{
 "byKey": {
  "cover": [
   {
    "say": "Welcome back to Business Model Design.",
    "item": 0
   },
   {
    "say": "In this chapter, we ask the question every AI founder eventually faces: how do you actually make money from AI, and keep what you earn?",
    "item": 0
   },
   {
    "say": "We'll cover pricing, unit economics, moats, and a full case study with real arithmetic.",
    "item": 0
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#1": [
   {
    "say": "Let's start with why AI profit cannot simply copy the traditional software playbook.",
    "item": 0
   },
   {
    "say": "Classic software is wonderful because once the code is built, selling one more copy costs almost nothing, so gross margins can be very high.",
    "item": 1
   },
   {
    "say": "AI is different. Every request calls a model, and that inference cost is billed by usage, so your cost rises as customers use more.",
    "item": 1
   },
   {
    "say": "That changes everything. How you charge must consider who drives the cost, and gross margin has to be calculated line by line.",
    "item": 2
   },
   {
    "say": "And your moat cannot be that you use a powerful model, because a competitor can call the same model tomorrow.",
    "item": 2
   },
   {
    "say": "So this chapter answers three questions: how you earn revenue, what eats your margin, and why others cannot copy you.",
    "item": 3
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#2": [
   {
    "say": "Line one is revenue: how you charge for value.",
    "item": 0
   },
   {
    "say": "Price is not cost plus a little profit. It is the share of the value customers get that they are willing to pay you.",
    "item": 0
   },
   {
    "say": "The first unit is the seat, a fixed amount per user per month.",
    "item": 1
   },
   {
    "say": "Next is usage, where you charge per request, per thousand words, per image, or per minute.",
    "item": 2
   },
   {
    "say": "Then outcome, where you are paid only when a support ticket is resolved or an order closes.",
    "item": 3
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#3": [
   {
    "say": "Two more charging units complete the list of five.",
    "item": 0
   },
   {
    "say": "Hybrid combines a base monthly fee with usage-based charges once the customer goes above an allowance.",
    "item": 1
   },
   {
    "say": "Freemium gives basic features away free, and charges for advanced features or higher usage.",
    "item": 2
   },
   {
    "say": "The rule to remember: pick a charging unit close to where the customer feels value, without letting your own costs run away.",
    "item": 2
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#4": [
   {
    "say": "Line two is cost: what eats your gross margin.",
    "item": 0
   },
   {
    "say": "The variable costs of an AI product include inference, the vector database and storage, payment processing fees, and any human review or support.",
    "item": 1
   },
   {
    "say": "Many teams calculate margin from the cloud bill alone, forget the humans, and discover that the more they sell, the less they earn.",
    "item": 1
   },
   {
    "say": "The third unit in this chapter shows how to itemize all of it, and protect margin with caching, model tiers, and usage caps.",
    "item": 1
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#5": [
   {
    "say": "Line three is the moat: why others cannot copy you easily.",
    "item": 0
   },
   {
    "say": "Model capability keeps improving and vendors converge, which is called model commoditization. Saying we also have AI won't keep customers for long.",
    "item": 0
   },
   {
    "say": "The first lasting moat is data. With legal use and customer authorization, you collect proprietary data others cannot get, like labeled industry data or correction records.",
    "item": 1
   },
   {
    "say": "That creates a feedback loop where the product gets more accurate the more it is used.",
    "item": 1
   },
   {
    "say": "The second is workflow. When your product sits inside the customer's systems, permissions, approvals, and reports, switching costs are high.",
    "item": 2
   },
   {
    "say": "A model moat, meaning a lead from training your own special model, is usually the most expensive and easiest to catch up to.",
    "item": 2
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#6": [
   {
    "say": "Now let's build a map of the seven monetization paths. The next unit covers them in detail.",
    "item": 0
   },
   {
    "say": "One: API and agent services. You wrap model capability as an interface or an agent that does tasks, charged by usage.",
    "item": 1
   },
   {
