AI Leadership · Mock Exam 1

This is a full-length, domain-weighted independent mock exam for the AI Leadership track. It is built from publicly available OpenAI learning objectives and is not an official OpenAI assessment or practice exam. Use it as your diagnostic: sit it first to find your two weakest domains. All 50 items are new and do not repeat the domain-page questions.

Instructions

Time: 60 minutes. This is our own suggested pacing for practice, not an official OpenAI limit.
Items: 50, executive judgement scenarios. Each item states how many answers to select — most are single-answer, some are select-two.
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. Remember that 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	Revisit the heaviest domains (Opportunity, Strategy, Governance) before re-attempting
70–79%	Building confidence	Close the gaps in your two weakest domains, then re-sit
80–89%	Assessment ready	You are at or above the Academy badge threshold; tidy up remaining weak spots
90%+	Strong readiness	Broad command of the material; proceed to the Academy assessment with confidence

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. See the assessment model page for how the real Academy assessment differs.

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 COO lists twelve candidate AI use cases and asks which to fund first. Nine promise 'efficiency', two promise 'innovation', one names a specific stated priority: cut invoice-processing cycle time by 30%. What should the leader anchor the shortlist on?
- A. Fund the two innovation ideas first, because differentiation matters more than efficiency
- B. Rank candidates by how directly and measurably they advance a stated business priority, starting with the invoice-cycle case
- C. Fund all twelve as small pilots so nothing is missed
- D. Rank purely by lowest implementation cost so early wins are cheap
Show answer
Answer: B.
Value is judged by contribution to a stated priority, and only the invoice case names one with a measurable target, so it anchors the shortlist. A chases novelty over traceability. C spreads scarce attention across unqualified ideas and produces no learning. D optimises cost, not value, and would fund cheap-but-irrelevant work.
Q2D1 · Opportunity and ValueSelect one
A sponsor claims an assistant will save 40,000 agent-hours a year, but the contact centre is already running below capacity and cannot reduce headcount this year. How should the benefit be characterised to the board?
- A. As a £-cashable saving equal to 40,000 hours times the loaded hourly rate
- B. As freed capacity that is only value if it is redeployed to higher-priority work, not as a cash saving this year
- C. As a headcount reduction to be realised immediately
- D. As innovation revenue, since freed time enables new products
Show answer
Answer: B.
Time saved in an under-utilised team is capacity, not cash, until it is deliberately redeployed; reporting it honestly protects credibility. A books a saving the P&L will never show. C promises cuts the plan explicitly rules out. D relabels capacity as revenue with no product or demand behind it.
Q3D1 · Opportunity and ValueSelect one
Two initiatives compete for one slot: a customer-facing chatbot that could lift NPS but touches regulated advice, and an internal meeting-summary tool with modest but certain time savings and low risk. The organisation has no AI track record yet. Which is the stronger FIRST bet?
- A. The chatbot, because customer-facing impact is more strategic
- B. The internal summary tool, because a low-risk, provable early win builds capability and credibility before higher-stakes work
- C. Neither; wait until a full governance framework is signed off
- D. Both at once, to maximise coverage
Show answer
Answer: B.
With no track record, sequencing a low-risk, quickly provable win earns the credibility and capability needed for regulated work later. A leads with the highest-risk option before the organisation can govern it. C freezes all progress pending perfection. D ignores capacity limits and dilutes the first proof point.
Q4D1 · Opportunity and ValueSelect two
A leader is screening a proposed use case for whether it is a strong EARLY AI candidate. Which two attributes most strengthen the case? (Select two)
- A. The task is high-volume and repeatable, so gains compound and are easy to measure
- B. A knowledgeable human already reviews the output, so errors are caught before impact
- C. The output is a legally binding decision with no human in the loop
- D. The use case is novel and has no comparable process to benchmark against
- E. The task requires access to data the organisation is not permitted to use
Show answer
Answer: A and B.
High volume makes benefit measurable and material, and an existing human review point makes early error rates safe to absorb. C removes the safeguard that makes early adoption defensible. D removes the baseline needed to prove benefit. E introduces a compliance blocker that disqualifies the case rather than strengthening it.
Q5D1 · Opportunity and ValueSelect one
A business-unit head insists their idea 'will obviously pay for itself' and refuses to state assumptions. The leader must decide whether to advance it to a funding gate. What is the most defensible action?
