AI for customer success
CS lives between sales and support, with renewal pressure and human relationships on both sides. AI helps with the scaffolding (call notes, QBR prep, onboarding plans, handoff briefs) so you spend time on what only you can do: actually listening to your customers and figuring out if they're going to renew.
~/.claude/skills/What AI handles well
QBR prep that's actually about the customer
The problem. QBR is in 3 days. Generic templates produce decks the customer skims. You want to bring real insights and a clear ask, not a status update.
What AI does. Use AI to synthesize the last 6 months: their usage patterns, support tickets, calls, what changed in their business. Then have it draft a QBR structured around: what worked, what didn't, what we're proposing for next quarter, the ask.
Renewal risk assessment
The problem. You have 30 accounts up for renewal in Q4. You don't have time to deep-dive each. The ones that look healthy might be the ones about to churn.
What AI does. Have AI score risk based on signals you provide: usage trends, executive turnover, support ticket sentiment, last QBR outcomes. Output: ranked list of which accounts need a conversation NOW vs which can run on autopilot.
Difficult conversations with churning customers
The problem. A customer signaled they might not renew. You need to have a hard conversation: what's actually wrong, can it be fixed, what would change their mind. Generic "I value our partnership" emails won't cut it.
What AI does. Use the tough conversation prep prompt. AI helps you walk in with the real question (not euphemisms), the specific examples, and how to avoid getting defensive when they push back.
Apology emails after incidents (without sounding corporate)
The problem. Outage. Bug. Missed commitment. You need to apologize without sounding like the customer-service handbook. "We sincerely apologize for any inconvenience" is the death of trust.
What AI does. Use the apology email prompt. It owns the issue plainly, explains impact specifically, says what we're doing about it (concrete, not vague), and avoids the tells of canned responses.
Onboarding plans tailored to customer maturity
The problem. Every customer gets the same 30-60-90 onboarding plan. Half the steps don't apply. The customer feels like a number.
What AI does. Have AI tailor the plan based on customer inputs: company size, technical maturity, primary use case, who their internal champion is. Output: a real plan, not a template.
Internal handoff briefs (Sales → CS, CS → Support)
The problem. A new customer just signed. Sales has all the context — their goals, their objections, who the players are. By the time it gets to you, half is lost in CRM notes.
What AI does. Use AI to structure handoff briefs from raw sales notes + call transcripts. Output: a one-pager covering customer's goals, the deal context, the people, the risks, what they expect in the first 90 days.
Educational content drafts for your customers
The problem. You promised a customer best-practices guide, a runbook, a how-to article. It's been three weeks. You're still "working on it."
What AI does. AI drafts educational content from your existing knowledge — what you'd say in a call, what other customers have asked, the FAQs. You edit for voice and ship in 90 minutes instead of three weeks.
Email triage when 80 customers email at once
The problem. New release ships. Inbox fills up. Some emails need real responses. Some are reactions you can ignore. Which is which?
What AI does. Use the email triage prompt. AI sorts inbox into respond-now / quick-reply / schedule-deep-work / delegate / archive. You execute on the sorted list.
Your AI stack
Start with the foundation. Add specialized tools as the work calls for them.
Foundation LLM
Specialized add-ons
Prompts ready to use
Get started in 30 minutes
Set up AI meeting notes for every customer call
10 minGranola, Otter, Fireflies — pick one your team approves. Stop typing during calls. Listen instead. AI handles the notes.
Build a Claude Project per top-10 account
15 minLoad each account's context: their goals, contract, key people, history. Any prompt about that account is now pre-loaded with relevant context.
Run the renewal risk prompt on your current pipeline
5 minOutput a ranked list. Schedule the top 3 risk conversations this week. The point isn't to predict perfectly — it's to surface accounts you've been on autopilot with.
Common mistakes
AI-generated QBRs that don't reference customer specifics. The customer can tell. A generic QBR says "you matter as much as a template."
Using AI to avoid the hard conversation. AI helps you PREPARE for it, not skip it. Sending a polished AI-drafted "we should talk" email instead of having the call IS the avoidance.
Trusting AI summary without spot-checking source. AI compresses. Compression loses the angry customer's exact wording — the wording that signals they're actually leaving.
Pasting customer data (names, contracts, internal struggles) into public AI. Use your company's approved tier. If you don't have one, anonymize.
Optimizing for response speed at cost of relationship. AI lets you reply faster. The customer doesn't want faster — they want to feel heard. Slow down where it matters.