AI Leadership · Mock Exam 2

This is the harder of the two independent mock exams for the AI Leadership track — more multi-constraint stems and more FIRST / BEST / MOST cost-effective / TWO qualifiers. It is built from publicly available OpenAI learning objectives and is not an official OpenAI assessment or practice exam. Use it as your readiness gate: reach 80%+ here before you book the real OpenAI Academy assessment. All 50 items are new and are distinct from Mock Exam 1 and the domain-page questions.

Instructions

Time: 60 minutes, our own suggested pacing for practice, not an official OpenAI limit.
Items: 50 executive judgement scenarios, deliberately more demanding than Mock Exam 1. Each item states how many answers to select.
Selection: for single-answer items pick the one best option; for select-two items you must select both correct options and no others to score the item.
No guessing penalty: answer every question. There is no deduction for a wrong answer.
Target: aim for ≥ 80% raw before taking the real OpenAI Academy assessment; the 80% line matches the Academy badge threshold. Academy badges and pathway certificates of completion are not certifications.

Domain distribution

#	Domain	Items here
D1	Opportunity and Value	9
D2	Strategy and Roadmap	9
D3	Governance and Risk	9
D4	Adoption and Change Management	8
D5	Workforce and Enablement	7
D6	Measurement and ROI	8

Total: 9 + 9 + 9 + 8 + 7 + 8 = 50 items.

Readiness interpretation

This is an independent readiness indicator, never an official score.

Raw score	Band	What it means
under 70%	Keep learning	The harder framing has exposed real gaps; revisit the weak domains before re-attempting
70–79%	Building confidence	Close your two weakest domains, then re-sit this harder set
80–89%	Assessment ready	At or above the Academy badge threshold on the harder mock; you are in good shape
90%+	Strong readiness	Strong command under multi-constraint pressure; proceed to the Academy assessment

Take the mock exam

Two ways to use the questions below: the interactive mode runs a timed sitting one question at a time and ends with your score, a per-domain breakdown and a full correction; the review mode underneath lists every question with its options one per line and the answer hidden until you ask for it. If you have not sat Mock Exam 1 yet, start there as your diagnostic.

Interactive mode

Take the practice exam

50 questions · one at a time · 60-minute countdown · results with per-domain breakdown and full correction at the end. Your progress is saved in this browser if you leave the page.

Time limit Shuffle question order

OAI-LEADPractice exam

Question 1 of 50

--:--

All questions (review mode)

Options are listed one per line. The answer and explanation stay hidden until you click Show answer. Use the interactive mode above for a timed sitting.