    "say": "Two: workflow and automation services, where you automate a business process like invoicing or reconciliation, charged monthly or by outcome.",
    "item": 2
   },
   {
    "say": "Three: vertical AI. You focus on one industry, such as dental practices or logistics, and build its knowledge, data, and workflow into the product.",
    "item": 3
   }
  ],
  "Principles: The Three Lines of AI Profit and the Moat#7": [
   {
    "say": "Here are the remaining paths on the map.",
    "item": 0
   },
   {
    "say": "Four: white-label and embedded, where other companies put your AI feature inside their product under their own brand.",
    "item": 1
   },
   {
    "say": "Five: content and template marketplaces, selling prompts, workflow templates, design assets, and other digital goods.",
    "item": 2
   },
   {
    "say": "Six: consulting plus product. You charge service fees while turning the repeatable parts into a product.",
    "item": 3
   },
   {
    "say": "Seven is the subscription AI assistant, a monthly assistant for individuals or small teams, usually grown through self-serve signup.",
    "item": 3
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#1": [
   {
    "say": "First, a quick note before the numbers.",
    "item": 0
   },
   {
    "say": "All prices, costs, and ratios in this unit are made-up numbers for practice, not any real product or market rate.",
    "item": 1
   },
   {
    "say": "Real model prices, plans, and features change and vary by vendor and version, so check the official documentation.",
    "item": 1
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#2": [
   {
    "say": "Let's walk through the five pricing models, starting with per seat.",
    "item": 0
   },
   {
    "say": "You charge a fixed amount per user per month.",
    "item": 1
   },
   {
    "say": "It fits when value grows with the number of users and each user's usage is similar, like a team collaboration tool.",
    "item": 2
   },
   {
    "say": "The upside is predictable revenue that customers easily understand.",
    "item": 3
   },
   {
    "say": "The risk is that AI usage varies widely, so one heavy user's inference cost can eat the margin of the whole seat. Customers will also share accounts.",
    "item": 4
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#3": [
   {
    "say": "Next, usage-based pricing.",
    "item": 0
   },
   {
    "say": "You charge per request, per thousand words, per image, or per minute.",
    "item": 1
   },
   {
    "say": "It fits APIs, agent services, and customers whose volume swings a lot.",
    "item": 2
   },
   {
    "say": "The upside is that cost and revenue grow together, so your margin is better protected.",
    "item": 3
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#4": [
   {
    "say": "Usage-based pricing has a downside, and then we move to outcome-based.",
    "item": 0
   },
   {
    "say": "Bills are hard to predict, and customers resist not knowing what they will owe at month-end. Small usage also means small revenue.",
    "item": 1
   },
   {
    "say": "Outcome-based means you are paid only for each resolved ticket or each qualified sales lead.",
    "item": 2
   },
   {
    "say": "It fits when the outcome is easy to define and both sides can measure it.",
    "item": 3
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#5": [
   {
    "say": "More on outcome-based pricing, then hybrid.",
    "item": 0
   },
   {
    "say": "The upside is low perceived risk for the customer and the strongest sense of value.",
    "item": 1
   },
   {
    "say": "The risk is real, though. You must agree on what resolved means, revenue drops directly when AI performance gets worse, and you absorb the swings in model quality.",
    "item": 2
   },
   {
    "say": "Hybrid is a base monthly fee that includes an allowance, with usage above it billed by volume.",
    "item": 3
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#6": [
   {
    "say": "Finishing the hybrid model and then freemium.",
    "item": 0
   },
   {
    "say": "The customer gets a predictable floor price and you get a cap on your cost exposure. It is the most common and most stable combination.",
    "item": 1
   },
   {
    "say": "Freemium uses a free tier to attract users and a paid tier for higher usage or advanced features.",
    "item": 2
   },
   {
    "say": "The risk is that free users also generate inference cost, so the free tier needs a clear usage cap, and conversion rate must be built into your profit and loss.",
    "item": 3
   }
  ],
  "Concepts: AI Product Pricing and Seven Ways to Monetize#7": [
   {
    "say": "Let's put numbers on this with one fictional customer, Hartwell Home Appliances. All figures are illustrative.",
