{
 "byKey": {
  "cover": [
   {
    "say": "Welcome back to Business Model Design. This is Chapter Ten, Validation, Risk and Launch, the last chapter of the course.",
    "item": 0
   },
   {
    "say": "We will turn your guesses into evidence, look honestly at the risks of an AI business, and get ready to launch.",
    "item": 0
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#1": [
   {
    "say": "Let's start with a common misunderstanding.",
    "item": 0
   },
   {
    "say": "Many people think the work is done once the canvas is drawn, the value proposition is written, and pricing and costs are worked out.",
    "item": 1
   },
   {
    "say": "But every box on that canvas is really a guess that says, we think it works like this. Until someone actually pays you, it is a well-organized guess.",
    "item": 1
   },
   {
    "say": "This chapter is about turning guesses into evidence, seeing the risks, and pushing your idea into the real world. It takes you from I can design this to I am ready to launch this.",
    "item": 2
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#2": [
   {
    "say": "Here is the core idea. A business model is a stack of assumptions.",
    "item": 0
   },
   {
    "say": "Rewrite each block of the canvas as an if then statement, and it becomes something you can test. Take customers first.",
    "item": 0
   },
   {
    "say": "For example, small shop owners spend six or more hours a month sorting receipts.",
    "item": 1
   },
   {
    "say": "For the value proposition, you might say they will pay twenty-nine dollars a month to get that time back.",
    "item": 2
   },
   {
    "say": "For channels, you might say accounting firms will be willing to refer clients. We will finish the list on the next slide.",
    "item": 3
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#3": [
   {
    "say": "Continuing the list, we have revenue and costs.",
    "item": 0
   },
   {
    "say": "Revenue might be that eighty percent of customers will renew for at least three months.",
    "item": 1
   },
   {
    "say": "Costs might be that variable cost per customer will stay below forty percent of revenue.",
    "item": 2
   },
   {
    "say": "A good assumption has three traits. It is specific, it is falsifiable, and it has a deadline.",
    "item": 2
   },
   {
    "say": "Everyone needs bookkeeping software is not testable. At least eight of twelve shop owners we interview spend six or more hours a month on receipts is.",
    "item": 2
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#4": [
   {
    "say": "Now let's sort assumptions into five types. This is a handy checklist.",
    "item": 0
   },
   {
    "say": "First, customer and problem. Who has the pain, how bad is it, and how do they solve it today?",
    "item": 1
   },
   {
    "say": "Second, value and solution. Does your solution really work, and how is it better than the current alternative?",
    "item": 2
   },
   {
    "say": "Third, revenue. How much will they pay, and is the way they pay workable? The next slide covers the last two.",
    "item": 3
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#5": [
   {
    "say": "The last two types are growth, and feasibility and cost.",
    "item": 0
   },
   {
    "say": "Growth asks how you will find customers, whether the acquisition cost is acceptable, and whether customers will stay.",
    "item": 1
   },
   {
    "say": "Feasibility and cost asks whether you can build it and whether the cost structure holds up.",
    "item": 2
   },
   {
    "say": "For AI products, pay special attention to inference cost, quality and compliance.",
    "item": 2
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#6": [
   {
    "say": "Your assumption list may have twenty items, but you only have the resources to test two or three. So how do you choose?",
    "item": 0
   },
   {
    "say": "Score importance from one to five. A five means that if this is wrong, we lose everything.",
    "item": 1
   },
   {
    "say": "Score uncertainty from one to five. A five means it is pure guesswork.",
    "item": 2
   },
   {
    "say": "Then multiply them to get a risk score, and sort from highest to lowest. If two scores are close, do the cheaper, faster test first.",
    "item": 3
   },
   {
    "say": "Say customers will pay scores importance five and uncertainty four, which is twenty. Cloud servers will be enough scores two times two, which is four. So test willingness to pay first.",
    "item": 3
   }
  ],
  "Principles: Your Business Model Is a Stack of Testable Assumptions#7": [
   {
    "say": "Here is the step people skip most often, and it matters most. Write the threshold first, then run the experiment.",
    "item": 0
   },
   {
    "say": "Before you start, write down what counts as a pass and what counts as a fail.",
    "item": 1
   },
   {
