💡 AI 模型在英文提示詞下表現最佳。因此,提示詞本文以英文呈現。使用英文輸入可獲得更準確、更詳細的回應。 處理問題的方式比處理成功的方式更能定義你的品牌。這些提示詞幫助你撰寫能降低挫折感的回覆、找出反覆出現問題的根本原因,以及將憤怒客戶轉化為你最大的品牌擁護者。已在 GPT-4.1、Gemini 2.5 Pro、Claude Sonnet 4 和 Grok 3 上測試,讓你知道哪個模型最能冷靜應對壓力。
提示詞
安撫憤怒客戶並化解衝突
Help me de-escalate this customer situation: What happened: [describe the issue] Customer's emotional state: [angry / frustrated / disappointed / threatening to leave] What they said: [paste their exact message or describe the call] What we can actually do: [your constraints and options] Relationship history: [new customer / long-time customer / VIP] Provide: 1. An opening response that validates their emotion without admitting fault prematurely 2. A bridge statement that transitions from empathy to problem-solving 3. Three resolution options to offer (good / better / best) 4. Language to avoid (phrases that make angry customers angrier) 5. A follow-up message to send 24 hours after resolution 6. An internal escalation template if I can't resolve it myself
進階技巧
配合客戶的溝通管道和情緒強度。如果他們寫了一封 500 字的投訴,不要用 2 句話回覆。回覆的篇幅代表你認真看待了他們的問題。AI 可以幫你讓回覆的份量匹配他們挫折感的份量。
已測試 Mar 15, 2026
建立常見投訴的回應範本
Build a complaint response library for [business type]. Top 10 complaints we receive: [List each complaint category with approximate frequency] Brand voice: [professional / friendly / casual] Response time goal: [hours/minutes] Channels: [email, chat, phone, social media] For each complaint, create: 1. An empathy-first response template (customizable with [brackets]) 2. Resolution steps the agent should take 3. Compensation guidelines: when to offer what 4. Escalation criteria: when this needs a manager 5. A prevention note: what to flag for the product/ops team 6. A 'win-back' follow-up message for after the issue is resolved
進階技巧
讓你最新的客服人員審閱每個範本。如果他們需要重寫才能聽起來像自己,那範本就太僵化了。好的範本是框架,不是劇本——客服人員應該用自己的語氣填充,同時遵循結構。
已測試 Mar 15, 2026
找出反覆出現的客戶問題的根本原因
Help me find the root cause of this recurring customer issue: The symptom: [what customers report] Frequency: [how often it happens] Affected segment: [which customers experience it] Timeline: [when it started or got worse] What we've tried: [fixes attempted so far] Data available: [logs, tickets, metrics you have] Conduct a root cause analysis: 1. Five 'why' analysis: drill from symptom to root cause 2. Contributing factors: what makes this issue worse or more frequent 3. A fishbone diagram structure (categories: people, process, product, environment) 4. The most likely root cause with evidence reasoning 5. Three fix options: band-aid, medium-term, permanent solution 6. A measurement plan: how to verify the fix actually worked
進階技巧
永遠比感覺舒適的次數多問一次「為什麼」。大多數團隊在第二或第三個「為什麼」就停下,修復的是症狀而非原因。真正的根本原因通常令人不安地系統性——這就是你知道已經挖得夠深的標誌。
已測試 Mar 15, 2026
建立清晰的升級處理流程
Help me build an escalation framework for our support team. Team structure: [tiers, roles, managers] Ticket volume: [daily/weekly] Current escalation process: [what exists now, or nothing] SLA targets: [response and resolution times] High-risk scenarios: [situations that need immediate attention] Design an escalation playbook: 1. Escalation tiers: what gets handled at each level 2. Trigger criteria: specific signals that require escalation 3. Response templates for each escalation level 4. Communication protocol: who gets notified and how 5. SLA adjustments: how timelines change at each tier 6. Post-escalation review: learning from every escalated case
進階技巧
為提到競爭對手、法律行動或公開在社群媒體上發言的客戶建立「快速通道」。這些情況對品牌影響極大,需要與一般升級不同的處理方式——此時速度和資深程度比什麼都重要。
已測試 Mar 15, 2026
撰寫真誠的道歉信以重建信任
I need to apologize to a customer (or group of customers) for [describe the mistake/issue]. What went wrong: [be specific about the failure] Impact on customers: [how it affected them] Our fault level: [fully our fault / partially / external factor] What we're doing to fix it: [concrete steps taken] Compensation available: [what you can offer] Write: 1. A direct apology that takes ownership without corporate hedging 2. A clear explanation of what happened (honest, not defensive) 3. Specific steps we're taking so it won't happen again 4. Compensation or goodwill gesture with proper framing 5. A mass email version (if this affects many customers) 6. A personal version for high-value customers with extra care
進階技巧
「但是」這個詞會摧毀每一個道歉。「我們很抱歉,但是這是發生的原因」聽起來像藉口。把每個「但是」換成「同時」——「我們很抱歉,同時這是我們正在做的」。這一個詞的改變,將防禦性的解釋轉變為負責任的態度。
已測試 Mar 15, 2026
將投訴轉化為長期忠誠
A customer had a bad experience and I want to turn them into an advocate. Original issue: [what went wrong] How we resolved it: [what we did to fix it] Customer's current sentiment: [still upset / neutral / pleasantly surprised] Customer value: [lifetime value or tier] Relationship length: [how long they've been a customer] Build a service recovery plan: 1. A surprise follow-up gesture 1 week after resolution 2. A personalized check-in message at 30 days 3. An invitation to provide input on preventing similar issues 4. A loyalty program or exclusive offer as a genuine thank-you 5. A referral opportunity that benefits them (not just you) 6. A long-term relationship plan: quarterly touches that show you remember
進階技巧
經歷過問題但被出色解決的客戶,忠誠度比從未遇過問題的客戶還高。這就是「服務恢復悖論」——有意識地運用它,在恢復過程中超額交付。
已測試 Mar 15, 2026
基於實際測試結果 — 非假設推測。 查看測試方法
Claude Sonnet 4
深度解讀情緒脈絡,撰寫的回覆真正有人情味。在降低衝突腳本和不帶企業套話的真誠道歉方面最強。
最佳降溫能力GPT-4.1
撰寫最自然的投訴回應資料庫和服務恢復計畫。在建立客服人員能快速個人化、不顯得機械化的範本方面最強。
最佳範本撰寫Gemini 2.5 Pro
產出最有邏輯的根本原因分析和升級處理框架。在客服團隊能立即執行的結構化系統方面最強。
最佳根因分析Grok 3
提供快速、實用的解決方案,不會過度複雜化。在速度比儀式更重要的直接問題解決場景中最強。
最快速解決先解決情緒,再解決問題。憤怒的客戶在感到被傾聽之前,聽不進你的解決方案。用前 2-3 句話純粹認同他們的挫折感,然後再轉向解決問題。
衡量首次聯繫解決率,而非工單數量。一個團隊關了 100 張工單但重開了 30 張,不如一個關了 70 張只重開 5 張的團隊有效。追蹤客戶多常帶著同樣問題回來。
為未來的客服人員記錄每一個解決方案。每個已解決的問題都是你團隊的訓練資料。建立可搜尋的歷史解決方案知識庫,讓客服人員不用重新發明已經解決過的問題的方案。