- A. Advance it on the strength of the sponsor's conviction
- B. Require a driver-based estimate with explicit assumptions, a counterfactual and a range before it can pass the gate
- C. Reject the idea outright for lack of a number
- D. Ask finance to invent a placeholder number so the gate can proceed
Show answer
Answer: B.
A fundable case needs a transparent, driver-based estimate with assumptions and a range so it can be challenged and tracked; that is the gate's whole purpose. A funds on conviction alone. C discards a potentially good idea rather than pricing it. D fabricates evidence, which corrodes the entire portfolio's credibility.
Q6D1 · Opportunity and ValueSelect one
An AI opportunity portfolio contains only large, multi-year transformation bets. A board member worries about the shape. What is the most balanced correction?
- A. Add several quick, low-risk wins alongside the big bets to fund learning and momentum while the large bets mature
- B. Cancel the big bets and do only quick wins
- C. Add even larger bets to raise the ambition
- D. Keep the portfolio unchanged; size signals seriousness
Show answer
Answer: A.
A healthy portfolio balances near-term provable wins against longer strategic bets, so it delivers learning and momentum without betting everything on slow maturation. B abandons strategic ambition. C worsens the concentration risk. D ignores the legitimate imbalance the board flagged.
Q7D1 · Opportunity and ValueSelect one
During opportunity sizing, a sponsor's estimate assumes AI handles 100% of cases with zero review. Experienced staff say realistically 60% could be automated with review on the rest. How should the leader treat the gap?
- A. Use the 100% figure to keep the business case attractive
- B. Re-base the estimate on the realistic automation rate and the cost of the review that remains, then present a range
- C. Split the difference at 80% without evidence
- D. Defer the case until the technology can reach 100%
Show answer
Answer: B.
Credible sizing uses the realistic automation share and accounts for residual human review, expressed as a range. A inflates the case on an unachievable assumption. C picks a number with no basis. D waits for a level of autonomy that may never be appropriate for the task.
Q8D1 · Opportunity and ValueSelect one
A functional leader wants to pursue a flashy generative use case that competitors are publicising, but it maps to none of the company's four stated priorities. What should the AI leader recommend?
- A. Pursue it to keep pace with competitors' announcements
- B. Decline or park it, because an initiative that traces to no stated priority should not consume scarce delivery capacity
- C. Reframe a priority around the use case so it appears aligned
- D. Fund it as a marketing expense outside the AI portfolio
Show answer
Answer: B.
Discipline means initiatives earn a place by advancing a stated priority; one that traces to nothing should be parked. A lets competitor PR set the agenda. C reverse-engineers strategy to justify a pet project. D just hides the same misalignment under a different budget line.
Q9D1 · Opportunity and ValueSelect one
A team reports that its AI pilot 'saved 20 minutes per report'. The leader wants to know whether that is real value. Which follow-up question matters MOST first?
- A. How much did the pilot cost to run?
- B. How was the 20 minutes measured, and against what baseline before AI?
- C. How many people liked using the tool?
- D. Which model version produced the reports?
Show answer
Answer: B.
Without a measured baseline, a per-report saving is an assertion, so the measurement method and pre-AI baseline come first. A matters, but a benefit with no baseline cannot be netted against cost meaningfully. C measures sentiment, not value. D is an implementation detail irrelevant to whether the benefit is real.
Q10D2 · Strategy and RoadmapSelect one
The executive team asks for an AI roadmap that reaches enterprise-wide scale in one quarter. Only one pilot has run and its benefit is not yet baselined. What sequencing should the leader defend?
- A. Commit to the one-quarter scale plan to show ambition
- B. Follow prove then scale then embed, gating the scale phase on a baselined, positive benefit from the pilot
- C. Skip proving and scale directly, since the pilot 'felt successful'
- D. Embed AI into every process first, then measure later
Show answer
Answer: B.
Scale should be gated on proven, baselined benefit, so prove then scale then embed is the defensible sequence. A promises a timeline the evidence cannot support. C scales on sentiment. D embeds unproven capability everywhere, multiplying risk before any evidence exists.
Q11D2 · Strategy and RoadmapSelect one
Four functions each want to buy a different AI point tool. The CIO warns of overlapping spend, fragmented governance and duplicated integration. What is the BEST strategic response?
