Gains:
- Ability to produce empathetic, clear and on-brand drafts of support responses
- Ability to adapt the same solution to different customer emotions and channels
- Ability to apply a style guide that balances tone, clarity and brevity
In customer service, how you say it is as important as what you say. The same fact ("your order will be delayed by 3 days") turns into a complaint when given in a cold and bureaucratic language, but turns into understanding when given in an empathetic and solution-oriented language. The support representative writes 60-80 replies per day; It is tiring to set up each one from scratch, with the right tone and without errors. Artificial intelligence (AI) is a speed and consistency multiplier right here: it gives the agent an empathetic, clear and on-brand outline; the agent reviews it and sends it off.
In this unit you will learn how to adapt a single solution to different customer emotions, different levels of formality and different channels; You'll learn how to anchor the tone in your brand's voice. The key idea is this: information (what) remains constant, tone (how) varies depending on the customer.
Note: The AI draft is a start, not the final text to be submitted. In particular, responses containing apology, compensation and commitment should not be sent without being read and approved by the representative.
Anatomy of the Empathic Response
A good support response usually follows this order:
- Acceptance/Empathy: Recognizing the customer's situation and emotion ("I understand the delay you are experiencing, you are right").
- Clarity: Saying clearly what happened and what to do.
- Action/Next step: Concrete solution or next step (“We expedited your shipment, you will receive it tomorrow”).
- Closing: Availability and a brief well-wishing.
This skeleton balances empathy with clarity. Only empathy (“we are so sorry” without resolution) is empty; only clarity ("that's politics") without empathy is cold.
Step by Step: Tone Management
- Define your brand voice. Warm, formal, simple? Make it concrete with a few example sentences.
- Separate the core information. Let the facts that will not change (date, amount, step) be clear.
- Read the emotion. Is the customer angry, concerned, or neutral?
- Adapt the tone to the emotion. Give more acceptance to the angry, more reassurance to the anxious.
- Format by channel. The email can be long; The conversation should be short and fragmented.
- Review. Read the draft; Remove any exaggerated, fraudulent or unauthorized statements.
Copiable Prompts
Basic response draft prompt adapted to the situation:
Role: You are the writing assistant of the "Aydın Elektronik" support team. Tone: warm, simple, solution-oriented; short sentences. Draft a response based on the information below. Follow this structure: (1) empathy/acceptance, (2) clear explanation, (3) concrete next step, (4) brief closing. Don't make promises (discounts, exact dates) beyond your authority. Situation: {{ customer_issue }}Facts we have: {{ known_information }}Customer's emotion: {{ angry / anxious / neutral }}
Prompt adapting the same solution to three different emotions:
Without changing the core information below, generate three separate responses: A) For the angry customer (more acceptance and assuming responsibility), B) For the anxious customer (more assurance and clarity), C) For the neutral customer (brief and direct). Core information: {{ date/amount/step }}
Prompt that humanizes a too formal/robotic text:
Make the following response more human and warm; Translate bureaucratic phrases ("you will be informed") into plain Turkish. Meaning andpreserve numerical information, do not extend. Text: {{ draft }}
Prompt that reviews the response and acts as a security filter:
Check this response draft before it is sent. Check the following:- Is there an over-promise (discount, exact delivery date, money-back guarantee)?- Is there unverified/fabricated information?- Is there an overly defensive or accusatory tone?- Is there a lack of empathy? Suggest a one-sentence correction for each problem. Draft: {{ draft }}
Weak Prompt / Strong Prompt
Weak prompt
Powerful prompt
"Reply to this customer"
Brand tone + empathy-clarity-action framework + authority boundary
Generic, toneless, sometimes cold text
Emotionally responsive, on-brand draft
Can promise beyond authority
Maintains commitment limit
Ignores emotion
First recognizes the emotion, then resolves it
The difference is that the strong prompt describes both tone, structure and boundary at the same time. Thus, the draft is both warm and safe.