    "say": "For example, you send a pre-sale invitation to fifty people on your waitlist, and at least eight paying is a pass.",
    "item": 1
   },
   {
    "say": "If you set the standard afterward, it is very easy to read a mediocre result as a success, because you love your idea.",
    "item": 1
   },
   {
    "say": "How do you set the threshold? There is no magic number. Work backward from your industry, price and costs.",
    "item": 2
   },
   {
    "say": "If you cannot explain where a threshold comes from, label it a provisional threshold and calibrate it with real data after the experiment.",
    "item": 2
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#1": [
   {
    "say": "Now let's talk about minimum experiments. Why minimum?",
    "item": 0
   },
   {
    "say": "The goal of validation is learning, not building a perfect product. You do only as much work as it takes to answer one assumption.",
    "item": 0
   },
   {
    "say": "When you choose an experiment, ask three questions. First, which assumption am I testing? Pick just one, not five at once.",
    "item": 1
   },
   {
    "say": "Second, what evidence would convince me? That is your pass threshold.",
    "item": 2
   },
   {
    "say": "Third, what is the cheapest, fastest way to get that evidence?",
    "item": 3
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#2": [
   {
    "say": "Method one is customer interviews, which is the cheapest experiment there is. You just talk to real people.",
    "item": 0
   },
   {
    "say": "The key is to ask about past behavior, not future intent. A good question is, when was the last time you sorted your receipts, how long did it take, and what did you do?",
    "item": 1
   },
   {
    "say": "A bad question is, if an app sorted your receipts for you, would you use it? Everyone says yes.",
    "item": 2
   },
   {
    "say": "Interviews tell you whether the problem exists and how painful it is. They cannot tell you whether someone will pay a particular price. That takes behavior.",
    "item": 2
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#3": [
   {
    "say": "Method two is the landing page test. Build a simple page with your value proposition, the price or plan, and one clear call to action.",
    "item": 0
   },
   {
    "say": "Then bring in real target customers with a small number of ads or social posts. Here are the metrics you can watch.",
    "item": 0
   },
   {
    "say": "Click-through rate tells you whether the headline and image grab attention.",
    "item": 1
   },
   {
    "say": "Conversion rate is the share of visitors who leave contact details or click the button.",
    "item": 2
   },
   {
    "say": "Cost tells you how much ad spend each conversion took.",
    "item": 3
   },
   {
    "say": "One thing to watch. Test only one thing at a time, like the wording or the price, or you will not know what made the difference.",
    "item": 4
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#4": [
   {
    "say": "A few more things to watch with landing pages.",
    "item": 0
   },
   {
    "say": "Traffic must come from your real target segment, or the conversion rate means nothing.",
    "item": 1
   },
   {
    "say": "Present the page honestly as a waitlist or not yet available. Do not pretend the product already exists.",
    "item": 2
   },
   {
    "say": "And remember that leaving contact details is a behavior, but it is still a weak commitment.",
    "item": 3
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#5": [
   {
    "say": "Method three is the pre-sale and willingness-to-pay test. This is the step from interested to will pay.",
    "item": 0
   },
   {
    "say": "A pre-order means customers pay early at a discount, with delivery promised after launch.",
    "item": 1
   },
   {
    "say": "A deposit is a small, refundable amount collected up front.",
    "item": 2
   },
   {
    "say": "A letter of intent is a non-binding statement from a business customer that they intend to buy. The next slide has the last option.",
    "item": 3
   },
   {
    "say": "Always be honest about delivery dates and scope, and offer a refund option.",
    "item": 3
   },
   {
    "say": "Also follow the consumer-protection and advertising rules for advance payments where you operate, such as FTC rules in the US. This is not legal advice, so check with counsel.",
    "item": 3
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#6": [
   {
    "say": "The last approach is a paid pilot. You charge a reasonable fee to a few customers for a simplified version.",
    "item": 0
   },
   {
    "say": "Taking money you cannot deliver on destroys trust and can cause legal trouble. So only sell what you can honestly deliver.",
    "item": 1
   }
  ],
  "Concepts: Minimum Experiments — Learn the Most for the Least#7": [
   {
    "say": "Method four is the concept test.",
    "item": 0
   },
   {