- A. Let each function buy independently to preserve autonomy and speed
- B. Assess whether a shared platform can meet most needs with common governance, and reserve point tools for genuinely distinct requirements
- C. Ban all AI tools until a single mandated vendor is chosen for everything
- D. Standardise on whichever tool the largest function already prefers
Show answer
Answer: B.
A platform-first assessment captures shared governance and integration economics while still allowing justified point tools. A produces the exact fragmentation the CIO flagged. C blocks all value while a monolith is chosen. D lets size, not fit, dictate the standard and may starve smaller functions of what they need.
Q12D2 · Strategy and RoadmapSelect two
A leader is deciding whether to BUILD a bespoke capability rather than buy a product. Which two conditions most justify building? (Select two)
- A. The capability is a genuine source of competitive differentiation the company must own
- B. No commercial product meets a well-understood, durable requirement, and the organisation can sustain the capability
- C. Building looks cheaper than a subscription in the first month
- D. The team wants the experience of building something in-house
- E. A capable product exists but procurement is slow this quarter
Show answer
Answer: A and B.
Build when the capability is differentiating and must be owned, or when no product fits a durable need and the org can maintain it. C ignores total cost of ownership over the lifetime. D lets curiosity drive a strategic decision. E is a temporary procurement friction, not a reason to take on permanent build-and-maintain cost.
Q13D2 · Strategy and RoadmapSelect one
An AI strategy draft lists many initiatives but never states how they connect to the company's priorities or to each other. A director calls it 'a shopping list'. What is the most important fix?
- A. Add more initiatives to look comprehensive
- B. Establish explicit traceability from each initiative to a stated priority and sequence them into a coherent roadmap
- C. Sort the list alphabetically for clarity
- D. Remove the priorities section to avoid constraining the teams
Show answer
Answer: B.
A strategy, not a list, requires each initiative to trace to a priority and to sit in a deliberate sequence. A adds volume, not coherence. C is cosmetic. D deletes the very anchor that would turn the list into a strategy.
Q14D2 · Strategy and RoadmapSelect one
A CFO will fund AI only if the leader picks one funding model. The programme spans several years, needs disciplined stage-gates, and must not let unproven bets run unchecked. Which model fits best?
- A. A single up-front lump sum released to teams at kickoff
- B. Staged funding released gate by gate as each phase proves benefit, so unproven work does not draw further capital
- C. Unlimited draw-down against a central pool with no gates
- D. Charge every cost to next year so this year looks free
Show answer
Answer: B.
Stage-gated funding matches capital release to proven progress, giving the discipline the CFO requires. A hands over all capital before any proof. C removes the control entirely. D is an accounting trick that hides, rather than manages, cost.
Q15D2 · Strategy and RoadmapSelect one
Halfway through the year, a scaled initiative is missing its benefit target and its assumptions have proven wrong. The sponsor wants to press on 'because we've already invested'. What should the leader advise?
- A. Continue, because stopping would waste the money already spent
- B. Re-evaluate against current evidence and be willing to stop or reshape it, treating sunk cost as irrelevant to the go-forward decision
- C. Double the budget to force the target
- D. Hide the miss until year-end reporting
Show answer
Answer: B.
Good portfolio discipline judges the go-forward case on current evidence and ignores sunk cost. A is the sunk-cost fallacy. C throws more capital at a failing set of assumptions. D conceals information the organisation needs to reallocate.
Q16D2 · Strategy and RoadmapSelect one
A leader is asked to produce an initial AI strategy draft after the AI Leadership course. Which content most belongs in that draft?
- A. A detailed vendor contract and signed budget approvals
- B. Stated priorities, a prioritised set of opportunities, a phased roadmap, governance approach, and how value will be measured
- C. A finished model architecture and API specification
- D. A list of every AI news article from the last quarter
Show answer
Answer: B.
An initial strategy draft aligns priorities, opportunities, sequencing, governance and measurement — exactly the leadership artefact the course targets. A and C are downstream execution detail, not strategy. D is background noise, not a plan.
