Handle Difficult Client Conversations with AI Coaching

1. Open https://claude.ai and start a **new chat** (keep everything in this single thread through Order 4). 2. Paste this prompt and fill in the brackets: "I have one specific difficult client conversation coming up. Help me prep — but do NOT give me wording yet. First mirror back what I'm telling you and flag missing context. Client (role + relationship length): [...] Situation in 3 lines (what happened, what they last said, what changed): [...] What I want (ideal outcome): [...] What I'd accept (acceptable outcome, still a yes): [...] My walk-away — what I will NOT accept under any framing (BATNA): [...] Channel: [live call / video / Loom / async written] Time pressure: [today / this week / no rush] Reply with: (a) a 1-paragraph mirror of the above in your words, (b) up to 3 missing facts you'd want before scripting, (c) NO wording yet." 3. Read Claude's mirror. If it softens your walk-away or invents context, reply: "Re-mirror with my exact walk-away and don't add facts I didn't give you." 4. Save the **walk-away line** somewhere you can glance at during the call — sticky note, top of doc, lockscreen.

1. In the **same chat**, paste: "Based on the situation I just described, do two things: 1) Classify this into ONE of: - scope-creep pushback - late payment / collections - deliverable rejection or rework - price renegotiation (theirs or mine) - refund request - missed deadline (mine) - relationship wind-down / firing the client 2) Recommend ONE stance for me to hold throughout: - firm-and-warm (default for ongoing relationships) - curious-and-direct (when facts are unclear) - professional-cold (when trust is broken or wind-down) Justify both in 2 sentences. If my situation actually contains 2 conversations bundled together, name them and tell me which to handle first." 2. If Claude bundles two conversations, **pick one** for this script — note the second one separately for later. 3. Write down the chosen **type** and **stance** as one line — you will paste them by name into Order 3 and Order 4.

1. In the **same chat**, paste (fill in the bracketed values from Orders 1 and 2): "Using the [stance from step 2] stance, write a one-page script for a [channel from step 1]: A) OPENER (max 3 sentences if live, max 5 sentences if async written) — no apology for things I didn't do, no 'just checking in,' no thanking them for 'patience' if I'm the one being patient. B) 3 TALKING POINTS — short, in my voice, each ending in a clean stop (no trailing questions that invite renegotiation). C) ONE ASK — a single concrete decision, date, amount, or action I want from them by end of the conversation. State it as a sentence I will actually say. D) WHAT I WILL NOT VOLUNTEER — list 2-3 things I should NOT bring up unprompted (discounts, extra scope, future favors, personal context) because they weaken the ask. Voice: like me, not customer-success. Short sentences. No 'I just wanted to…', no 'Sorry to bother…', no 'Totally understand if…'." 2. Read the opener **out loud**. Rewrite any line you wouldn't actually say. 3. Confirm the **ask** is one sentence and contains a number, a date, or a binary decision — not a feeling.

1. In the **same chat**, paste: "Switch roles: YOU are now the client. Push back on my ask the way this specific client most likely will. Run THREE objections in a row, one at a time: 1) The emotional objection (frustration, 'I thought we were past this', 'this feels personal') 2) The financial / contractual objection ('the contract says…', 'we already paid for…', 'this is out of scope for what we agreed') 3) The leverage objection (urgency, mentioning a competitor, hinting at ending the engagement, threatening a bad review) For each: after you push back as the client, switch back to coach role and propose MY reply in [stance from step 2]. Reply must NOT lower my ask, must NOT apologize for things outside my control, and must be one sentence I would actually say. Then propose ONE fallback position — what I would accept if they refuse the ask but I don't want to walk yet — and confirm it is ABOVE the walk-away from Order 1. If you can't find a fallback above the walk-away, say so explicitly so I know the answer is to walk." 2. Rewrite any reply that sounds like you'd cave. If a reply starts with "I completely understand…" or quietly drops the ask, kill it. 3. Verify the fallback is above your walk-away. If not, write **WALK** at the bottom of your script and rehearse the polite exit line instead.

1. Open https://docs.google.com (or Apple Notes / Notion — pick one and stick to it). Create a new doc titled `[Client] — [Topic] script — [DATE]`. 2. Paste the script content from your Claude chat in this exact order: - **Opener** - **3 talking points** - **The ask** - **Objection replies (1 / 2 / 3)** - **Fallback position** - **Walk-away** (keep it visible on the page during the call) 3. Switch back to Claude **one last time** and paste: "Pre-draft a 4-line summary email I'll send within 30 minutes after the call. Subject + body. Must include: (1) what was decided, (2) the next step + owner, (3) the date or amount tied to it, (4) one sentence putting the agreed boundary in writing. Tone: [stance from step 2]. No softeners. No reopen language like 'happy to discuss further.' Also write a 2-line VERSION-B for the case where they refused the ask and I accepted the fallback position. Same structure." 4. Paste both versions into the doc under headings **Summary email — agreed** and **Summary email — fallback**. 5. Close every other tab. The doc is the only thing you need open during the call.