AFM 241: Exam Study Reference

The puzzle of why successful, well-run companies get beaten by cheaper, "worse" products. The answer: the same rational financial logic that makes them great at improving for their best customers blinds them to disruption from below.

Judging a new investment against the cheap marginal cost (extra cost) of reusing an existing asset, instead of the full cost of building the new capability. Makes the obsolete asset look better.

Chasing short-term EPS to prop up the share price, buybacks, even sacrificing long-term value to smooth reported earnings.

Problem 1: Parmenides' Fallacy: assuming the "do nothing" future = today's healthy status quo. Really it's a non-linear decline.  Problem 2: early disruptive cash flows are underestimated. Fix: in disruption, judge qualitatively, not by DCF alone.

Run 3 tests: ① Low-end? No, same/better quality than a cab. ② New-market? No, didn't serve non-consumers. ③ Head-to-head with the incumbent? Yes, and disruptors win by flying under the radar. → It's a market-share grab, won via regulatory arbitrage (skipping commercial insurance/licensing costs) + cheap investor money.

The structural/technological advantage that lets a disruptor keep its edge as it "creeps up-market." If it has one, it's a real threat; if not, it's a "one-trick pony." (Online universities have one but undermine the degree's scarcity → not a sure win.)

Ask why a customer "hires" a product, the real job it does. Milkshake: morning commuters hire it to be thick, filling, one-handed; afternoon parents hire it as a quick kids' treat → same product, opposite improvements.

Sustaining work uses a heavyweight (co-located experts, e.g. Toyota Prius) or lightweight team. Disruptive work needs an autonomous team, a separate org with its own leadership & metrics (a new RPV). Toyota's Echo failed without one; HP's inkjet succeeded with one.

A reusable instruction file (SKILL.md, written in Markdown) that Claude loads only when relevant, turning a repeated task into a consistent, structured system. "Progressive disclosure" = it reads only what's needed, not everything every time.

Role · Context · Instructions · Examples · Format. Use XML-style tags to separate your data from your commands; iterate rather than expecting one perfect shot. The Truth Protocol = a 7-rule instruction set that makes Claude flag uncertainty instead of fabricating (covers sources, stats, quotes, code, etc.).

An independent third-party audit report confirming a software vendor actually runs strong security controls. It's the first evidence you ask for before trusting a vendor with client data. Type 1 = controls are designed properly at one point in time (≈ a review). Type 2 = controls operated effectively over 6–12 months (≈ an audit).

A powerful AI model can be pulled by government order almost overnight, and strict data-retention terms can make it unusable for client data. Lesson for accountants: "most capable ≠ most usable", governance and auditability decide adoption, so always keep a fallback (Plan B).

A graph of the predictable emotional arc every new technology goes through: over-excitement, a crash, then real durable value. It exists to help organizations time their adoption, invest too early and you pay for immature tech ("bleeding edge"); dismiss it as a fad and you miss the next big thing. The curve tracks human expectations, not the tech itself.

A chart that rates each innovation by potential benefit (Transformational > High > Moderate > Low) × years to mainstream ("plateau"). High-priority bets sit top-left (big impact, mature soon). Its real job: defend against "personality-driven" decisions, an exec pushing a pet project.

Novelty (the new thing feels limitless: Railway Mania 1840s, Internet 1990s) · Social contagion (bandwagon, no one wants to miss out) · Heuristics (mental shortcuts: "manage by magazine," e.g. a Big Data pitch "must mention Hadoop").