Q17D2 · Strategy and RoadmapSelect two
Two departments propose nearly identical AI initiatives with separate budgets and vendors. Before funding, the leader wants to consolidate them well. Which two conditions should the consolidated initiative satisfy? (Select two)
- A. A single named owner and shared budget, so the effort is not duplicated across two teams
- B. One shared governance and measurement approach, so learning and controls are common
- C. Two parallel vendors kept in competition throughout to hedge bets
- D. Separate success metrics per department to preserve autonomy
- E. A doubled budget to reflect the two original requests combined
Show answer
Answer: A and B.
Consolidation works when there is one owner and shared budget plus common governance and measurement, capturing shared learning and avoiding duplicate spend. C reintroduces the duplication being removed. D fragments the metrics that should now be shared. E inflates cost rather than consolidating it.
Q18D2 · Strategy and RoadmapSelect one
A leader must decide whether to build on a general-purpose AI platform or buy a narrow point solution for a single, unusual regulated workflow that the platform cannot handle. Everything else in the org is well served by the platform. What is the best call?
- A. Force the regulated workflow onto the platform anyway to keep one vendor
- B. Keep the platform for the majority and buy the point solution only for the genuinely distinct regulated workflow
- C. Replace the platform with point tools everywhere for consistency
- D. Drop the regulated workflow from scope because it complicates the standard
Show answer
Answer: B.
Platform-for-the-many plus point-tool-for-the-genuinely-distinct is the balanced pattern. A bends a critical regulated need to a tool that cannot meet it. C throws away platform economics. D abandons a required workflow to protect tidiness.
Q19D6 · Measurement and ROISelect one
A programme dashboard shows only monthly licence counts and total prompts. The board asks whether these prove the programme is delivering value. What is the leader's most accurate response?
- A. Yes, both are rising so value is clearly being delivered
- B. No; those are access and activity counts, and value requires baselined benefit measures, quality-paired outcomes and net cost — none of which the dashboard shows
- C. Yes, because high prompt volume proves engagement equals value
- D. The dashboard is fine as long as the numbers keep rising
Show answer
Answer: B.
Licences and prompt counts measure access and activity, not value; value needs baselined benefit, quality-paired outcomes and net cost. A and C mistake access and volume for delivered value. D treats a rising vanity metric as sufficient evidence of return.
Q20D3 · Governance and RiskSelect one
Staff report that summarising a public press release inside the sanctioned workspace still requires monthly risk-committee sign-off, so some have started pasting content into personal ChatGPT accounts. What is the ROOT problem a leader should fix?
- A. Staff are non-compliant and should be disciplined
- B. Governance applies one heavy control to all risk levels, pushing low-risk work into ungoverned shadow tools; controls should be tiered to risk
- C. The workspace is too slow and should be replaced
- D. Public documents should never be summarised by AI
Show answer
Answer: B.
Uniform heavy controls on low-risk tasks create shadow AI; the fix is tiering controls to risk so trivial tasks are frictionless. A punishes a symptom of bad design. C misdiagnoses speed for a policy problem. D bans a plainly safe activity.
Q21D3 · Governance and RiskSelect one
A firm must ensure that when an employee leaves, all access to the AI workspace is removed automatically from the corporate identity system. Which enterprise control most directly delivers this?
- A. A written offboarding policy in the HR handbook
- B. SCIM provisioning tied to the identity provider, so deprovisioning is automatic on departure
- C. Encryption of data at rest
- D. A monthly manual audit of active users
Show answer
Answer: B.
SCIM automates provisioning and deprovisioning through the identity provider, removing access the moment the identity is disabled. A is a policy without enforcement. C protects stored data but not access lifecycle. D catches lapses late and manually, not automatically.
Q22D3 · Governance and RiskSelect one
Legal fears that customer data entered into ChatGPT will be used to train OpenAI's models. In a properly configured Business or Enterprise workspace, what is the accurate position a leader should state?
- A. All data entered is always used for training regardless of plan
- B. Business and Enterprise workspaces do not train on business data by default, so the concern is addressed by using the governed workspace rather than personal accounts
- C. Training only stops if every user disables it individually each session
- D. There is no way to prevent training on entered data
Show answer
Answer: B.
Business and Enterprise workspaces do not train on business data by default, which is precisely why sanctioned workspaces beat personal accounts. A is false for those plans. C misstates the default as opt-out per session. D is simply incorrect for the governed plans.