    "say": "Before you have a product, show customers a mockup, a sixty-second demo video, or a two-page description, and watch how they react.",
    "item": 1
   },
   {
    "say": "Ask which of their problems it solves, what is missing, and how they handle this today. Compare reactions to two or three different concepts.",
    "item": 1
   },
   {
    "say": "Concept tests are good for checking wording and positioning, but the feedback is mostly verbal opinion. So pair it with behavioral metrics.",
    "item": 1
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#1": [
   {
    "say": "Now let's turn to the risks of an AI business. Principles first.",
    "item": 0
   },
   {
    "say": "AI makes many new businesses possible, but it brings risks that traditional models did not have, or did not feel as sharply.",
    "item": 1
   },
   {
    "say": "This unit covers principles and ways of thinking only. It is not legal advice, and laws differ by country, state and industry, and they change.",
    "item": 1
   },
   {
    "say": "For legal questions, consult a qualified professional and check the latest official guidance. The process mirrors assumption testing: list the risks, rate likelihood and impact, decide on a response, and keep tracking.",
    "item": 2
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#2": [
   {
    "say": "Risk one is model commoditization.",
    "item": 0
   },
   {
    "say": "When many vendors offer models of similar quality, we use AI alone is hard to sustain as a differentiator. Falling prices pressure products that resell model capability.",
    "item": 0
   },
   {
    "say": "So do not bet your moat on the model itself. Ask whether your edge lies in data others cannot easily get, workflow integration, brand and trust, switching costs, channels, or vertical expertise.",
    "item": 1
   },
   {
    "say": "Treat cheaper, better models as an opportunity. Your costs may fall, but so does everyone else's barrier to entry.",
    "item": 2
   },
   {
    "say": "Ask yourself regularly, if a general-purpose tool adds this feature next year, what value does my product still have?",
    "item": 3
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#3": [
   {
    "say": "Risk two is vendor dependence. If your product relies on a single model or platform, their price changes, rule changes or model swaps can hit your cost and quality.",
    "item": 0
   },
   {
    "say": "First, wrap model calls in your own interface layer, so switching vendors does not mean rewriting the whole product.",
    "item": 1
   },
   {
    "say": "Second, keep your own evaluation question set, and run it every time you change a model or version.",
    "item": 2
   },
   {
    "say": "Third, read the terms on data use, price changes, usage limits and service guarantees. Two more principles are on the next slide.",
    "item": 3
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#4": [
   {
    "say": "Two more principles for vendor dependence.",
    "item": 0
   },
   {
    "say": "Keep critical data, like customer data and your accumulated labels and feedback, in your own hands.",
    "item": 1
   },
   {
    "say": "And have a backup plan. At least know the alternatives and the switching cost.",
    "item": 2
   },
   {
    "say": "Remember that prices, features and terms vary by vendor and version, so check the official documentation.",
    "item": 2
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#5": [
   {
    "say": "Risk three is copyright and intellectual property.",
    "item": 0
   },
   {
    "say": "The questions around AI-generated content and training data are still evolving, so do not assume anything AI produces is automatically safe for commercial use.",
    "item": 0
   },
   {
    "say": "Understand what your tools say about licensing and commercial use. The official terms are what count.",
    "item": 1
   },
   {
    "say": "Do not enter prompts that imitate specific artists, brands or protected content.",
    "item": 2
   },
   {
    "say": "For your own core materials, confirm the source and license, and keep records.",
    "item": 3
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#6": [
   {
    "say": "Two more principles on copyright.",
    "item": 0
   },
   {
    "say": "Before you promise customers exclusive ownership or zero infringement risk, make sure you can actually guarantee it.",
    "item": 1
   },
   {
    "say": "And have a professional review high-stakes uses.",
    "item": 2
   }
  ],
  "Concepts: AI Business Risks and Responsible Launch#7": [
   {
    "say": "Risk four is privacy and data protection. AI services often handle customer data, which may include personal information or trade secrets.",
    "item": 0
   },
   {
    "say": "Practice data minimization. Collect and send only what is necessary to do the job.",
    "item": 1
   },
   {