Q1D1 · Opportunity and ValueSelect one
A CEO gives the leader one funding slot and three candidates: a high-value regulated use case with no review capacity yet, a medium-value low-risk case with a clear baseline, and a low-value flashy demo competitors are running. Constraints: build credibility, prove measurable value, stay within risk appetite. Which should be funded FIRST?
- A. The regulated high-value case, because upside is largest
- B. The medium-value low-risk case, because it satisfies all three constraints — provable value, within risk appetite, and it builds the track record the regulated case will later need
- C. The flashy demo, to answer competitor announcements
- D. None; wait for capacity to appear for the regulated case
Show answer
Answer: B.
Only the medium, low-risk, baselined case satisfies value, risk appetite and credibility-building simultaneously, which is what FIRST demands under all three constraints. A breaks the risk-appetite and review-capacity constraints. C fails the measurable-value test. D forfeits all near-term learning while waiting.
Q2D1 · Opportunity and ValueSelect one
Two sponsors present benefit cases. Sponsor A claims £5m from full automation with no assumptions listed. Sponsor B claims £1.2m–£1.8m with a stated automation rate, a counterfactual and residual review costs. The board asks which case the leader trusts MORE for funding. What is the best answer?
- A. Sponsor A, because the number is larger
- B. Sponsor B, because a transparent, driver-based range with a counterfactual is more defensible than a larger unsupported point estimate
- C. Neither, because both use estimates
- D. Average the two figures for balance
Show answer
Answer: B.
Credibility, not magnitude, decides trust; B's transparent range with counterfactual can be challenged and tracked. A is a hero number with no basis. C rejects the legitimate practice of estimating. D blends an unsupported figure into the credible one, contaminating it.
Q3D1 · Opportunity and ValueSelect two
A leader must separate cashable savings from freed capacity in a benefit case. Which two statements are TRUE about how to treat each? (Select two)
- A. Cashable savings reduce budgeted spend and can be booked once realised
- B. Freed capacity is only value if it is redeployed to higher-priority work, otherwise it is unrealised
- C. Freed capacity can always be reported as cash in the same year
- D. Cashable savings and freed capacity are interchangeable for board reporting
- E. Freed capacity should be counted at twice its hourly value to reflect motivation gains
Show answer
Answer: A and B.
Cashable savings hit the budget once realised; freed capacity only becomes value on redeployment. C books capacity as cash the P&L never sees. D conflates two fundamentally different benefit types. E fabricates a multiplier with no basis.
Q4D1 · Opportunity and ValueSelect one
A leader is pressured to fund the initiative with the single largest headline number. It concentrates the whole portfolio in one multi-year bet with unproven assumptions. Which reasoning should MOST guide the decision?
- A. Biggest number wins, since it maximises expected value
- B. Portfolio balance and risk-adjusted value matter more than a single headline figure; concentrating everything in one unproven bet is fragile
- C. Fund it and cancel all smaller work to focus
- D. Fund it because the sponsor is the most senior
Show answer
Answer: B.
Risk-adjusted value and portfolio balance beat a raw headline; an all-in unproven bet is fragile. A ignores probability and risk. C removes the learning and momentum that diversification provides. D decides by seniority, not merit.
Q5D1 · Opportunity and ValueSelect one
An opportunity looks strong on paper but the only data that would make it work is data the organisation is contractually barred from using for AI. What is the FIRST thing the leader should do?
- A. Proceed and seek forgiveness later if challenged
- B. Treat the data restriction as a disqualifying constraint and either redesign the use case without that data or drop it
- C. Assume the contract will be renegotiated in time
- D. Anonymise a sample and hope it is sufficient
Show answer
Answer: B.
A binding data restriction is a disqualifying constraint that must be resolved before value is even discussed. A invites contractual and reputational harm. C bets the case on an uncertain future renegotiation. D risks breaching the restriction with unverified anonymisation.
Q6D1 · Opportunity and ValueSelect one
A leader must decide the MOST cost-effective way to test whether a high-uncertainty opportunity is worth pursuing at all. Time and budget are tight. What is the best approach?
- A. Commission a full enterprise rollout to get real data
- B. Run a small, time-boxed pilot with a clear baseline and a decision gate, so a cheap test either kills or advances the idea
- C. Build a custom model first to be sure
- D. Decide on intuition to save the cost of testing
Show answer
Answer: B.
A small, baselined, time-boxed pilot is the cheapest way to generate a go/no-go decision on an uncertain idea. A spends enterprise cost before proving anything. C is the most expensive path for a high-uncertainty bet. D discards evidence entirely.
Q7D1 · Opportunity and ValueSelect one
The value of a proposed initiative is real but small, while a lower-scoring alternative would unlock learning that many future initiatives depend on. Capacity allows only one this quarter. What consideration should tip the decision?
- A. Always pick the higher immediate value regardless of context
- B. Weigh the option value of learning that unlocks a pipeline of future initiatives, not just this quarter's standalone benefit
- C. Pick whichever sponsor complains loudest
- D. Split the capacity and half-fund both
Show answer
Answer: B.