Q23D3 · Governance and RiskSelect two
A leader is drafting an AI risk taxonomy. Which two categories most belong in a working taxonomy for a large enterprise? (Select two)
- A. Data and privacy risk, including confidential data leaving governed systems
- B. Output-quality risk, including hallucination and unsafe or biased outputs reaching decisions
- C. The risk that the AI vendor's marketing is too enthusiastic
- D. The risk that competitors also use AI
- E. The risk that employees enjoy the tools too much
Show answer
Answer: A and B.
Data/privacy and output-quality are core risk categories any taxonomy must cover. C is not an enterprise risk category. D is a competitive-strategy consideration, not a governance risk. E is not a risk at all.
Q24D3 · Governance and RiskSelect one
A global firm must ensure EU customers' regulated data is processed within the EU. The leader is choosing an approach. Which capability makes this achievable in a governed workspace?
- A. Turning on prompt caching to speed up responses
- B. Data residency options that allow processing within a specified region such as the EU
- C. Asking staff to only work during EU business hours
- D. Using a longer context window
Show answer
Answer: B.
Regional data residency lets an organisation keep processing within a specified region such as the EU. A is a performance/cost feature. C confuses working hours with data location. D is unrelated to where data is processed.
Q25D3 · Governance and RiskSelect one
A newly formed AI governance committee insists on reviewing every individual prompt use case before approval, creating a growing backlog. Adoption is stalling. How should the leader reshape governance?
- A. Add more committee members to clear the backlog faster
- B. Move to tiered decision rights: pre-approve low-risk patterns, reserve committee review for high-risk cases, and publish clear guardrails
- C. Disband governance entirely to remove the bottleneck
- D. Approve everything automatically to clear the queue
Show answer
Answer: B.
Governance should enable by tiering decision rights and pre-approving low-risk patterns, reserving scrutiny for genuine risk. A scales a broken process. C removes needed controls. D abandons risk management altogether.
Q26D3 · Governance and RiskSelect one
A regulated organisation wants staff to use AI but must enforce that only approved user groups can access certain sensitive projects. Which control set most directly enforces this?
- A. Role-based access control combined with SSO, so identity and permissions govern who reaches what
- B. A shared team password written in a runbook
- C. A reminder email asking people to be careful
- D. Increasing the model's reasoning effort
Show answer
Answer: A.
RBAC plus SSO ties access to verified identity and role, enforcing who can reach sensitive projects. B is an insecure anti-pattern. C is guidance without enforcement. D affects model behaviour, not access.
Q27D3 · Governance and RiskSelect one
The audit team asks how the organisation will demonstrate, months later, exactly who accessed which AI workspace data and when. Which enterprise capability answers this?
- A. The analytics dashboard showing weekly active users
- B. Compliance and audit logs that record access and administrative events for later review
- C. The model's knowledge cutoff date
- D. A screenshot taken during the launch demo
Show answer
Answer: B.
Compliance and audit logs provide the durable, reviewable record auditors require. A shows usage counts, not access events. C is a model property. D is anecdotal and not a control.
Q28D3 · Governance and RiskSelect one
A vendor pitches an AI agent capability but the leader learns it runs only with US data residency and no zero-data-retention option. The workload involves EU-regulated data that must not leave the region. What should the leader conclude?
- A. Adopt it anyway; residency rules are usually flexible in practice
- B. Treat the residency and retention constraints as a hard fit-check the workload fails, and either change the workload or choose a compliant approach
- C. Ignore the constraint because the agent is technically impressive
- D. Ask the EU regulator to make an exception for this vendor
Show answer
Answer: B.
Data-residency and retention constraints are hard fit-checks; a workload that must stay in-region cannot use a US-only, no-ZDR service. A treats a compliance boundary as negotiable. C lets capability override compliance. D is unrealistic and misplaces the responsibility.
Q29D4 · Adoption and Change ManagementSelect one
Six months after a company-wide launch email, AI usage is high in the analytics team and near zero everywhere else. The sponsor is surprised. What does this pattern indicate?
- A. The tool is defective for most teams
- B. Adoption has stalled at the enthusiasts and has not crossed to the mainstream majority, which needs enablement beyond an announcement
- C. The analytics team is hoarding licences
- D. The launch email was sent to the wrong distribution list
Show answer
Answer: B.
Early enthusiasts adopt from an announcement, but the mainstream needs role-specific enablement to cross the chasm. A blames the tool for a change-management gap. C invents a motive with no evidence. D is a trivial explanation that does not fit high enthusiast usage.