Three Mini Cases
Case 1 — Same reality, two worlds. “My item was 2 days late,” one customer wrote. First draft: "Your order has been delayed due to logistics congestion." The second draft produced with the empathy prompt: "It's annoying that the product you expected is delayed, I'm sorry. We have prioritized your cargo; you will receive it tomorrow at the latest, your tracking number is: ...". The customer who received the second response thanked him and the survey score was 5/5.
Case 2 — Acceptance of anger. "I wrote it three times, it's still not resolved!" message, the three-emotion prompt produced the "angry" version: first he took responsibility with the sentence "It's our fault you had to write so many times, I'm sorry for that", and then he gave the clear solution. This opening, which did not place the blame on the customer, defused an escalating conversation.
Case 3 — The security filter worked. “We will definitely solve your problem today and define a 20% discount,” a representative’s AI sketch said. The audit prompt flagged these two statements as "commitment beyond authority". The representative removed “today” and “20%” and replaced it with “we will resolve it and get back to you as soon as possible”; He was saved from making a promise he couldn't keep.
Tip: Include your brand's list of "forbidden words and phrases" in the prompt: expressions such as "absolutely", "never a problem", "I guarantee" are invitations to broken promises. Instead, suggest controllable statements like "we're doing our best," "we've taken this step," "we'll get back to you by this date."
Balance of Brevity and Clarity
Empathy does not mean length. The customer wants his problem solved; The emotional three-paragraph introduction backfires because it delays the resolution. Good outline: one sentence empathy, one or two sentences clear information, one sentence next step. Giving the model a concrete limit, such as “no more than 5-6 sentences” or “no more than 90 words,” improves both readability and the agent's review speed.
Caution: The excessive empathy produced by AI sometimes comes across as fake (“I feel your pain wholeheartedly”). The customer senses the exaggeration and trust is lost. Keep the tone cordial but measured; A real apology is more effective than ten fancy sentences.
Breaking Negative News Well
The hardest moment in support is saying “no”: no refunds, item out of stock, request out of policy. There is a technique to delivering bad news, and AI can apply it consistently. The formula is three steps: (1) give the news clearly and honestly, do not offer fraudulent hope; (2) explain why briefly, without being accusatory; (3) be sure to offer an alternative (“we cannot supply this product, but this similar model is in stock today”). People accept "no" much more easily when a door is left open. Instructing the AI to “give the negative news but always suggest at least one alternative or next step” makes the disappointment manageable. Delaying bad news or leaving it vague ("let's see, maybe") is the worst; The customer both sets expectations and ends up being more disappointed. Clarity is kindness in the long run.
Common mistakes
- Not defining the brand tone and each response coming out with a different voice.
- Just giving empathy and skipping the concrete solution/next step (empty courtesy).
- Just giving information and skipping empathy (coldness).
- Leaving commitments that exceed authority (discount, final date, guarantee) in the draft.
- Sending the draft without reviewing it, with excessive/fake empathy.
- Pasting long email text into the chat without considering the channel.
In summary
- A good response has four steps: empathy/acceptance, clear explanation, concrete next step, brief closure.
- The core knowledge (what) remains constant; The tone (how) is adapted to the customer's emotion.
- Give acceptance and responsibility to the angry, reassurance and clarity to the anxious.
- Clean up over-authorization commitments with an audit step; Don't make promises you can't keep.
- Empathy must be balanced with brevity; A measured and sincere tone is more effective than exaggeration.
Application task
Choose a real customer complaint (or example) and write the core information (what happened, what to do) in one sentence. Then produce angry, anxious, and neutral versions with the three-emotion prompt. Finally, run each version through the audit prompt and remove any intrusive language, false empathy, or accusatory tone. Write down, in one sentence, which version you can send to a real customer and why.
checklist
- [ ] I defined the brand tone in concrete sentences.
- [ ] I applied the empathy-clarity-action-closure framework.
- [ ] I adapted the tone to the emotion without distorting the core information.
- [ ] I cleared the over-authorization commits with the audit step.
- [ ] I chose the appropriate length and format for the channel.
- [ ] I weeded out excessive/fake empathy and kept it sincere and measured.