Strategic value includes option value — learning that unlocks a downstream pipeline can outweigh a small standalone benefit. A ignores that pipeline effect. C decides by politics. D under-funds both to the point neither proves anything.
Q8D1 · Opportunity and ValueSelect one
A leader reviews a benefit estimate that quietly assumes the freed hours convert to revenue at the team's full billing rate, though the team is salaried and not billable. What is the MOST accurate correction?
- A. Keep the billing-rate valuation; it is a standard proxy
- B. Value freed salaried time at its actual cost or redeployment value, not at a billing rate the team never charges
- C. Double the estimate to reflect efficiency gains
- D. Remove the benefit entirely because it is intangible
Show answer
Answer: B.
Freed salaried time is valued at cost or redeployment value, not an unearned billing rate. A applies a rate the team never charges, inflating the case. C compounds the error. D over-corrects and hides a genuine, if modest, benefit.
Q9D1 · Opportunity and ValueSelect one
A leader must present the single most defensible sizing artefact to a sceptical CFO for a mid-sized initiative. Which artefact best withstands challenge?
- A. A one-line 'expected annual saving' figure
- B. A driver-based model showing volumes, automation rate, unit values, a counterfactual, key assumptions, and a low-to-high range
- C. A competitor case study with no local data
- D. A vendor's ROI calculator output
Show answer
Answer: B.
A transparent driver-based model with assumptions, counterfactual and range is the most challengeable and therefore most defensible artefact. A hides all logic in one number. C imports external data with no local validity. D inherits a vendor's optimistic assumptions.
Q10D2 · Strategy and RoadmapSelect one
The board wants scale next quarter; the leader has one pilot with strong user enthusiasm but no baselined benefit, and a second pilot with a modest but baselined and positive benefit. Which is the correct sequencing decision under a prove-scale-embed model?
- A. Scale the enthusiastic pilot, since energy predicts success
- B. Scale the baselined-positive pilot and keep the enthusiastic one in prove until its benefit is baselined, because scale gates on proven benefit, not sentiment
- C. Scale both to satisfy the board's timeline
- D. Embed both directly and measure afterwards
Show answer
Answer: B.
Scale is gated on proven, baselined benefit, so only the baselined-positive pilot qualifies now. A scales on enthusiasm. C scales an unproven pilot to hit a date. D embeds unmeasured work everywhere before any evidence exists.
Q11D2 · Strategy and RoadmapSelect two
A leader is arbitrating platform versus point-tool decisions across the enterprise. Which two conditions most favour buying a separate POINT tool rather than extending the platform? (Select two)
- A. A regulated workflow has requirements the platform genuinely cannot meet
- B. A specialised need is durable, distinct, and forcing it onto the platform would degrade the whole platform
- C. A single team simply prefers a different user interface
- D. The platform already meets the need but procurement of an add-on feels faster
- E. A leader wants more vendors to reduce dependence on one supplier as a matter of principle
Show answer
Answer: A and B.
Point tools are justified for genuine, durable, distinct needs the platform cannot serve without harm. C is a preference, not a requirement. D duplicates a capability that already exists. E adds fragmentation on principle rather than need.
Q12D2 · Strategy and RoadmapSelect one
A three-year custom-build has consumed significant budget, has no production use, and now traces to none of the current stated priorities after a strategy refresh. The sponsor cites the money already spent. What should the leader recommend?
- A. Continue because so much has already been invested
- B. Stop or repurpose it; sunk cost is irrelevant and an initiative tracing to no priority should not draw further capital
- C. Increase the budget to finish it faster
- D. Re-write the priorities so the build appears aligned
Show answer
Answer: B.
The go-forward decision ignores sunk cost, and work tracing to no priority should not be funded further. A is the sunk-cost fallacy. C escalates commitment to a misaligned bet. D corrupts the strategy to protect the project.
Q13D2 · Strategy and RoadmapSelect one
A CFO offers two funding structures for a multi-year AI programme with uncertain later phases: (1) full multi-year budget approved now, or (2) stage-gated release tied to proven benefit. The leader wants both discipline and momentum. Which is the BEST choice and why?
- A. Option 1, so teams have certainty and never pause
- B. Option 2, because releasing capital gate by gate on proven benefit preserves discipline while still funding momentum for phases that earn it
- C. Neither; fund it all from next year's budget
- D. Option 1 but with no benefit tracking to reduce overhead
Show answer
Answer: B.
Stage-gated funding keeps discipline over uncertain later phases while still funding earned momentum. A commits all capital before uncertain phases prove out. C is an accounting deferral, not a funding model. D removes the very tracking that would justify continued spend.
Q14D2 · Strategy and RoadmapSelect one
A strategy draft is strong on opportunities and roadmap but silent on governance and measurement. A director says it will fail review. Which combination of additions MOST completes it as a leadership artefact?
- A. A longer opportunity list and more diagrams
- B. A governance approach (decision rights, controls) and a measurement approach (baselines, indicators, attribution) tied to the priorities
- C. A detailed model-selection appendix
- D. A marketing plan for the launch
Show answer
Answer: B.
