{
 "byKey": {
  "cover": [
   {
    "say": "Welcome back to Practical AI for Business.",
    "item": 0
   },
   {
    "say": "In this chapter, we're tackling sales conversion and smart customer service.",
    "item": 0
   },
   {
    "say": "You'll build a support bot that answers from your own documents, and then learn how to write cold email that is personal and responsible.",
    "item": 0
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#1": [
   {
    "say": "Let's start with episode seven, building a product support bot with a knowledge base.",
    "item": 0
   },
   {
    "say": "The short video opens with a hook about customer questions arriving at midnight, with nobody there to answer.",
    "item": 1
   },
   {
    "say": "The pain point is familiar: the same questions about shipping, delivery time, warranty, and sizing, every single day.",
    "item": 2
   },
   {
    "say": "Messages sent after hours wait until tomorrow.",
    "item": 2
   },
   {
    "say": "In the hands-on part, you import a product FAQ and an after-sales manual into a knowledge base and build the bot.",
    "item": 3
   },
   {
    "say": "Then a customer asks whether something can go in the oven, and the bot answers from the manual and cites the passage.",
    "item": 3
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#2": [
   {
    "say": "Here's the second half of the video plan.",
    "item": 0
   },
   {
    "say": "You close by recapping the value: common questions get answered automatically, and complex ones go to a human.",
    "item": 1
   },
   {
    "say": "The call to action is to comment for the support-bot setup checklist.",
    "item": 1
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#3": [
   {
    "say": "Now let's correct the claim in the hook.",
    "item": 0
   },
   {
    "say": "Saying the bot knows the product better than a new hire is only true if your knowledge base is complete, accurate, and up to date.",
    "item": 1
   },
   {
    "say": "When data is missing, an AI support agent can be wrong, miss the point, or confidently make things up.",
    "item": 1
   },
   {
    "say": "The video also says eighty percent of common questions are repeats, but that is a scenario assumption, not a universal number.",
    "item": 1
   },
   {
    "say": "Estimate your own ratio from your own support logs.",
    "item": 1
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#4": [
   {
    "say": "Next, who is this for?",
    "item": 0
   },
   {
    "say": "It fits e-commerce, courses, appointment-based services, and B to B products that get lots of repetitive inquiries.",
    "item": 1
   },
   {
    "say": "It's a poor fit for final answers on high-risk decisions, like medical, legal, or refund disputes.",
    "item": 1
   },
   {
    "say": "Those must always go to a human.",
    "item": 1
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#5": [
   {
    "say": "The core concept here is RAG, short for retrieval-augmented generation.",
    "item": 0
   },
   {
    "say": "Here's the idea: the system first finds the relevant passages in your documents, then has the AI answer based on those passages.",
    "item": 1
   },
   {
    "say": "That keeps answers grounded, and updating is easy, because you just edit the document instead of retraining a model.",
    "item": 1
   },
   {
    "say": "Dify is one platform for building RAG apps, but other knowledge-base chatbot tools and Google's document and search integrations work too.",
    "item": 1
   },
   {
    "say": "The flow is the same everywhere: documents, chunking, retrieval, and an answer with a cited source.",
    "item": 1
   },
   {
    "say": "Features, plans, and data handling vary by version and region, so check the official documentation.",
    "item": 1
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#6": [
   {
    "say": "Let's walk through the workflow.",
    "item": 0
   },
   {
    "say": "First, the input: a cleaned-up FAQ, product specs, shipping and return policies, and the after-sales manual.",
    "item": 1
   },
   {
    "say": "If you use PDFs, confirm that the text can actually be copied.",
    "item": 1
   },
   {
    "say": "Then let the AI answer according to your rules, cite its sources, and hand off to a human when it's unsure.",
    "item": 2
   },
   {
    "say": "Human checks matter most. Verify answers one by one with a test question set.",
    "item": 3
   },
   {
    "say": "Every week, review real conversations where the bot answered wrong or couldn't answer, and add that content to the knowledge base.",
    "item": 3
   },
   {
    "say": "After launch, set up a human handoff process, an after-hours message inbox, and a named owner.",
    "item": 4
   }
  ],
  "EP07 | Build a Product Support Bot with a Knowledge Base#7": [
   {
    "say": "Finally, some principles for organizing your knowledge base.",
    "item": 0
   },
   {
    "say": "Let one document cover one topic, with a clear title.",
    "item": 1
   },
   {
    "say": "Write the exceptions down, for example, that shipping to Alaska and Hawaii is charged separately.",
    "item": 2
   },
   {
    "say": "Mark every document with a version and an update date, and take down anything expired.",
