Full scripts. Read them, or play them when the video version drops.
Lesson 14 min
The 60-80% rule
Audit your last 50 customer-service interactions. I guarantee 60-80% of them are repeat questions you've answered dozens of times: 'where's my order?', 'do you accept [thing]?', 'how do I [common action]?', 'what's your return policy?'. These are exactly the questions AI handles perfectly — they have predictable answers from your existing FAQ. Today: we install an AI customer-service layer that auto-resolves the 60-80%, leaving only the genuinely tricky cases for you. Time saved: 5-15 hours per week for most small businesses. Quality of customer experience: actually IMPROVES, because AI replies in 30 seconds instead of you replying in 24 hours.
Key points
- →60-80% of support is repeat questions with FAQ answers
- →AI handles those instantly, frees humans for tricky cases
- →Time saved: 5-15 hours/week typical
- →Customer experience improves (faster + consistent)
Lesson 210 min
Build your support knowledge base
Before installing the tool, gather your knowledge. Spend 30 minutes pulling together: 1) Your top 20 repeat questions and the answers, 2) Your return/refund policy, 3) Shipping or service-delivery policies, 4) FAQ from your website (if any), 5) Tone guidelines — how YOU answer customers (formal? casual? funny?). Put it all in one Google Doc or Notion page. Don't be lazy — be specific. Don't just write 'we accept returns within 30 days.' Write 'we accept returns within 30 days for unused items in original packaging. Customer pays return shipping unless the item was damaged in transit. Refunds are issued within 5 business days of receipt.' AI needs specifics. Vague = vague answers customers won't accept. This step takes 30-60 minutes. Worth every minute — it pays back every day forever.
Key points
- →Spend 30-60 min gathering: top 20 questions + answers, policies, tone
- →Be specific — vague knowledge = vague AI answers
- →Single source of truth: one Google Doc or Notion page
- →This step is the highest-leverage 60 min in the course
Lesson 310 min
Intercom Fin — the polished choice
Intercom Fin is the best AI customer-service tool in 2026. Why: high resolution rate (60-80% of conversations fully resolved without human handoff), excellent fallback to humans for complex cases, integrates with your existing helpdesk (Zendesk, Help Scout, Intercom). Pricing is per-resolution ($0.99/resolution) which sounds expensive until you realize it replaces multiple support-hour costs. Setup: sign up, upload your knowledge base (the doc you built in lesson 2), connect to your support channel (email, chat, both), set the handoff rules ('escalate to human if customer says X, Y, Z'). Test 10 sample questions. Iterate where it fails. Live within an hour.
Key points
- →Intercom Fin: $0.99/resolution, 60-80% auto-resolve
- →Connects to: Zendesk, Help Scout, Intercom, Front
- →Setup: upload KB, connect channels, set handoff rules
- →Test 10 questions before going live
Lesson 48 min
Help Scout AI — the budget-friendly alternative
If Intercom Fin's per-resolution cost is unpredictable for your volume, Help Scout's built-in AI is the alternative. $50/mo per user includes AI drafting + categorization + summarization. AI doesn't auto-send — it drafts replies that you approve in one click. Slightly less automated than Fin (you're still in the loop), but cheaper and more predictable. For businesses doing <50 tickets/day, Help Scout often costs 30-50% less than Fin. Setup: same knowledge base, different tool. The platform's docs walk you through training in 20 minutes.
Key points
- →Help Scout AI: $50/mo per user, predictable
- →Drafts replies; you approve in one click (slower than Fin but cheaper)
- →Better for <50 tickets/day
- →Same KB upload, different platform
Lesson 58 min
Handoff rules — when AI gives up
Bad handoffs are where AI customer-service breaks. The AI tries to handle something it shouldn't, frustrates the customer, AND then dumps them on you with an angry message. Configure handoff rules carefully. Hand off to a human when: 1) Customer explicitly asks to speak to a person, 2) Tone shifts to angry / frustrated (most platforms detect this), 3) Question involves a refund or compensation, 4) Question involves anything not in your knowledge base. The AI's last message before handoff should NEVER be 'I can't help with that.' It should be 'Let me get [name] to look at this for you — they'll respond in [timeframe]. In the meantime, anything else I can help with?' Warm handoffs preserve customer trust. Cold handoffs torch it.
Key points
- →Bad handoffs make customers angrier than no AI at all
- →Hand off when: customer asks, tone shifts angry, refund/comp, outside KB
- →Warm handoff language: 'let me get [name] to look at this'
- →Never 'I can't help with that' — always 'human is coming'
Lesson 65 min
Review responses (Google/Yelp)
Bonus layer: AI responses to Google and Yelp reviews. Most small businesses leave reviews unanswered. Big mistake — responding to reviews boosts your local SEO ranking AND signals to future customers that you're attentive. Set up: use Birdeye or Podium (paid platforms, $250-500/mo) for full automation, OR a simple Zapier flow ($20/mo): when a new Google review lands → trigger Claude → draft a personalized response → email it to you for one-click approval. Approve, post. 5 minutes per week, your review section looks alive. Worth the investment.
Key points
- →Most small businesses skip review responses — easy SEO win
- →Birdeye/Podium full automation ($250-500/mo) or DIY Zapier ($20/mo)
- →Workflow: AI drafts response → you approve → post
- →5 min/week, boosts local SEO + customer perception