Q30D4 · Adoption and Change ManagementSelect one
Usage has plateaued after a rollout. The leader can invest in only one lever. Which has the STRONGEST effect on getting the mainstream majority to adopt?
- A. A one-off all-hands webinar repeating the launch message
- B. Managers actively modelling use in their own work and expecting it in team routines
- C. A larger licence purchase
- D. A prize draw for the most prompts submitted
Show answer
Answer: B.
Manager modelling and embedding into routines is the strongest driver of mainstream adoption. A repeats a message that already failed to move the majority. C buys access, not behaviour. D rewards volume, which invites gaming rather than valuable use.
Q31D4 · Adoption and Change ManagementSelect one
A programme reports success as '6,000 licences issued' to the board. The leader is uneasy. Why is this metric misleading as a measure of success?
- A. Because licences are expensive
- B. Because issuing a licence measures access, not active or valuable use, and many issued licences may sit unused
- C. Because 6,000 is too small a number to matter
- D. Because licences expire monthly
Show answer
Answer: B.
Licences count access; adoption is measured by active, valuable use, and unused seats inflate the picture. A is about cost, not the metric's validity. C misreads scale. D invents a licensing rule irrelevant to the point.
Q32D4 · Adoption and Change ManagementSelect two
A leader is designing a champion network to drive adoption. Which two design choices most improve its effectiveness? (Select two)
- A. Choose champions embedded in each function who understand that function's real work
- B. Give champions time, recognition and a direct line to the programme team
- C. Appoint only senior executives as the champions
- D. Keep the champion list secret so it does not create expectations
- E. Rotate champions weekly so no one builds context
Show answer
Answer: A and B.
Embedded, credible champions plus real time, recognition and support make a network effective. C removes the peer credibility that makes champions work. D hides the very people colleagues should turn to. E destroys the context champions need to help.
Q33D4 · Adoption and Change ManagementSelect one
A staff survey finds many people say 'I don't know what I'd use it for in my role'. Usage is low. What is the BEST response?
- A. Send another motivational email about the technology's potential
- B. Build a role-specific use-case library with concrete, relevant examples people can copy the same day
- C. Mandate a minimum number of prompts per week
- D. Wait for staff to discover uses on their own
Show answer
Answer: B.
Concrete, role-specific examples remove the 'what would I use it for' barrier directly. A repeats generic messaging that already failed. C forces activity without relevance, inviting gaming. D leaves the barrier in place.
Q34D4 · Adoption and Change ManagementSelect one
An enthusiastic early cohort loves the tool, but the leader notices the same people are producing all the reported wins while most licence-holders never log in. What is the most useful next move?
- A. Celebrate the wins and consider the rollout complete
- B. Focus enablement on the un-engaged majority, since crossing the chasm — not delighting enthusiasts — is where value now comes from
- C. Remove licences from the enthusiasts to force others to step up
- D. Increase the model tier for the enthusiasts only
Show answer
Answer: B.
Value at scale comes from moving the majority, so enablement should target the un-engaged. A mistakes early-adopter success for organisational adoption. C punishes the people succeeding. D deepens the divide rather than closing it.
Q35D4 · Adoption and Change ManagementSelect one
A manager says AI adoption is 'not my job — that's IT's programme'. Team usage is the lowest in the company. How should the leader address this?
- A. Escalate to IT to push harder on the team
- B. Make adoption an explicit expectation of the manager's role, with support, because manager behaviour drives team adoption
- C. Bypass the manager and train the team directly and permanently
- D. Accept that some teams will simply never adopt
Show answer
Answer: B.
Adoption is a line-management responsibility; making it an explicit, supported expectation aligns the strongest lever. A treats a leadership issue as a tooling issue. C works around the manager rather than engaging the key influence. D gives up on a solvable problem.
Q36D4 · Adoption and Change ManagementSelect one
The board wants a single number to track whether the AI rollout is 'working'. The leader must recommend something better than licences issued. Which is the strongest single adoption signal?
- A. Number of licences purchased
- B. Active use tied to real work, such as regularly recurring usage in core workflows across most of the target population
- C. Total prompts submitted, regardless of purpose
- D. Number of people who attended the launch event
Show answer
Answer: B.
