Study Guide Sample
The AI Dividend: Enterprise AI Transformation
A companion for revision, practice and applied management decisions
Study Guide Sample
The AI Dividend: Enterprise AI Transformation
Chapter 4: AI Economics and Business Value
Return on Investment, Quick Wins, and Strategic Options
What is this study guide?
A chapter study guide is a companion to the book, designed to help readers review the key ideas, test their understanding and practise applying them to management decisions. Use it after reading the corresponding chapter for independent study, revision or classroom discussion.
What this sample contains
The complete chapter summary, four selected MCQs with an answer key, three short-answer questions with model answers, and two long-answer questions with structured hints. The full Chapter 4 guide contains 10 MCQs, five short-answer questions and two long-answer questions. Chapter guides cover all 20 chapters.
How to use it
Attempt the questions before consulting the answers. Compare your reasoning with the model answers and use the essay hints to plan a longer response. Return to the chapter where you need a fuller explanation.
Chapter summary
The chapter argues that a disciplined business case, not the technology choice, decides whether an AI investment should be made at all. Value reduces to five mechanisms (the Five-Lever Value Map): revenue growth, productivity, cost reduction, quality improvement, and risk reduction. Each carries different measurement and attribution demands, and conflating them leads enterprises to overfund the easily justified cost cases and underfund the strategically important revenue and risk cases.
Standard return-on-investment arithmetic captures only the direct return of the first model; it has no language for strategic option value: the data assets, organizational capability, and competitive positioning that compound and make later investments cheaper or possible. Three plain techniques make that option value visible without option-pricing mathematics: the capability roadmap, competitive benchmarking, and a real-options framing. Every portfolio runs two clocks at once: quick wins that ship within a budget cycle and build credibility, and capability-building investments whose returns arrive later but lower the marginal cost of everything downstream; favoring one starves the other.
The Dividend Stop Test supplies the discipline most business cases omit, declining investments that fail any of four conditions: a simpler tool would suffice, the data is not there, the organization will not act on the output, or regulation and ethics close the door. Indian conditions enter the business case as first-order inputs: digital public infrastructure lowers the data bill, the served addressable market must be distinguished from the total, wage economics tilt the case from automation toward augmentation, and the MSME access dividend means first-generation AI must be measured by access return rather than optimization return.
Multiple-choice questions
1. What does the chapter identify as the prior question that decides whether an AI technology choice should be made at all?
2. A bank weighs a chatbot with a six-month payback against a customer data platform that enables five downstream applications but returns nothing for eighteen months. Which framing best justifies the platform?
3. A proposal would deploy a learned model where a deterministic rule set captures most of the value at a fraction of the cost. Which Dividend Stop Test condition does it fail?
4. An automation business case assumes full adoption, zero ongoing run cost, and instant realization from day one. What does the chapter advise?
Check the MCQ answer key
1: D | 2: B | 3: B | 4: D
Short-answer questions
1. Explain why conflating the five value mechanisms leads to poor capital allocation, and what discipline corrects it.
Read the model answer
Model answer. The five mechanisms (revenue growth, productivity, cost reduction, quality, and risk) differ in how easily their value can be measured and attributed. Cost-reduction AI is easy to justify on its face; revenue-growth AI is harder to attribute and carries higher strategic upside; risk-reduction value is probabilistic. When a business case blurs them, the organization tends to fund the easily measured cases and starve the strategically important ones, then misread the results because no one specified which mechanism was meant to drive value. The correcting discipline is to name which mechanism is primary, which are secondary, and how each will be measured after deployment, with owners and a review cadence. The managerial implication is that a case listing three primary mechanisms is usually hiding the fact that the sponsors do not know which one really drives the value, and a case listing none is not a case at all.
2. A sponsor proposes a sophisticated learned model for a problem a rule engine could largely solve. Apply the Dividend Stop Test to advise the investment committee.
Read the model answer
Model answer. The Dividend Stop Test declines investments that fail any of four conditions, and this proposal triggers the first: a simpler tool would do the job. If a deterministic policy engine captures most of the value at a fraction of the build cost and with far better explainability, the engine is the better investment, and a proposal that has not engaged seriously with that alternative is incomplete. The pattern is the capability trap, deploying AI because the organization now can, rather than because the problem requires it. The advice is to return the proposal for rework that either justifies why the learned model is necessary or adopts the simpler tool, and to protect scarce AI talent from problems that do not need it. Running the test at the front door is cheaper than discovering the mismatch in production, since a meaningful share of proposals fail at least one condition.
3. Explain how strategic option value differs from direct ROI, and how a manager can make it visible without option-pricing mathematics.
Read the model answer
Model answer. Direct ROI divides measurable benefits by measurable costs over a fixed horizon and captures only what the current model does. Strategic option value is the set of follow-on moves the investment makes possible, operating through three compounding channels: data assets that deepen over time, organizational capability that lowers the cost of later deployments, and competitive positioning that buys time-to-market. Standard arithmetic ignores all three because it asks only what the first investment returns, not what it enables. Three plain techniques make the value visible: a capability roadmap that lists the downstream initiatives the investment unlocks with a probability for each; competitive benchmarking that prices the cost of falling two years behind a rival; and a real-options framing that treats infrastructure as a prepaid call option on future opportunities. In practice, option value should be narrated explicitly and kept separate from the contractable direct return, never blended into one optimistic total.
Long-answer questions
1. A mid-sized enterprise’s AI steering committee proposes to evaluate every prospective AI initiative solely on direct return on investment, on the grounds that strategic option value is speculative and invites inflated cases. Construct the counter-argument and recommend an evaluation approach that keeps discipline while crediting option value honestly.
Read the answer hints
Frameworks to deploy: Strategic option value (data, capability, positioning) with the capability-roadmap, competitive-benchmarking, and real-options framings, and the quick-win versus capability-building distinction.
Word count: 600–900 words.
Pitfall to avoid: Do not argue that option value should be folded into the headline ROI figure; the chapter’s discipline is to keep direct return and option value distinct, not to inflate one with the other.
Structural beats:
Open by conceding the committee’s legitimate concern: option value is real but not contractable, and treating it as committed revenue damages credibility.
Argue which categories of investment a direct-ROI-only rule systematically rejects, including data platforms, governance, and literacy, and why rejecting them harms the long-term position.
Show how the three plain framings make option value defensible without pretending it is committed return.
Close with a recommended approach that numbers the direct return honestly, narrates option value separately, and attaches an explicit stop rule.
2. An Indian fintech is pitching an investor on an AI credit-scoring product for the MSME market. The investor applies US fintech benchmarks and concludes the unit economics are marginal. Using the chapter’s Indian-context arguments, rebuild the case and take a position on how AI return should be measured in this segment.
Read the answer hints
Frameworks to deploy: The access-return versus optimization-return distinction, the digital public infrastructure data-cost effect, and the served-versus-total addressable market distinction.
Word count: 600–900 words.
Pitfall to avoid: Do not claim the entire MSME population is immediately addressable; the served addressable market is smaller than the total, and the case must distinguish them or it will not survive contact with operating conditions.
Structural beats:
Open by identifying what the Western benchmark misses: it measures optimization return against an existing baseline, but much of the MSME segment has no analytical baseline to improve on.
Argue that the relevant measure is access return, the value created when formal credit reaches enterprises that previously had none, which standard templates undervalue by an order of magnitude.
Bring in the DPI data-cost advantage and the need to size the served addressable market honestly rather than quote an undifferentiated population count.
Close by taking a position on the right return framing for first-generation AI in underserved markets, with the caveats the chapter attaches.