    "say": "Give notice and get consent. Explain in plain language what you collect, what it is used for, whether it may train models, and how long you keep it.",
    "item": 2
   },
   {
    "say": "Understand how your vendor handles the data you send, including training use, retention, and where it is stored.",
    "item": 3
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#1": [
   {
    "say": "Let's see all of this in action with a case study. TallyTap is a fictional startup.",
    "item": 0
   },
   {
    "say": "The company, people and numbers are all made up for learning. The costs are teaching assumptions, so use your own actual quotes. All amounts are in US dollars.",
    "item": 1
   },
   {
    "say": "Maria, a former bookkeeping assistant, and Dev, a software engineer, want to help small shops that struggle with piles of receipts every month.",
    "item": 2
   },
   {
    "say": "TallyTap would use AI to read and categorize receipts, do a first-pass reconciliation, flag doubtful items for a person to review, and charge twenty-nine dollars a month.",
    "item": 2
   },
   {
    "say": "They have thirty-six thousand dollars in cash, and fixed costs of about four thousand dollars a month.",
    "item": 3
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#2": [
   {
    "say": "Step one is to break the canvas into assumptions.",
    "item": 0
   },
   {
    "say": "They wrote six assumptions, each with a pass threshold, and scored each by importance times uncertainty.",
    "item": 1
   },
   {
    "say": "Two tie for the highest score of twenty. H two is that shop owners will pay twenty-nine dollars, and H four is that they can get a customer for no more than seventy-five dollars.",
    "item": 2
   },
   {
    "say": "Then come H five at sixteen, H three at fifteen, H six at twelve, and H one at ten.",
    "item": 2
   },
   {
    "say": "H one is cheapest, so confirm it with interviews. Test H two with a pre-sale, and H three and H six with Wizard of Oz.",
    "item": 2
   },
   {
    "say": "Use a small ad test for H four, and log H five as long-term tracking, since churn cannot be tested until after launch.",
    "item": 2
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#3": [
   {
    "say": "Here is the table. Each row has an assumption, a threshold, importance, uncertainty and a score.",
    "item": 0
   },
   {
    "say": "H one says shop owners spend six or more hours a month on receipts. The threshold is at least eight of twelve interviewees, and the score is five times two, or ten.",
    "item": 2
   },
   {
    "say": "H two says shop owners will pay twenty-nine dollars a month. The threshold is a pre-sale conversion of at least fifteen percent, scoring five times four, or twenty.",
    "item": 3
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#4": [
   {
    "say": "Continuing down the table.",
    "item": 0
   },
   {
    "say": "H three says AI receipt-reading quality can be controlled. The threshold is first-pass accuracy of at least ninety percent, with low-confidence items flagged. It scores fifteen.",
    "item": 1
   },
   {
    "say": "H four says they can get a paying customer for no more than seventy-five dollars. That is judged by a trend from a small ad test, and it scores twenty.",
    "item": 2
   },
   {
    "say": "H five says monthly churn is no higher than five percent, tracked for three months after launch. It scores sixteen.",
    "item": 3
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#5": [
   {
    "say": "The last row is H six. Variable cost per customer should stay under forty-five percent of revenue.",
    "item": 0
   },
   {
    "say": "They will estimate it from Wizard of Oz data, and it scores twelve.",
    "item": 1
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#6": [
   {
    "say": "Step two is the experiments and results. Let's go through them.",
    "item": 0
   },
   {
    "say": "In weeks one and two, they interviewed twelve shop owners, and nine spend six or more hours a month. That is seventy-five percent, above the threshold.",
    "item": 1
   },
   {
    "say": "But with only twelve people, the confidence interval is wide, roughly forty-seven to ninety-one percent. So the conclusion is directionally right, not an exact seventy-five.",
    "item": 1
   },
   {
    "say": "In weeks three and four, three hundred dollars of ads brought four hundred clicks, and fifty-two people joined the waitlist. That is a thirteen percent conversion.",
    "item": 1
   },
   {
    "say": "They invited the waitlist to prepay three months at eighty-seven dollars. Eleven paid, which is twenty-one point two percent, above the fifteen percent threshold.",
    "item": 1
   },
   {
    "say": "They were honest about timing and offered a full refund. Their measured cost was about twenty-seven dollars per customer, but they kept the conservative seventy-five in the model.",
