Everything covered across Modules 1–10 is a lot of material to simply restate in sequence, and restating it is not what a board wants from a closing report. What it wants is a judgement: now that real evidence exists, which trends still deserve investment, and which looked promising on paper but have earned less attention than a generic industry forecast would give them? That judgement is the same one Module 1's innovation-process audit found Thornfield structurally weak at — strong on search, weak on select and resource — so revisiting the trends is this capstone's first opportunity to do the selecting the organisation has historically avoided, this time with evidence rather than enthusiasm to weigh.

The evidence does not rank neatly by how exciting each trend sounded at the time it was introduced. AI in marketing has the strongest case for continued investment: the segmentation pilot went live once the depot-level "on-time delivery" inconsistency was fixed, and the proactive fairness audit of the telematics scorecard turned a governance risk into demonstrated capability rather than a live exposure. Web3 is not dead, but it is correctly stalled: the proof-of-delivery ledger's customer value was real, yet it was deferred, not rejected, because Pathfinder's own unreliable delivery-milestone data — a legacy-system problem distinct from the AI pilot's depot-level definition gap — would only be made permanently and verifiably wrong once written to an unalterable ledger; its case rests on that legacy data problem being fixed, not on the technology itself. Emerging technology splits three ways: Doncaster's warehouse robotics is a proven, low-risk sustaining win whose rollout is now an operational decision; the customer-service chatbot and, more so, opening the platform to independent hauliers are the most genuinely disruptive opportunities reviewed anywhere in the course, and also the least evidenced; the platform reassessment, the bigger of the two, is not yet board-approved. Sustainability's reframed, substantiated claim is low-risk and credible, but modest in scale next to the others.

Rumelt's (2011) point about diagnosis applies as much to this exercise as it did to Module 2's strategic bets: a plan that calls every trend equally urgent has not done the strategic work of choosing, it has merely listed.

A roadmap earns the name only once it stops being a folder of separate worksheets and becomes one argument a board can follow start to finish. That argument has a fixed shape: a vision narrowed into a small number of strategic bets, each traceable to a named driver (Rumelt, 2011); an operating model naming who owns each bet, including any gap the exercise finds (Ross, Mocker and Beath, 2018); a current-state-to-target-state sequence in which foundational, dependency-creating work precedes the capability built on it, consistent with dynamic-capabilities thinking (Teece, Pisano and Shuen, 1997); and a delivery and leadership plan that names the team, its sprint structure, its riskiest assumption (Ries, 2011) and how its people will actually be led through the friction a real programme produces, not just the tasks it requires.

Assembling these pieces in one document, rather than leaving them as separate artefacts from separate modules, is itself the discipline this lesson teaches. A roadmap that lists strategic bets but never names an operating-model owner reads as confident until someone asks who is accountable; one that sequences initiatives by excitement rather than dependency reads as ambitious until the foundational gap it skipped resurfaces downstream, exactly as an unresolved data-quality problem resurfaces in an AI project that depends on it. The delivery and leadership sections deserve the same rigour as the strategy sections: a cross-functional team with a named riskiest assumption and a Mendelow's Matrix for its stakeholders (Mendelow, cited in Johnson and Scholes, 1991) is a credible delivery plan; a generic promise to "manage change well" is not.

A roadmap that has only ever been presented to people who already agree with it has not yet been tested. Stress-testing means deliberately asking what could derail each strategic bet, and sorting the answers by how likely and how damaging they are — a probability-impact matrix, of the kind set out in the Project Management Institute's PMBOK Guide (Project Management Institute, 2017), gives that sorting a disciplined shape rather than a gut feeling about which risk to worry about most. The risks worth stress-testing go well beyond delivery slippage. Stakeholder risk asks whether a Mendelow placement made early in delivery still holds once a bet's scope grows, since a stakeholder's power and interest can shift as a programme proceeds. Governance and ethical risk asks whether a new initiative gets the same proactive scrutiny a fairness audit gave an existing system, rather than scrutiny only after a complaint arrives. Reputational risk asks whether any claim the roadmap makes — financial, environmental or otherwise — would survive the same substantiation test a sustainability claim had to pass before publication.

Financial risk deserves its own line, because it is the easiest to understate without anyone intending to. Flyvbjerg (2006) found that cost and benefit forecasts are distorted by optimism bias and, sometimes, strategic misrepresentation; a roadmap's own benefits case carries both risks, and naming them explicitly, with a named owner for each mitigation, is what turns a risk register from a formality into a document a board can actually rely on when deciding how much confidence to place in the numbers behind it.

A finished roadmap still has to survive the ten minutes a board actually gives it. Minto's structure for business communication — state the answer first, then the argument that supports it, rather than building up to a conclusion the audience must wait for (Minto, first published 1985, commonly cited as 1987) — matters more at this stage than at any earlier one: a board presentation that opens with methodology and works towards a recommendation loses the room before the recommendation arrives. The answer that belongs at the top is the honest one. Module 10's reporting discipline applies directly here: metrics are presented against the targets set at the outset, not reframed against an easier goal chosen because the original one looks unflattering, and no single flattering number — a strong engagement score, say — is allowed to stand in for a financial result that has not yet met its target.

The ask itself needs the same specificity Ward and Peppard's (2002) business-case structure demands of costs and benefits: not "more investment" but a named sponsor, a stated amount, and what it unlocks, following the same sponsorship logic Kotter's (1996) change model and Project Northstar's own COO sponsorship both depend on. A board that is shown the mixed results plainly, and asked for a specific, defensible commitment in response, is in a position to say yes with confidence; one shown only the good news, or an ask with no number attached, is not.