Governance and measurement are the missing pillars of a complete strategy; adding both ties the draft together. A adds volume to a section already strong. C is execution detail. D is communications, not strategy substance.
Q15D2 · Strategy and RoadmapSelect one
Two business units each demand to own the enterprise AI platform decision. Delay is now blocking every downstream initiative. What is the BEST governance response for the strategy?
- A. Let the units keep negotiating until they agree
- B. Establish clear decision rights — a named owner and a defined process — so platform decisions are made and downstream work can proceed
- C. Give each unit its own platform to end the dispute
- D. Escalate every future AI decision to the full board
Show answer
Answer: B.
Clear decision rights with a named owner unblock the enterprise; strategy needs someone accountable to decide. A leaves the blockage unresolved. C creates the fragmentation the enterprise choice was meant to avoid. D overloads the board and slows everything.
Q16D2 · Strategy and RoadmapSelect one
A leader must decide whether to build, buy or partner for a capability that is important but not differentiating, where a mature product exists and no internal team can maintain a bespoke system. What is the BEST call?
- A. Build it in-house to retain full control
- B. Buy the mature product, because the capability is non-differentiating, a fit-for-purpose product exists, and the org cannot sustain a build
- C. Partner on a bespoke joint venture for a commodity need
- D. Do nothing until an internal team can be hired to build
Show answer
Answer: B.
Buy when the capability is non-differentiating, a product fits, and the org cannot maintain a build. A takes on maintenance the org cannot sustain for no differentiation. C over-engineers a commodity. D blocks value waiting to build what should be bought.
Q17D2 · Strategy and RoadmapSelect one
The roadmap embeds AI deeply into a core process before the pilot benefit is baselined, on the sponsor's confidence. A risk officer objects. Which principle most supports the objection?
- A. Embedding early always accelerates value
- B. Embedding is the final phase and should follow proven, scaled benefit, because deep integration multiplies both value and risk
- C. Confidence from a senior sponsor is sufficient evidence
- D. Baselines are only needed at year-end
Show answer
Answer: B.
Embed comes last because deep integration multiplies risk as well as value and should follow proof at scale. A ignores the risk multiplier of premature embedding. C substitutes authority for evidence. D delays measurement past the point it can guide the decision.
Q18D2 · Strategy and RoadmapSelect one
A leader is told the strategy must fix the exact models and prices for the next three years to give teams certainty. Given how fast the model lineup changes, what is the BEST way to handle this in the strategy?
- A. Hard-code specific model IDs and prices for three years
- B. State selection principles and a review cadence, so teams choose the best fit as the lineup evolves rather than binding to fast-decaying specifics
- C. Remove all mention of models from the strategy
- D. Pick the most expensive model to be safe long-term
Show answer
Answer: B.
Durable strategy states selection principles and a review cadence rather than binding to fast-changing model specifics. A locks in facts that will be stale within months. C omits guidance teams need. D wastes cost and still hard-codes a choice.
Q19D3 · Governance and RiskSelect one
After tightening controls, shadow use of personal AI accounts rises, not falls. Investigation shows every task — even trivial public-data summaries — now needs committee approval. What is the MOST effective governance change?
- A. Ban personal accounts more strictly and monitor devices
- B. Tier controls to risk so low-risk tasks are frictionless in the sanctioned workspace, removing the incentive to go around governance
- C. Approve the backlog faster with more reviewers
- D. Remove all controls to eliminate friction
Show answer
Answer: B.
Shadow AI grows when governance over-controls low-risk work; tiering controls removes the incentive to bypass it. A treats symptoms with surveillance while the cause remains. C scales a mis-designed process. D swings to no control, which is unacceptable for higher-risk work.
Q20D3 · Governance and RiskSelect two
A leader must map two requirements to enterprise controls: (1) automatically remove access when staff leave, and (2) keep EU-regulated data processed in the EU. Which two control choices correctly satisfy them respectively? (Select two)
- A. SCIM provisioning tied to the identity provider satisfies automatic access removal on departure
- B. Regional data residency satisfies keeping EU-regulated data processed in the EU
- C. Prompt caching satisfies automatic access removal
- D. A longer context window satisfies EU data residency
- E. An acceptable-use policy document satisfies both automatically
Show answer
Answer: A and B.
SCIM automates deprovisioning and regional residency keeps processing in-region — each maps to its requirement. C is a performance feature unrelated to access. D confuses model capacity with data location. E is unenforced policy, not an automatic control.
Q21D3 · Governance and RiskSelect one
A leader must decide whether a proposed control set is 'enabling governance' or 'governance theatre'. The set requires a full risk-committee review for every prompt, produces long queues, and has never rejected a case. Which judgement is MOST accurate?
- A. It is strong governance because every case is reviewed
- B. It is theatre: high friction, no risk-based triage, and no evidence it changes outcomes; it should be replaced with tiered, risk-based controls
- C. It is fine as long as the committee meets weekly
- D. It is enabling because it slows adoption to a safe pace
Show answer
Answer: B.