    "item": 1
   },
   {
    "say": "In weeks five to eight, a Wizard of Oz test handled twelve hundred receipts. The AI got ninety-one point five percent right, above the ninety percent threshold, and flagged nine percent for review.",
    "item": 1
   },
   {
    "say": "Among the rest, only six were wrong, and each human review took about one point two minutes.",
    "item": 1
   }
  ],
  "Case Study: How TallyTap Validated an AI Bookkeeping Idea in Eight Weeks#7": [
   {
    "say": "Step three is to update the unit economics with those real numbers, assuming one hundred fifty receipts per customer each month.",
    "item": 0
   },
   {
    "say": "AI reading costs three dollars, human review costs six dollars seventy-five, payment processing is eighty-four cents, and cloud and support add one dollar.",
    "item": 1
   },
   {
    "say": "That is eleven dollars fifty-nine in variable cost, so gross profit is seventeen dollars forty-one, a gross margin of sixty percent.",
    "item": 1
   },
   {
    "say": "With five percent churn, lifetime is twenty months, so L T V is about three hundred forty-eight dollars. Against a C A C of seventy-five, L T V over C A C is four point six four, and payback is about four point three months.",
    "item": 1
   },
   {
    "say": "Break-even needs at least two hundred thirty customers. With forty customers, monthly burn is about thirty-three hundred dollars, and the cash lasts about ten point nine months.",
    "item": 1
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#1": [
   {
    "say": "Now let's collect the formulas and templates in one place.",
    "item": 0
   },
   {
    "say": "This unit condenses the earlier methods into formulas, worked examples and fill-in templates. Work the examples out yourself, then copy the templates into your own project.",
    "item": 1
   },
   {
    "say": "All amounts are in US dollars, and AI prices and features vary by vendor and version, so check the official documentation.",
    "item": 1
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#2": [
   {
    "say": "Here are the core formulas. You will recognize them from the case study.",
    "item": 0
   },
   {
    "say": "In words, gross profit is the monthly fee minus variable costs. Lifetime is one divided by churn. L T V is gross profit times lifetime.",
    "item": 1
   },
   {
    "say": "Payback is C A C divided by monthly gross profit, and break-even is fixed costs divided by gross profit per customer.",
    "item": 1
   },
   {
    "say": "A common rule of thumb is an L T V over C A C of at least about three, with a payback that is not too long. But it is a rule of thumb, not a law, and it varies by industry and funding.",
    "item": 1
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#3": [
   {
    "say": "Let's read them one by one, starting with the first three.",
    "item": 0
   },
   {
    "say": "Monthly variable cost is inference cost, plus human review, plus payment processing fees, plus other items like cloud and support.",
    "item": 1
   },
   {
    "say": "Monthly gross profit is the monthly fee minus that variable cost.",
    "item": 2
   },
   {
    "say": "Gross margin is monthly gross profit divided by the monthly fee.",
    "item": 3
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#4": [
   {
    "say": "Next, lifetime, L T V and C A C.",
    "item": 0
   },
   {
    "say": "Average lifetime in months is one divided by the monthly churn rate.",
    "item": 1
   },
   {
    "say": "L T V is monthly gross profit times average lifetime. This is a simplified version with no discounting.",
    "item": 2
   },
   {
    "say": "C A C is total acquisition spend divided by new paying customers.",
    "item": 3
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#5": [
   {
    "say": "Now the ratios that tell you whether the model works.",
    "item": 0
   },
   {
    "say": "L T V over C A C is simply L T V divided by C A C.",
    "item": 1
   },
   {
    "say": "Payback in months is C A C divided by monthly gross profit.",
    "item": 2
   },
   {
    "say": "Break-even customers is monthly fixed costs divided by gross profit per customer, rounded up.",
    "item": 3
   }
  ],
  "Formulas and Templates: Unit Economics, Sensitivity Analysis and Launch Templates#6": [
   {
    "say": "Finally, runway.",
    "item": 0
   },
   {
    "say": "Runway in months is your cash divided by monthly net burn, where net burn is fixed costs minus total monthly gross profit.",
    "item": 1
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#1": [
   {
    "say": "Let's wrap up with the key takeaways. First, the chapter map.",
    "item": 0
   },
   {
    "say": "One, a business model is a stack of assumptions to test. They should be specific, falsifiable, have a deadline, and be ranked by importance times uncertainty.",
    "item": 1
   },
   {