    "item": 3
   },
   {
    "say": "One more tip: fast-changing information like prices and inventory should be pulled live from your systems, not hard-coded in documents.",
    "item": 3
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#1": [
   {
    "say": "Now on to episode eight, personalized outreach and responsible cold email.",
    "item": 0
   },
   {
    "say": "The hook asks: sent a hundred cold emails and got zero replies? The reason is that you're blasting the same generic ad to everyone.",
    "item": 1
   },
   {
    "say": "The pain point is reps copying and pasting the same email, getting names wrong, and having mass sends flagged as spam.",
    "item": 2
   },
   {
    "say": "In the hands-on part, you collect a prospect's public company news and ask AI to write one relevant icebreaker opening line.",
    "item": 3
   },
   {
    "say": "After that comes a short value statement.",
    "item": 3
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#2": [
   {
    "say": "Here's the closing part of the video plan.",
    "item": 0
   },
   {
    "say": "The recap is simple: send fewer, send better, and send with sincerity.",
    "item": 1
   },
   {
    "say": "The call to action is to comment for the outreach email template.",
    "item": 1
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#3": [
   {
    "say": "Let's correct the claim in this hook too.",
    "item": 0
   },
   {
    "say": "Triple your reply rate is marketing language, and no number is guaranteed to apply to every company.",
    "item": 1
   },
   {
    "say": "Reply rates depend heavily on industry, list quality, product, and timing.",
    "item": 1
   },
   {
    "say": "A sensible approach is to run your own small A B test.",
    "item": 1
   },
   {
    "say": "Try fifty generic emails versus fifty personalized ones, and compare using your own data.",
    "item": 1
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#4": [
   {
    "say": "So who should use this?",
    "item": 0
   },
   {
    "say": "It's a good fit for B to B sales, distributor recruitment, and small companies that need to reach prospects proactively.",
    "item": 1
   },
   {
    "say": "It's a poor fit if you have no clear target audience and just want to blast out mass emails.",
    "item": 1
   },
   {
    "say": "That's a problem for both results and compliance.",
    "item": 1
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#5": [
   {
    "say": "Let's look at the core tool concepts.",
    "item": 0
   },
   {
    "say": "The outline mentions Clay, Perplexity, and ChatGPT. Clay leans toward organizing prospect data and automated enrichment.",
    "item": 1
   },
   {
    "say": "Perplexity can help you search public information with sources, and ChatGPT or another large language model writes the emails.",
    "item": 1
   },
   {
    "say": "You can also use company websites, press releases, trade media, public job postings, and Google search as sources.",
    "item": 1
   },
   {
    "say": "The concept stays the same: first find what just happened to them, then connect it to your value.",
    "item": 1
   },
   {
    "say": "Features and plans vary by version and region, so check the official documentation.",
    "item": 1
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#6": [
   {
    "say": "Before you learn to write the email, let's settle compliance and ethics.",
    "item": 0
   },
   {
    "say": "First, personal data. Use only information that is lawful, public, and relevant to your business purpose, and know where it came from.",
    "item": 1
   },
   {
    "say": "Privacy laws like CCPA and CPRA in California, and GDPR if you serve EU customers, may apply.",
    "item": 1
   },
   {
    "say": "Rules differ, so check with counsel. This is not legal advice.",
    "item": 1
   },
   {
    "say": "Second, don't bypass platform rules. Don't use scrapers to pull users' data in bulk, and don't buy lists of unknown origin.",
    "item": 2
   },
   {
    "say": "Third, anti-spam basics. Identify yourself clearly, offer a way to opt out, honor opt-outs, and never email people who said no.",
    "item": 3
   },
   {
    "say": "Rules like CAN-SPAM in the US are a starting point, but they differ by country, so confirm before sending across borders.",
    "item": 3
   }
  ],
  "EP08 | Personalized Outreach: Responsible Cold Email#7": [
   {
    "say": "Two more principles on the same topic.",
    "item": 0
   },
   {
    "say": "Keep your content truthful: don't impersonate anyone, and don't claim you've met or share mutual friends when you haven't.",
    "item": 1
   },
   {
    "say": "And keep volume in check. Send few, relevant, valuable emails, not as many as possible.",
    "item": 2
   }
  ],
  "closing": [
   {
    "say": "That wraps up this chapter.",
    "item": 0
   },
   {
    "say": "You built the idea of a support bot grounded in your own documents, and you learned to do outreach that is personal and respectful.",
    "item": 1
   },
   {
    "say": "Please do the lab and then take the chapter quiz to check how much you've learned.",
    "item": 1,
    "pause": 1
   }
  ]
 }
}