Reviewing everything with no triage and no rejections adds friction without managing risk — the definition of theatre. A mistakes activity for control. C tunes cadence, not design. D reframes a bug (slow adoption) as a feature.
Q22D3 · Governance and RiskSelect one
A vendor offers a durable cloud-agent capability but only with US data residency and no zero-data-retention option. The organisation has one workload that must stay in the EU and another with no residency constraint. What is the BEST governance decision?
- A. Use the capability for both workloads to standardise
- B. Use it only for the unconstrained workload and choose a compliant path for the EU-residency workload
- C. Use it for the EU workload since agents are more secure
- D. Reject the capability entirely for all workloads
Show answer
Answer: B.
Match the capability to workloads whose constraints it satisfies: the unconstrained one, not the EU-residency one. A puts EU-regulated data into a US-only, no-ZDR service. C directly violates the residency requirement. D forgoes value where the capability is perfectly compliant.
Q23D3 · Governance and RiskSelect one
A leader is assembling a risk taxonomy and wants the category that most directly covers the danger of a confident but wrong AI output driving a business decision. Which category is it?
- A. Data-residency risk
- B. Output-quality risk, covering hallucination and unsafe or biased outputs reaching decisions without adequate review
- C. Vendor-lock-in risk
- D. Licence-cost risk
Show answer
Answer: B.
A confident-but-wrong output reaching a decision is squarely output-quality (accuracy/hallucination) risk. A concerns where data is processed. C concerns supplier dependence. D concerns spend, not correctness.
Q24D3 · Governance and RiskSelect one
Legal, worried about training on inputs, proposes banning AI outright. The leader knows the sanctioned Business/Enterprise workspace does not train on business data by default and offers SSO, RBAC and audit logs. What is the MOST balanced recommendation?
- A. Ban AI as legal requests, to be safe
- B. Channel usage into the governed workspace whose default no-training, SSO, RBAC and audit-log controls address the concern, rather than banning and driving shadow use
- C. Allow personal accounts but ask staff to be careful
- D. Disable audit logs to reduce data collected
Show answer
Answer: B.
The governed workspace's default no-training plus SSO, RBAC and audit logs directly address the concern while enabling value. A drives shadow use and forfeits benefit. C is the ungoverned path legal fears. D removes a control the organisation needs for assurance.
Q25D3 · Governance and RiskSelect one
An organisation must both restrict sensitive projects to approved groups and be able to prove after the fact who accessed what. Which pairing of controls MOST directly meets both needs together?
- A. Prompt caching plus a longer context window
- B. RBAC with SSO to enforce who can access, plus compliance and audit logs to prove access after the fact
- C. A shared password plus a monthly reminder email
- D. A larger licence pool plus a satisfaction survey
Show answer
Answer: B.
RBAC/SSO enforces access and audit logs provide the after-the-fact record — together they meet both needs. A is unrelated to access or audit. C is insecure and unauditable. D addresses neither enforcement nor evidence.
Q26D3 · Governance and RiskSelect one
A leader is asked to approve an agentic workload that will act with limited human oversight on customer records. Which condition should the leader treat as a hard prerequisite before approval?
- A. That the agent uses the newest available model
- B. That data-handling, residency, retention and a human-oversight or review mechanism appropriate to the risk are all in place first
- C. That the workload is faster than the manual process
- D. That the vendor's demo was impressive
Show answer
Answer: B.
Higher-autonomy work on sensitive records requires data-handling, residency/retention and risk-appropriate oversight as prerequisites. A is a model choice, not a control. C is a benefit, not a safeguard. D is not evidence of governance readiness.
Q27D3 · Governance and RiskSelect one
The governance model routes every AI decision — trivial or critical — to a single central committee, and it is now the bottleneck for the whole programme. What redesign BEST preserves control while removing the bottleneck?
- A. Keep central review but meet daily
- B. Delegate low-risk decisions to pre-approved patterns and local owners, reserving central review for high-risk cases
- C. Abolish the committee and let anyone decide anything
- D. Route decisions to whichever executive is free that day
Show answer
Answer: B.
Tiered delegation — local owners and pre-approved patterns for low risk, central review for high risk — preserves control and removes the bottleneck. A speeds a centralised choke point without fixing it. C removes control. D introduces inconsistent, unaccountable decisions.
Q28D4 · Adoption and Change ManagementSelect one
A rollout crossed the early enthusiasts but stalled at roughly a third of the target population. Two options remain within budget: a second broadcast launch email, or manager-led embedding with role-specific use cases. Which is MOST likely to move the mainstream, and why?
- A. The broadcast email, because reach is what is missing
- B. Manager-led embedding with role-specific use cases, because the mainstream adopts through local leadership and relevance, not repeated broadcasts
- C. Neither; the tool has reached its natural ceiling
- D. A prize for the most active users
Show answer
Answer: B.
The chasm is crossed by manager modelling and role relevance, not by repeating a broadcast that already failed. A repeats an approach the mainstream ignored. C concedes defeat prematurely. D rewards volume and invites gaming rather than valuable use.
Q29D4 · Adoption and Change ManagementSelect two