    "item": 0
   },
   {
    "say": "They have ten support agents and four thousand tickets a month. Per seat at forty dollars per agent comes to four hundred dollars.",
    "item": 1
   },
   {
    "say": "Usage at twenty-five cents per ticket comes to one thousand dollars.",
    "item": 2
   },
   {
    "say": "Outcome at sixty-five cents per resolved ticket, with a fifty-five percent resolution rate, comes to one thousand four hundred thirty dollars. If resolution falls to thirty percent, that drops to seven hundred eighty.",
    "item": 3
   },
   {
    "say": "Hybrid, a hundred fifty dollar base plus ten cents per ticket, comes to five hundred fifty dollars.",
    "item": 4
   },
   {
    "say": "After inference cost and a payment fee, gross margins are about fifty-two, seventy-nine, eighty-four, and sixty-four percent, in that order. Outcome looks best, but it swings with quality.",
    "item": 4
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#1": [
   {
    "say": "A quick note again before we do the math.",
    "item": 0
   },
   {
    "say": "These model prices and costs are made-up numbers for practice, not any vendor's quote. Check the official documentation for real prices.",
    "item": 1
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#2": [
   {
    "say": "Unit economics means checking whether each additional sale makes or loses money. For AI, start with the smallest unit: one request.",
    "item": 0
   },
   {
    "say": "Input tokens are what you send to the model, including the system prompt, the user's question, and any retrieved data.",
    "item": 1
   },
   {
    "say": "Output tokens are what the model replies with.",
    "item": 2
   },
   {
    "say": "Inference cost per request is input tokens times the input price, plus output tokens times the output price, with prices listed per million tokens.",
    "item": 3
   },
   {
    "say": "Illustration: a large model costs three dollars per million input tokens and fifteen dollars per million output. A request with three thousand input and four hundred output costs one point five cents.",
    "item": 3
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#3": [
   {
    "say": "Now let's see that cost is more than inference.",
    "item": 0
   },
   {
    "say": "Say the monthly fee is a hundred dollars and covers a thousand requests at one point five cents each, so inference is fifteen dollars.",
    "item": 1
   },
   {
    "say": "Add a payment fee of two dollars ninety, four dollars for vector database and hosting, and about eight dollars of human review.",
    "item": 1
   },
   {
    "say": "Total cost is twenty-nine ninety, gross profit is seventy dollars ten cents, and margin is about seventy percent.",
    "item": 1
   },
   {
    "say": "That looks fine, but it is the typical customer. Next, the heavy one.",
    "item": 1
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#4": [
   {
    "say": "Here is the risk of a usage spike.",
    "item": 0
   },
   {
    "say": "Suppose one customer makes ten thousand requests, ten times the allowance, but still pays a hundred dollars.",
    "item": 1
   },
   {
    "say": "Inference alone is a hundred fifty dollars, total cost is about one sixty-five, and gross profit is negative sixty-four ninety. This customer costs you nearly sixty-five dollars every month.",
    "item": 1
   },
   {
    "say": "Spikes come from customers running batches through automation, abuse or scraping, or your own agent looping and calling the model repeatedly.",
    "item": 1
   },
   {
    "say": "So design monitoring and protections before launch.",
    "item": 1
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#5": [
   {
    "say": "There are four levers for controlling cost. Here are the first three.",
    "item": 0
   },
   {
    "say": "Caching. When the system prompt and fixed knowledge repeat, many vendors offer a cheaper cached price, though that varies by vendor and version.",
    "item": 1
   },
   {
    "say": "Illustration: with two thousand of three thousand input tokens cached at ten percent of price, cost drops from one point five cents to about zero point nine six cents.",
    "item": 1
   },
   {
    "say": "Model tiers and routing. Send simple requests, like store hours, to a small model, and save the large model for hard ones.",
    "item": 2
   },
   {
    "say": "If seventy percent go to the small model, average cost falls to about half a cent per request.",
    "item": 2
   },
   {
    "say": "Third, slimmer prompts and outputs. Cut unnecessary context, limit output length, and retrieve only the pieces you need.",
    "item": 3