Recurring, work-embedded active use across the target population reflects real adoption. A counts access. C rewards volume that may be noise. D measures a one-off event, not sustained use.
Q37D5 · Workforce and EnablementSelect one
A company plans identical two-hour AI training for all 2,500 staff regardless of role. The leader questions the design. What is the BEST improvement?
- A. Make the same session mandatory and add a test
- B. Segment training by role and capability level so each group learns what its actual work requires
- C. Cut the training to one hour to save time
- D. Deliver it only to managers and let it cascade informally
Show answer
Answer: B.
Role- and level-segmented enablement teaches each group what its work actually needs, which drives real capability. A enforces a one-size-fits-none design. C trims duration without fixing relevance. D relies on unmanaged cascade that usually decays.
Q38D5 · Workforce and EnablementSelect one
A manager wants to announce the team is 'OpenAI certified' after everyone earned the Academy AI Foundations badge. What is the accurate position the leader must insist on?
- A. The announcement is fine; badges are certifications
- B. Academy badges and pathway certificates of completion are not certifications, so claiming the team is 'certified' would be inaccurate
- C. The badge counts as certification only for internal use
- D. Certification is automatic once three badges are earned
Show answer
Answer: B.
OpenAI states plainly that Academy badges and pathway certificates of completion are not certifications; the claim would be inaccurate. A, C and D all misrepresent a badge as a certification in different ways, which the leader must correct.
Q39D5 · Workforce and EnablementSelect two
A leader is choosing training formats to produce durable behaviour change, not just awareness. Which two formats are strongest? (Select two)
- A. Hands-on practice on the person's own real tasks with feedback
- B. Ongoing peer support and role-specific examples reinforced over time
- C. A single recorded lecture watched once
- D. A slide deck emailed to all staff
- E. A one-line mention in a newsletter
Show answer
Answer: A and B.
Applied practice on real tasks and sustained peer reinforcement change behaviour. C, D and E are all one-off, passive awareness activities that rarely alter how people work.
Q40D5 · Workforce and EnablementSelect one
As AI drafts routine documents faster, a leader must redesign an analyst role. Which capability should the redesigned role most emphasise?
- A. Typing the documents faster than before
- B. Reviewing, judging and improving AI output, and handling the exceptions and higher-value work AI cannot
- C. Producing more first drafts manually to stay busy
- D. Avoiding AI to preserve the old workflow
Show answer
Answer: B.
Role redesign shifts humans toward review, judgement and higher-value exception work. A optimises a task AI now does. C manufactures redundant work. D refuses the change entirely.
Q41D5 · Workforce and EnablementSelect one
Morale has dropped and staff are asking 'are they going to replace us?'. Leadership must respond. What is the MOST effective response?
- A. Reassure everyone that nothing will ever change
- B. Be honest about how roles will change, name the reskilling and support on offer, and explain where human judgement remains essential
- C. Avoid the topic until decisions are final
- D. State that anyone worried can look for another job
Show answer
Answer: B.
Honest framing of role change plus concrete reskilling support builds trust and engagement. A makes a promise that will be broken. C leaves a vacuum that rumour fills. D is corrosive and drives out talent.
Q42D5 · Workforce and EnablementSelect one
A leader wants to route employees onto appropriate OpenAI Academy learning. A team of API developers and a team of business analysts need different paths. What is the best routing principle?
- A. Send everyone to the longest available course to be thorough
- B. Map each group to the courses that match their role — for example builder-oriented API courses for developers and workplace application courses for analysts
- C. Let people pick any course at random
- D. Require the leadership course for all technical staff
Show answer
Answer: B.
Matching learning paths to role and product context is the point of a capability-based enablement plan. A wastes time on irrelevant depth. C produces uneven, unaligned capability. D misroutes technical staff into a strategy course.
Q43D5 · Workforce and EnablementSelect one
Finance asks the leader to justify a training budget when 'the tools are intuitive'. What is the strongest justification?
- A. Training is a compliance requirement so it must be funded
- B. Without role-relevant enablement, purchased licences go unused and the expected value is never realised, so training protects the larger tool investment
- C. Competitors spend on training so we must too
- D. Training keeps staff busy between projects
Show answer
Answer: B.
Enablement is what converts licence spend into realised value; skipping it wastes the larger investment. A invents a compliance mandate. C argues by imitation. D trivialises training as filler.
