Chapter 36 Exercises

Work these the way the chapter taught you to think about the future: ask of every new tool what it does better than a person and what it cannot do at all, and of every climate or insurability question whether the price the math demands is one the world will pay. Items marked with a dagger (†) have worked solutions in Appendix: Answers to Selected Exercises; the rest are for discussion or self-test. Section references like (§36.3) point you back to the relevant part of the chapter. Several items reach back into earlier chapters — the future only makes sense against the craft you already know.

A. Recall and definitions

  1. † Define continuous underwriting in one sentence, and explain how the assessment it produces differs from the traditional annual snapshot. (§36.1)
  2. Define AI-augmented underwriting, and state the "co-pilot, not autopilot" division of labor in your own words. (§36.2)
  3. † Define insurability, and explain why a risk can become uninsurable even though it remains perfectly modelable. Which clause of the definition does the work? (§36.4)
  4. Explain what it means for the catastrophe baseline to be non-stationary, and why that breaks an assumption the law of large numbers (Chapter 1) and catastrophe modeling (Chapter 30) both rely on. (§36.3)
  5. Name the three products on the "new product frontier" (§36.5) and, in a phrase each, the weakness of the traditional indemnity policy that each one addresses.
  6. Define basis risk as it applies to parametric insurance, and give one example of it cutting against the insured and one cutting in their favor. (§36.5)
  7. List four of the six skills the chapter argues will matter most in 2035, and note for each whether it is a skill the machine is making more or less valuable. (§36.6)

B. Continuous underwriting and IoT

  1. † The chapter says continuous underwriting's greatest value is usually not mid-term re-pricing. Explain why not (what does the policy contract of Chapter 5 prevent?), and state what the live data is actually most valuable for instead. (§36.1)
  2. Explain how continuous underwriting "collapses the gap between underwriting and loss control" (Chapter 9). Give a single sensor signal that does both jobs at once and say what each job is. (§36.1)
  3. Name two kinds of risk for which continuous/sensor-based underwriting will arrive slowly or not at all, and explain what it is about those risks that no sensor can see. (§36.1)
  4. † A colleague proposes writing a marginal, deteriorating risk you would otherwise decline, "because we can just monitor it continuously and react." Identify the two flaws the chapter names in this reasoning, and connect one of them to adverse selection (Chapter 1). (§36.1)

C. AI as co-pilot

  1. An LLM reads a 90-page submission and a five-year loss run and produces a two-page summary, a draft risk assessment, and a list of broker questions in seconds. List two tasks here the co-pilot does well and one specific way its output could be confidently wrong (use the chapter's word). (§36.2)
  2. † Rewrite the slogan "AI will replace underwriters" into the chapter's preferred formulation, and explain in three or four sentences which kind of underwriter is being automated away and which is being amplified. Tie this to the book's fifth theme. (§36.2; Ch. 1, Ch. 32)
  3. The chapter says "you sign the quote, not the model." Explain why accountability cannot be delegated to an AI, drawing on the override discipline of Chapter 32 and the documentation discipline of Chapter 13. (§36.2)

D. Climate and the property lines

  1. † Explain, using trend versus event, why an underwriter should not spike the coastal rate after one bad hurricane season and then cut it after two quiet ones. What is the disciplined alternative, and which earlier-chapter concept (Chapter 11) is it an application of? (§36.3)
  2. For each of hurricane, wildfire, flood, and severe convective storm, state in one line the climate mechanism that is changing its frequency or severity, and name the line(s) of business most exposed. (§36.3; Ch. 15, Ch. 19)
  3. "A clean five-year coastal loss run can be a trap." Explain the trap precisely, and connect it to the Chapter 1 error of pricing for the average year while the tail gets fatter. (§36.3; Ch. 1, Ch. 8)

E. Underwrite this submission

  1. † Underwrite this. A homeowners renewal on the coast: a well-built, fully-code-compliant home, no losses in the owner's ten years. The climate-conditioned named-storm AAL for the location has roughly doubled in a decade; reinsurance has hardened; the state has approved only part of the indicated rate increase, so the risk-adequate premium is far above both the payable and the filed rate. State (a) whether the risk is measurable, (b) what specifically has failed, (c) the company-level decision, and (d) the system-level mechanism that is supposed to catch this risk. (§36.4)
  2. Underwrite this — the same risk, a different lever. Take the account in #18. Identify the single lever that could lower the adequate price (not merely shift who pays it), and explain how the credit/debit logic of Chapter 11 would reward it. Why is this the only durable answer? (§36.4; Ch. 11)
  3. A small coastal business wants protection against a hurricane shutting it down. Recommend whether a parametric wind supplement, a traditional business-income policy (Chapter 19), or both is the better answer, and name the trade-off (speed/certainty vs. basis risk) the business is accepting either way. (§36.5; Ch. 19)