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#6": [
   {
    "say": "The fourth lever is cost ceilings and guardrails.",
    "item": 0
   },
   {
    "say": "That means allowances, overage billing, fair-use rules, per-customer daily caps, and anomaly alerts.",
    "item": 1
   },
   {
    "say": "But quality is the floor. Bad routing makes the small model answer hard questions badly, so measure quality with an evaluation set, not cost alone.",
    "item": 1
   }
  ],
  "Concepts: AI Unit Economics, Productizing Services, and Common Traps#7": [
   {
    "say": "Before we leave this unit, let's check your moats with five questions.",
    "item": 0
   },
   {
    "say": "First: the more customers use the product, do your data or accuracy improve? That's a feedback loop.",
    "item": 1
   },
   {
    "say": "Second: is the way you obtain data legal, and have customers authorized it?",
    "item": 2
   },
   {
    "say": "Third: how long and how much would it cost a customer to replace you?",
    "item": 3
   },
   {
    "say": "The last two ask which customer systems you've integrated, and why customers would stay if someone built a similar product with the same model tomorrow.",
    "item": 3
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#1": [
   {
    "say": "Now let's follow a case study, AnswerDesk.",
    "item": 0
   },
   {
    "say": "The company, the people, and all numbers are fictional teaching material. Prices and costs are illustrative, so check official documentation for real ones.",
    "item": 1
   },
   {
    "say": "The calculations in the case have been verified with Python.",
    "item": 1
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#2": [
   {
    "say": "Meet Clearpath Digital, a two-person studio.",
    "item": 0
   },
   {
    "say": "Founder Rachel Torres knows customer service processes, and her partner Marcus Chen is an engineer.",
    "item": 1
   },
   {
    "say": "They build AI customer support for retail stores, where AI drafts replies that a support rep approves. Each project is five thousand dollars and takes about a hundred twenty hours.",
    "item": 1
   },
   {
    "say": "After the fourth customer, Rachel notices, as an assumption, that about seventy percent of the work is the same at every store.",
    "item": 2
   },
   {
    "say": "She wonders: could this become a product?",
    "item": 2
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#3": [
   {
    "say": "Step one: validate demand before writing code. They do three things.",
    "item": 0
   },
   {
    "say": "They list the repeated steps across the four projects and the hours each takes.",
    "item": 1
   },
   {
    "say": "They interview customers, asking what they would pay monthly if they set it up themselves, and whether they would prepay.",
    "item": 2
   },
   {
    "say": "Then they estimate the build cost of a product version, assuming twenty thousand dollars.",
    "item": 3
   },
   {
    "say": "The order matters. Repetition decides whether it can be standardized, willingness to pay decides if it's worth building, and build cost decides payback.",
    "item": 3
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#4": [
   {
    "say": "Step two: design the pricing. They choose a hybrid plan called AnswerDesk.",
    "item": 0
   },
   {
    "say": "A one thousand dollar onboarding fee, including knowledge base import and testing.",
    "item": 1
   },
   {
    "say": "A hundred dollars a month, including a thousand answered conversations.",
    "item": 2
   },
   {
    "say": "Five cents per conversation above the allowance.",
    "item": 3
   },
   {
    "say": "Why not per seat? Store support is often just the owner, so seats don't reflect value or protect cost. Why not pure outcome-based? Resolved is fuzzy in retail, which invites disputes.",
    "item": 3
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#5": [
   {
    "say": "Step three: calculate the unit economics, using the earlier illustrative prices.",
    "item": 0
   },
   {
    "say": "The typical customer has a thousand conversations. Total cost is twenty-nine ninety, gross profit seventy dollars ten cents, and margin about seventy percent.",
    "item": 1
   },
   {
    "say": "The heavy customer has ten thousand conversations with no overage billing. Total cost is about one sixty-five, and gross profit is negative sixty-four ninety.",
    "item": 2
   },
   {
    "say": "With overage billing, nine thousand extra conversations at five cents add four hundred fifty dollars, so revenue is five hundred fifty.",
    "item": 3
   },
   {
    "say": "Total cost is a hundred seventy-seven ninety-five, gross profit three seventy-two oh five, and margin about sixty-seven point six percent.",
