly and drop the decoration. No exclamation marks, no emoji, no
cheerfulness that clashes with the customer’s state. Short paragraphs, one idea each.
A wall of justification is read as defensiveness regardless of what it says.
Saying no without a fight
Sometimes the answer really is no. The shape that holds up: acknowledge, give the reason in
one sentence, state the decision plainly, then offer the best available alternative.
“I understand why you expected it to be covered. The warranty runs 12 months from
delivery and this was purchased in March last year, so I cannot replace it free of charge. What
I can do is get you the repair rate and check whether the goodwill discount applies —
would that help?”
Padding the refusal with excessive apology invites negotiation, because it signals the
decision might move. Clear and kind beats soft and ambiguous.
Know when to stop de-escalating
De-escalation language has a limit, and pushing past it burns agents out. When someone is
abusive rather than angry, the goal changes from calming to closing. Agents need explicit
permission and a scripted line — “I want to help, but I am not able to continue
if the messages stay personal. I am going to close this chat now” — plus the
certainty that a manager will back the decision. Without that backing, agents absorb abuse
because ending the chat feels riskier than enduring it.