Q44D6 · Measurement and ROISelect one
A team deployed an AI assist and now wants to prove it saved time, but no measurement was taken before launch. What is the core problem?
- A. The model version was not recorded
- B. Without a pre-deployment baseline, any claimed saving cannot be substantiated against a before state
- C. The team used the wrong dashboard
- D. Not enough people were surveyed about satisfaction
Show answer
Answer: B.
A benefit claim needs a baseline; without a before-measurement there is nothing to compare against. A is a minor detail. C is tooling, not the measurement gap. D measures sentiment, not the time saving in question.
Q45D6 · Measurement and ROISelect two
A leader wants leading indicators to watch early, before financial results arrive. Which two are LEADING indicators for an AI initiative? (Select two)
- A. Active, recurring use in the target workflows
- B. Reduction in the time taken to complete the targeted task
- C. Year-end reported profit
- D. The annual audited financial statement
- E. The share price at the next earnings call
Show answer
Answer: A and B.
Recurring use and task-time reduction show up early and predict later results, so they are leading indicators. C, D and E are lagging outcomes reported long after and shaped by many factors beyond the initiative.
Q46D6 · Measurement and ROISelect one
A support team's average handle time dropped 30%, but the rate of reopened tickets rose sharply over the same period. How should the leader judge the result?
- A. As a clear success, since handle time is the headline metric
- B. As incomplete: speed must be paired with a quality measure, and rising reopens suggest faster-but-worse resolutions that may erase the gain
- C. As irrelevant, because reopen rate is not a real metric
- D. As a reason to cut handle time targets further
Show answer
Answer: B.
Cycle-time gains must be paired with quality; rising reopens can cancel or reverse the apparent benefit. A reports a partial metric as a win. C dismisses a valid quality signal. D doubles down on the metric that may be causing the harm.
Q47D6 · Measurement and ROISelect two
A pilot 'was basically free' on a trial, but at production scale the bill grew. Which two cost drivers must the leader model before scaling? (Select two)
- A. Per-use consumption cost that grows with volume, such as tokens or usage-based fees
- B. Recurring per-seat licence cost across the full user population
- C. The colour of the dashboard theme
- D. The number of slides in the launch deck
- E. Which conference room hosted the kickoff
Show answer
Answer: A and B.
Usage-based consumption and per-seat licensing are the drivers that scale the bill with volume and headcount. C, D and E are irrelevant to unit economics.
Q48D6 · Measurement and ROISelect one
A region grew revenue 12% in the same period it adopted AI, but it also hired two senior reps and the whole market rose. The leader must report to the board. What is the MOST honest attribution approach?
- A. Attribute the full 12% to AI to strengthen the case
- B. Separate the AI contribution from hiring and market effects, state assumptions, and report a defensible range rather than the whole delta
- C. Attribute nothing to AI to be safe
- D. Wait several years before saying anything about the effect
Show answer
Answer: B.
Honest attribution isolates confounders and reports a defensible range with assumptions. A overclaims by ignoring hiring and market lift. C under-reports and hides real benefit. D forfeits the ability to steer the programme with evidence.
Q49D6 · Measurement and ROISelect one
The board asks for a single credible line on an AI initiative's return. Which board line best reflects disciplined measurement?
- A. 'AI transformed the business and delivered massive savings.'
- B. 'Against a measured baseline, we estimate £X–£Y benefit this year, attributing conservatively, at a net run cost of £Z.'
- C. 'Everyone loves the tool, so it must be paying off.'
- D. 'We issued thousands of licences, so the return is clear.'
Show answer
Answer: B.
A credible board line cites a baseline, a range, conservative attribution and net run cost. A is a hero claim with no evidence. C substitutes sentiment for value. D counts access, not return.
Q50D6 · Measurement and ROISelect one
A leader reports gross benefit from an AI programme but omits the licence, usage and enablement costs. The CFO objects. What is the correct fix?
- A. Keep reporting gross benefit; costs sit in another budget
- B. Report net benefit — gross value less the full run cost including licences, usage and enablement — so the return is honest
- C. Remove the costs from the company accounts
- D. Report only the costs to be conservative
Show answer
Answer: B.
ROI must net the full run cost against gross benefit to be honest. A overstates return by hiding cost. C is not a real option and would misstate accounts. D swings to the opposite distortion and hides the benefit.