    "item": 3
   },
   {
    "say": "The decision: before launch, overage billing and usage guardrails are a must.",
    "item": 3
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#6": [
   {
    "say": "Step four: protect margin with caching and routing.",
    "item": 0
   },
   {
    "say": "Marcus finds, as an assumption, that about seventy percent of questions are simple. Those go to the small model.",
    "item": 1
   },
   {
    "say": "The system prompt and policy documents are cached, two thousand of three thousand input tokens at ten percent of the normal price.",
    "item": 2
   },
   {
    "say": "The large model drops to about one cent, and the small model to under a tenth of a cent.",
    "item": 3
   },
   {
    "say": "Average inference is about three point five five dollars per thousand conversations. Gross margin rises to about eighty-one point five percent, and the heavy customer reaches about fifty percent even without overage.",
    "item": 3
   },
   {
    "say": "They also build a two hundred question evaluation set to confirm that quality doesn't drop before going live.",
    "item": 3
   }
  ],
  "Case Study: AnswerDesk Goes From Consulting Project to Subscription Product#7": [
   {
    "say": "Step five: break-even.",
    "item": 0
   },
   {
    "say": "Fixed costs are six thousand dollars a month. Each typical customer contributes about eighty-one fifty-five with caching and routing, so they need six thousand divided by eighty-one fifty-five, about seventy-three point six, rounded up to seventy-four customers.",
    "item": 1
   },
   {
    "say": "Without optimization, each contributes seventy dollars ten cents, which means eighty-six customers.",
    "item": 1
   },
   {
    "say": "That's twelve more customers, about sixteen percent more, just from controlling cost.",
    "item": 1
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#1": [
   {
    "say": "This unit gives you formulas you can reuse.",
    "item": 0
   },
   {
    "say": "All prices, costs, and ratios here are made-up numbers for practice, not any vendor's quote, and they don't guarantee any actual margin.",
    "item": 1
   },
   {
    "say": "Check the official documentation for real prices, then swap in your own numbers. The formulas stay the same.",
    "item": 1
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#2": [
   {
    "say": "Formula one is the inference cost per request.",
    "item": 0
   },
   {
    "say": "In words: input tokens times the input price, plus output tokens times the output price, divided by one million, because prices are listed per million tokens.",
    "item": 1
   },
   {
    "say": "Example one: the large model costs three dollars per million input tokens and fifteen per million output, for three thousand input and four hundred output tokens.",
    "item": 2
   },
   {
    "say": "Input cost is nine tenths of a cent, output cost is six tenths, so the total is one point five cents.",
    "item": 2
   },
   {
    "say": "Example two, the small model at thirty cents and a dollar fifty per million, with the same tokens, costs about a tenth of a cent and a half.",
    "item": 3
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#3": [
   {
    "say": "Here is formula one written out, step by step.",
    "item": 0
   },
   {
    "say": "The formula is input tokens times input price, plus output tokens times output price, all divided by one million.",
    "item": 1
   },
   {
    "say": "For the input side, three thousand times three divided by one million equals zero point zero zero nine.",
    "item": 2
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#4": [
   {
    "say": "Continuing the large model example.",
    "item": 0
   },
   {
    "say": "For output, four hundred times fifteen divided by one million equals zero point zero zero six.",
    "item": 1
   },
   {
    "say": "Add the two and inference cost is zero point zero one five dollars per request.",
    "item": 2
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#5": [
   {
    "say": "And the small model example.",
    "item": 0
   },
   {
    "say": "Three thousand times point three, plus four hundred times one point five, each divided by one million, gives zero point zero zero one five dollars.",
    "item": 1
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#6": [
   {
    "say": "Formula two is average cost after caching and routing.",
    "item": 0
   },
   {
    "say": "Caching swaps reusable input for a lower price, and routing blends the large and small model costs by share.",
    "item": 1
   },
   {