A leader is choosing metrics to judge whether adoption is genuinely progressing across the organisation. Which two are the STRONGEST signals? (Select two)
- A. The share of the target population using AI regularly in core workflows
- B. Breadth of adoption across functions, not concentration in one enthusiast team
- C. The total number of licences purchased
- D. The number of people who opened the launch email
- E. The raw count of prompts submitted, regardless of purpose
Show answer
Answer: A and B.
Regular use in core workflows and breadth across functions are the strongest adoption signals. C counts access. D counts a one-off open. E rewards volume that may be meaningless activity.
Q30D4 · Adoption and Change ManagementSelect one
A programme celebrates '6,000 licences issued' and '2 million prompts submitted' as proof of success. The CEO asks the leader whether the programme is truly working. What is the MOST accurate answer?
- A. Yes; both numbers are large and rising
- B. Unknown from those figures alone, because licences measure access and raw prompt counts measure activity — neither shows valuable, recurring use tied to real work
- C. Yes, because prompt volume proves engagement
- D. No; the numbers are too small to matter
Show answer
Answer: B.
Access and raw activity counts cannot confirm valuable, recurring, work-embedded use. A and C mistake access and volume for adoption. D misjudges scale; the issue is the metric type, not the magnitude.
Q31D4 · Adoption and Change ManagementSelect one
A leader inherits a champion network that is all senior executives, meets quarterly, and has no time allocated. Adoption in the field is flat. What single change would MOST improve it?
- A. Add more executives to the network
- B. Recruit embedded, credible peers in each function and give them real time and recognition to help colleagues
- C. Meet monthly instead of quarterly
- D. Rename it a 'centre of excellence'
Show answer
Answer: B.
Effective champions are embedded peers with real time and recognition, close to the daily work. A deepens the wrong composition. C changes cadence, not credibility or capacity. D is a relabel with no substance.
Q32D4 · Adoption and Change ManagementSelect one
Managers in one division treat AI adoption as optional and never use it themselves; their teams have the lowest usage. Another division's managers use it daily and expect it in routines; their teams lead adoption. What does this MOST strongly imply for the leader's plan?
- A. The difference is random and cannot be influenced
- B. Manager behaviour is a primary driver of team adoption, so making adoption an explicit, supported managerial expectation should be central
- C. The leading division simply has better people
- D. IT should force usage in the lagging division
Show answer
Answer: B.
The contrast points to manager modelling and expectation as the primary lever, so it should anchor the plan. A ignores a clear pattern. C attributes to talent what the evidence attributes to leadership behaviour. D treats a management issue as a tooling mandate.
Q33D4 · Adoption and Change ManagementSelect one
A survey shows two barriers: 'I don't know what to use it for' (majority) and 'I don't trust the output' (a vocal minority). Budget allows one primary intervention this quarter. Where should the leader focus FIRST for the biggest adoption gain?
- A. The trust concern, because it is voiced most loudly
- B. The relevance barrier, with a role-specific use-case library, because it affects the majority who are blocking mainstream adoption
- C. Neither; run another survey
- D. Buy more licences to widen access
Show answer
Answer: B.
The majority barrier — 'what would I use it for' — gates mainstream adoption, so a role-specific use-case library yields the biggest gain first. A serves the loudest minority, not the largest blocker. C delays action. D adds access nobody is using.
Q34D5 · Workforce and EnablementSelect one
A leader must design enablement for 2,500 staff with very different roles, under a fixed budget. Which design gives the BEST capability return per pound spent?
- A. One identical mandatory course for everyone
- B. Role-and-level segmented paths with hands-on practice on real tasks, so each group gains the capability its work needs
- C. The longest available course for all staff
- D. A voluntary open library with no structure
Show answer
Answer: B.
Segmented, applied paths concentrate spend on the capability each role actually needs, maximising return. A spends on irrelevance for most. C wastes time on unneeded depth. D leaves capability to chance.
Q35D5 · Workforce and EnablementSelect two
A director wants marketing to say the workforce is 'OpenAI certified' and 'certification-ready' after completing Academy pathways. Which two statements must the leader insist are TRUE and used instead? (Select two)
- A. Academy badges and pathway certificates of completion are not certifications
- B. The accurate framing is that staff completed OpenAI Academy courses and assessments, earning badges
- C. Completing pathways guarantees eligibility for a future certification
- D. The team can be described as 'OpenAI certified' internally
- E. Badges are equivalent to a psychometrically validated certification exam
Show answer
Answer: A and B.
OpenAI states badges and pathway certificates are not certifications; the accurate claim is course/assessment completion earning badges. C contradicts OpenAI's explicit wording. D still calls a badge a certification. E overstates a completion badge as a validated exam.
Q36D5 · Workforce and EnablementSelect one
As AI automates routine drafting, a leader redesigns a knowledge-worker role. Which redesign MOST protects value and morale together?
- A. Cut the role and redistribute the drafting to fewer people