F. Price / structure this risk

  1. † Structure this. You are designing an on-demand drone policy that a hobbyist switches on for the hours they fly. Name the single largest threat to the product's loss ratio (Chapter 3), explain why it is just an old enemy in a new app (Chapter 1), and propose two design rules that would push back on it. (§36.5; Ch. 1, Ch. 3)
  2. A parametric hurricane cover pays a fixed \$500,000 if a Category 3+ storm passes within 25 miles of the insured location. The storm passes at 26 miles and causes \$900,000 of damage. (a) What does the insured collect? (b) What is this gap called? (c) State one structuring change that would reduce the insured's exposure to it, and the cost of that change. (§36.5)
  3. Design the judgment upstream. The chapter argues that parametric, embedded, and on-demand products "relocate underwriting judgment upstream into the design." For an embedded shipping-protection product that quotes and binds via API in milliseconds with no human in the loop, list the three design decisions that now carry the underwriting judgment that a desk used to carry. (§36.5; Ch. 20)

G. Find the red flag

  1. † Find the red flag. An InsurTech pitches a fully-automated coastal-homeowners product: an LLM reads the application and aerial imagery, scores the risk, and binds in under a minute, "with continuous sensor monitoring after bind." Name three distinct red flags a senior underwriter would raise — at least one about climate/insurability (§36.4), one about the AI (§36.2), and one about the continuous-monitoring claim (§36.1). (§36.1, §36.2, §36.4)
  2. A pricing model's coastal output is described as "climate-aware." On inspection, its only forward-looking input is the last ten years of the carrier's own losses in the zone, trended at a flat historical rate. Explain why this is not genuinely forward-looking, and what is missing. (§36.3; Ch. 30)

H. Memo and communication

  1. † Write the memo. In about 200 words, draft a note to your underwriting manager recommending that your team adopt an LLM co-pilot for summarizing large commercial submissions. State the benefit, the two guardrails you would insist on (verification and accountability), and the single compliance risk (Chapter 35) that must be managed. (§36.2; Ch. 35)
  2. A broker complains that your renewal quote on a loss-free coastal account jumped sharply "for no reason — they've never had a claim." In a short spoken explanation (4–6 sentences), explain the climate/insurability reality to them without losing the relationship (Chapter 39). (§36.3, §36.4; Ch. 39)

I. Ethics

  1. † Ethics dilemma. Your AI-augmented system produces consistently higher prices for a particular coastal community, and the model's explanation cites only "location-based catastrophe risk." A colleague says, "It's just the cat model — it's actuarially fair, end of discussion." Using the actuarial-vs-social fairness tension (Chapter 35), explain why "end of discussion" is the wrong posture, what you would actually check, and where the genuine, unresolved tension lies. (§36.4; Ch. 35)
  2. Is it ethical to sell an embedded insurance product that binds in milliseconds to a customer who gives it almost no thought? Argue both sides — the access/convenience case and the suitability/value case — and state the regulatory question the chapter says is beginning to be asked. (§36.5)
  3. The chapter says a risk-based price no one will pay is "the market telling the truth about the physical risk," and raises the possibility that some places should not be rebuilt. Is delivering that signal part of the underwriter's job, or a harm the underwriter inflicts? Defend a position, and distinguish the underwriting decision from the societal one. (§36.4)

J. The Underwriting File

  1. † File extension. Take the Harbor Steel forward look (The Underwriting File, this chapter) and write the three-to-four-sentence "2035 addendum" you would clip to the file. It must (a) name one thing continuous/IoT underwriting would change about the two fires that drove this account, (b) name what the AI co-pilot would and would not do, and (c) state honestly that the climate repricing raises the catastrophe load — without pre-empting the capstone's binding decision (Chapter 40). (The Underwriting File; Ch. 27, Ch. 30, Ch. 40)
  2. The forward look says a parametric wind supplement (floated in Chapter 34) becomes a live option for Harbor Steel by 2035. Sketch the trigger you would propose (peril, threshold, distance), state what it would sit on top of (which deductible from the terms already set), and name the basis risk the company would be accepting. (The Underwriting File; Ch. 12, Ch. 34)
  3. In a 2035 world, the Port Hadley named-storm AAL on a climate-conditioned view is higher than today's file carries. Explain, using portfolio thinking (Chapter 29) and the cat aggregate (Chapter 30), why that makes Harbor Steel's fate depend less on the account's own quality and more on whether the zone still has room. (The Underwriting File; Ch. 29, Ch. 30)