    "say": "Example three: with two thousand cacheable tokens at ten percent of the price, the large model costs zero point zero zero nine six dollars and the small model zero point zero zero zero nine six.",
    "item": 2
   },
   {
    "say": "Example four: seventy percent to the small model and thirty to the large, both cached, gives an average of zero point zero zero three five five two dollars.",
    "item": 3
   },
   {
    "say": "Routing alone, without caching, would average zero point zero zero five five five.",
    "item": 3
   }
  ],
  "Formulas and Templates: AI Profit Calculations and Ready-to-Use Templates#7": [
   {
    "say": "One last example to close this formula.",
    "item": 0
   },
   {
    "say": "Routing only, with no caching, seventy percent at zero point zero zero one five and thirty percent at zero point zero one five, equals zero point zero zero five five five dollars.",
    "item": 1
   }
  ],
  "Key Takeaways: Monetizing AI#1": [
   {
    "say": "Let's recap the chapter, starting with the map.",
    "item": 0
   },
   {
    "say": "First, the three lines: revenue, cost, and moat.",
    "item": 1
   },
   {
    "say": "Second, five pricing models and seven monetization paths.",
    "item": 2
   },
   {
    "say": "Third, AI unit economics: inference cost, gross margin, spike risk, caching, routing, and caps.",
    "item": 3
   },
   {
    "say": "Fourth, data and workflow moats, and productizing a service.",
    "item": 4
   },
   {
    "say": "Fifth, common traps and when not to use AI, plus formulas and templates. Prices are illustrative, so check the official documentation.",
    "item": 5
   }
  ],
  "Key Takeaways: Monetizing AI#2": [
   {
    "say": "The chapter map continues.",
    "item": 0
   },
   {
    "say": "The last piece is the formulas and templates, which you can reuse with your own numbers.",
    "item": 1
   }
  ],
  "Key Takeaways: Monetizing AI#3": [
   {
    "say": "Takeaway one: how AI differs from traditional software.",
    "item": 0
   },
   {
    "say": "Every AI request has an inference cost, so marginal cost is not zero.",
    "item": 1
   },
   {
    "say": "Calculate gross margin line by line: inference, fees, hosting, and human review.",
    "item": 2
   },
   {
    "say": "And a moat cannot rest on we use a certain model.",
    "item": 3
   }
  ],
  "Key Takeaways: Monetizing AI#4": [
   {
    "say": "Takeaway two: the five pricing models.",
    "item": 0
   },
   {
    "say": "Per seat works when value scales with headcount and usage is similar.",
    "item": 1
   },
   {
    "say": "Usage pricing means cost follows usage and revenue keeps pace.",
    "item": 2
   },
   {
    "say": "Outcome-based is low risk for the customer, but you must define the outcome, and you absorb the swings.",
    "item": 3
   },
   {
    "say": "Hybrid is a base price plus overage, and it is the most common.",
    "item": 4
   }
  ],
  "Key Takeaways: Monetizing AI#5": [
   {
    "say": "One more pricing takeaway.",
    "item": 0
   },
   {
    "say": "For freemium, calculate the break-even conversion rate. The illustrative example in the chapter is two point four percent.",
    "item": 1
   }
  ],
  "Key Takeaways: Monetizing AI#6": [
   {
    "say": "Takeaway three: the seven monetization paths.",
    "item": 0
   },
   {
    "say": "They are API or agent, workflow automation, vertical AI, white-label or embedded, template marketplace, consulting plus product, and subscription assistant.",
    "item": 1
   },
   {
    "say": "When choosing, weigh pain intensity, willingness to pay, margin, your capability, and the moat.",
    "item": 1
   }
  ],
  "Key Takeaways: Monetizing AI#7": [
   {
    "say": "Takeaway four: the unit economics formulas.",
    "item": 0
   },
   {
    "say": "Cost per request is input times input price plus output times output price, divided by one million.",
    "item": 1
   },
   {
    "say": "Gross profit is the monthly fee minus inference, fees, hosting, and review.",
    "item": 2
   },
   {
    "say": "Request cap is your inference budget divided by cost per request.",
    "item": 3
   },
   {
    "say": "Break-even customers is fixed costs divided by contribution per customer, rounded up. Freemium break-even conversion is free user cost divided by paid contribution plus free user cost.",
    "item": 3
   }
  ],
  "closing": [
   {
    "say": "That's the end of this chapter on monetizing AI.",
    "item": 0
   },
   {
    "say": "You've seen how to charge, how to protect your margin, and how to build a moat.",
    "item": 0
   },
   {
    "say": "Please complete the lab, then take the chapter quiz to check how much you've learned.",
    "item": 1,
    "pause": 1
   }
  ]
 }
}