    "say": "Two, minimum experiments include interviews, a landing page, pre-sale, concept test and Wizard of Oz. Write the threshold first, then run the experiment.",
    "item": 2
   },
   {
    "say": "Three, the AI risks are commoditization, vendor dependence, copyright, privacy, hallucination and accountability, and regulation. Principles matter more than statutes.",
    "item": 3
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#2": [
   {
    "say": "Four, financial modeling and sensitivity analysis. That means unit economics, break-even, runway, and finding the most sensitive assumption.",
    "item": 1
   },
   {
    "say": "Five, the twelve-week launch roadmap and the capstone project, with your canvas, unit economics, risk register and pitch.",
    "item": 2
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#3": [
   {
    "say": "Key point one is assumptions and priorities.",
    "item": 0
   },
   {
    "say": "Every block of the canvas can be rewritten as an if then assumption, and there are five types of assumptions.",
    "item": 1
   },
   {
    "say": "Rank them by importance times uncertainty, from highest to lowest.",
    "item": 3
   },
   {
    "say": "Remember that evidence runs from weak to strong. Opinion is weakest, then stated intent, behavior, commitment, and finally payment and renewal.",
    "item": 3
   },
   {
    "say": "Write pass and fail thresholds in advance. There are only three outcomes, go, pivot or stop.",
    "item": 3
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#4": [
   {
    "say": "Two more notes on this point.",
    "item": 0
   },
   {
    "say": "Evidence runs from weak to strong, starting with opinion, then stated intent, behavior, commitment, and finally payment and renewal.",
    "item": 1
   },
   {
    "say": "And write pass and fail thresholds in advance. There are only three outcomes, go, pivot or stop.",
    "item": 2
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#5": [
   {
    "say": "Key point two is minimum experiments.",
    "item": 0
   },
   {
    "say": "In interviews, ask about past behavior, not future intent.",
    "item": 1
   },
   {
    "say": "In a landing page test, test one thing at a time, bring traffic from your target segment, and honestly show the product is not yet available.",
    "item": 2
   },
   {
    "say": "A pre-sale is the key step from interested to will pay. Deliver honestly and offer a refund option.",
    "item": 3
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#6": [
   {
    "say": "A few more points on experiments.",
    "item": 0
   },
   {
    "say": "Concept tests are good for checking value-proposition wording and positioning.",
    "item": 1
   },
   {
    "say": "Wizard of Oz can measure AI accuracy, review rate and cost before development. When data is involved, tell customers in advance and get consent.",
    "item": 2
   },
   {
    "say": "And with a small sample, a conclusion is a direction, not a precise number.",
    "item": 3
   }
  ],
  "Key Takeaways: Validation, Risk and Launch#7": [
   {
    "say": "Key point three is the AI risk principles.",
    "item": 0
   },
   {
    "say": "For commoditization, put the moat in data, workflow, trust and switching costs.",
    "item": 1
   },
   {
    "say": "For vendor dependence, use an interface layer, an evaluation set, a review of terms, and a backup plan.",
    "item": 2
   },
   {
    "say": "For copyright, confirm tool licenses, avoid imitating protected content, and keep records.",
    "item": 3
   },
   {
    "say": "Privacy comes down to data minimization, notice and consent, knowing how your vendor handles data, and supporting deletion. For hallucination, tier by risk and keep a human in the loop on high-risk tasks.",
    "item": 3
   },
   {
    "say": "Regulation means tracking regulators' announcements and building compliance into requirements. This course does not provide legal advice.",
    "item": 3
   }
  ],
  "closing": [
   {
    "say": "And that brings us to the end of the course. Congratulations, you made it!",
    "item": 0
   },
   {
    "say": "Let's look back at the journey. You started with a customer problem and a value proposition, then designed the canvas, pricing, costs and channels.",
    "item": 0
   },
   {
    "say": "In this last chapter, you learned to turn those designs into testable assumptions, run minimum experiments, manage AI risks, and check the unit economics, all the way to AI profit.",
    "item": 0
   },
   {
    "say": "Before you go, take the chapter quiz to check what you have learned, and then do the final review.",
    "item": 1
   },
   {
    "say": "Then start your capstone project, where you put the canvas, unit economics, risk register and pitch together for your own idea. Thank you for learning with me, and good luck with your launch.",
    "item": 1,
    "pause": 1
   }
  ]
 }
}