- B. Shift the role toward reviewing and improving AI output, owning exceptions, and higher-judgement work, with reskilling support
- C. Keep the role identical and forbid AI to preserve jobs
- D. Add unrelated manual busywork to justify the headcount
Show answer
Answer: B.
Redesigning toward review, exceptions and judgement, backed by reskilling, protects both value and morale. A cuts capability without redesign. C refuses the change and forgoes value. D manufactures make-work that fools no one.
Q37D5 · Workforce and EnablementSelect one
Anxiety is spreading: staff hear rumours of layoffs 'because of AI'. Leadership has not decided any cuts. What is the MOST effective communication approach?
- A. Say nothing until any decision is final
- B. Communicate honestly and early about how roles will evolve, what reskilling is available, and where human judgement stays essential, without over-promising
- C. Promise no role will ever change
- D. Tell worried staff to focus on their work and stop speculating
Show answer
Answer: B.
Honest, early communication with concrete reskilling and a realistic picture builds trust without over-promising. A lets rumour fill the silence. C makes an unkeepable promise. D dismisses legitimate concern and erodes trust.
Q38D5 · Workforce and EnablementSelect one
A leader must route three groups to appropriate OpenAI Academy learning: API developers, workplace users, and Codex-using engineers. Which routing is MOST role-appropriate?
- A. Send all three groups to the AI Leadership course
- B. Route developers to Build-with-AI API courses, workplace users to Apply-AI-at-work courses, and engineers to the Codex courses
- C. Let each person choose any course at random
- D. Send everyone to the shortest course to save time
Show answer
Answer: B.
Matching each group to the product-relevant Academy track is the core of capability-based routing. A misroutes technical staff into a strategy course. C produces unaligned capability. D optimises time over relevance.
Q39D5 · Workforce and EnablementSelect one
Finance will fund enablement only if the leader shows it is not 'nice-to-have'. Which argument is the STRONGEST link between enablement and realised value?
- A. Enablement improves employee satisfaction scores
- B. Unused licences produce zero value; enablement is what converts purchased access into the adoption and benefit the business case assumed
- C. Enablement is standard practice in the industry
- D. It gives HR something to administer
Show answer
Answer: B.
Enablement is the mechanism that turns licence spend into the adoption the business case depends on, protecting the larger investment. A is a soft benefit, not the value link. C argues by convention. D trivialises it as administration.
Q40D6 · Measurement and ROISelect one
A team wants to claim a 25% time saving from an AI tool but only started measuring after launch, comparing this month to last month. Multiple process changes also happened. What is the MOST defensible way to salvage a credible claim?
- A. Report the 25% as-is; month-over-month is close enough
- B. Reconstruct or run a proper baseline, control for the other process changes, and report a conservative range rather than the raw month-over-month figure
- C. Drop the claim entirely; nothing can be recovered
- D. Round the figure up to 30% for a stronger headline
Show answer
Answer: B.
A credible claim needs a proper baseline, controls for confounders, and a conservative range. A ignores confounding process changes. C over-corrects and discards recoverable evidence. D inflates an already-suspect number.
Q41D6 · Measurement and ROISelect two
A leader must pick the two metrics that best show, early and honestly, whether an AI initiative is on track before annual financials land. Which two? (Select two)
- A. Recurring active use in the targeted workflow (a leading indicator)
- B. Task cycle-time change paired with a quality measure so speed is not bought at the cost of quality
- C. The audited year-end financial statement
- D. The board's overall confidence rating
- E. Total prompts submitted regardless of outcome
Show answer
Answer: A and B.
Recurring use and quality-paired cycle-time are early, honest signals of progress. C is a lagging annual outcome. D is sentiment, not measurement. E rewards volume, not value or quality.
Q42D6 · Measurement and ROISelect one
An AI assist cut average handling time 30%, but customer satisfaction and reopened-ticket rates worsened. Finance still wants to book the full time-saving as benefit. What should the leader insist on?
- A. Book the full time-saving; quality is a separate issue
- B. Net the quality deterioration against the speed gain, because a faster-but-worse process may deliver little or negative real benefit
- C. Ignore satisfaction; it is subjective
- D. Cut handle-time targets further to lock in the gain
Show answer
Answer: B.
Speed and quality must be judged together; worsening quality can erase or reverse a time-saving. A books a gross figure that quality losses may cancel. C dismisses a real quality signal. D worsens the quality problem it should address.
Q43D6 · Measurement and ROISelect two
A pilot ran on a free trial. Before approving scale, the leader must model the true run cost. Which two cost drivers MOST determine the at-scale bill? (Select two)
- A. Usage-based consumption (for example token or per-call charges) that scales with volume
- B. Per-seat licence cost multiplied across the full user population
- C. The number of pilot participants who gave positive feedback
- D. The brand of laptops the team uses
- E. The number of meetings held during the pilot
Show answer
Answer: A and B.
Consumption charges and per-seat licensing scale with volume and headcount, driving the at-scale bill. C is a sentiment count. D and E are irrelevant to unit economics.
Q44D6 · Measurement and ROISelect one
A region using AI grew 15%, but it also launched a new product, gained two competitors' customers, and rode a rising market. The leader must brief the board on AI's contribution. What is the MOST honest framing?
- A. Credit the full 15% to AI to make the strongest case
- B. Decompose the growth, isolate a defensible AI contribution with stated assumptions, and present it as a conservative range with the confounders named
- C. Credit nothing to AI to avoid any risk of overclaiming
- D. Say the effect is 'clearly significant' without numbers
Show answer
Answer: B.
Honest attribution decomposes growth, names confounders, and reports a conservative, assumption-backed range. A overclaims by ignoring the product launch and market lift. C hides genuine benefit. D offers a vague claim that cannot be defended or steered by.
Q45D6 · Measurement and ROISelect one
The board wants a single quarterly line that captures return without overclaiming. Which construction is BEST?
- A. 'AI is delivering huge value across the business.'
- B. 'Against baseline, estimated net benefit this quarter is £X–£Y, attributed conservatively after controlling for known confounders, at a run cost of £Z.'
- C. 'We are the most AI-advanced firm in our sector.'
- D. 'Adoption is high, so return must be strong.'
Show answer
Answer: B.
A defensible board line states baseline, net range, conservative attribution and run cost. A is a hero claim. C is a competitive boast, not a return figure. D infers return from adoption without measuring benefit or cost.
Q46D4 · Adoption and Change ManagementSelect one
A leader must justify continuing an adoption-support budget when a finance partner argues the tool 'sells itself'. Usage sits at a third of the target population and has plateaued. Which argument is MOST compelling?
- A. Adoption support is standard practice and should continue
- B. The plateau shows the mainstream has not crossed, and crossing it requires active enablement — manager expectations, champions and role-specific use cases — which is exactly what the budget funds
- C. The tool is defective and needs replacing, so the budget is moot
- D. Withdraw support and let adoption find its natural level
Show answer
Answer: B.
A plateau at a third signals the mainstream has not crossed the chasm, and active enablement is what moves them — the budget's purpose. A argues by convention. C blames the tool for a change gap. D abandons the majority to a level that will not rise on its own.
Q47D6 · Measurement and ROISelect two
A sponsor presents only leading indicators after two quarters and asks the board to declare victory. Before any success claim, which two additional evidence types should the leader require? (Select two)
- A. Lagging outcome measures tied to the stated priority (for example realised benefit)
- B. Net benefit after full run cost, attributed conservatively for confounders
- C. A larger count of prompts submitted across the organisation
- D. A survey of how satisfied the sponsor is with the tool
- E. The number of licences issued since launch
Show answer
Answer: A and B.
A proven success needs lagging outcomes tied to the priority and net benefit after cost with conservative attribution. C is a raw activity count. D is sentiment, not outcome evidence. E measures access, not delivered value.
Q48D6 · Measurement and ROISelect one
Two initiatives report the same gross benefit. Initiative X uses a high-cost model at high volume; initiative Y uses a low-cost model for the same task quality. The leader must advise on portfolio value. What matters MOST?
- A. They are equal because gross benefit is identical
- B. Net benefit and unit economics differ: Y likely delivers more net value by achieving the same quality at lower run cost
- C. X is better because it uses a more capable model
- D. Neither; model choice is irrelevant to value
Show answer
Answer: B.
With equal gross benefit and quality, the lower run cost makes Y the higher net-value choice — unit economics decide. A ignores cost. C equates model capability with value even when quality is matched at lower cost. D denies the clear cost difference.
Q49D4 · Adoption and Change ManagementSelect one
A leader launches AI across ten country offices at once. Three months later, usage is strong in two offices whose leaders actively use and expect the tool, and near zero in the rest. What is the FIRST corrective move?
- A. Send a stronger global mandate email from headquarters
- B. Engage the local office leaders in the low-usage sites to model use and set team expectations, since local leadership drives mainstream adoption
- C. Assume the tool is culturally unsuited to the low-usage offices and withdraw it there
- D. Buy more licences for the successful offices
Show answer
Answer: B.
The success/failure split tracks whether local leaders model and expect use, so engaging low-usage office leaders is the first fix. A repeats a broadcast the mainstream ignores. C abandons offices without addressing the leadership gap. D rewards the already-adopting sites and ignores the laggards.
Q50D5 · Workforce and EnablementSelect one
A leader must build a capability map before routing training. A team's members range from AI-sceptical novices to a few power users already building workflows. What is the BEST enablement design for this single team?
- A. Give the whole team the beginner course so no one is lost
- B. Differentiate by current capability: foundational, applied practice for novices and advanced or peer-leadership roles for the power users
- C. Give the whole team the advanced course to stretch them
- D. Train only the power users and let them teach the rest ad hoc
Show answer
Answer: B.
Capability mapping tailors enablement to current levels and can turn power users into peer leaders. A bores and under-serves the advanced members. C loses the novices. D relies on unstructured cascade that usually decays and leaves novices unsupported.