Task 4.3: Lead enterprise-wide change and build AI-ready workforce capabilities.

• AAdd more engineers so that the program can deliver faster

  Incorrect

  Delivery capacity is not the constraint. The program stalls at cross-department decisions that engineers cannot make.

• BKeep sponsorship in IT, because AI is mainly a technology initiative

  Incorrect

  AI transformation needs both technology and business leadership; an IT-only sponsor cannot secure cross-department decisions.

• CGain visible executive sponsors who tie AI to strategic outcomes

  Correct

  AWS guidance calls for active and visible executive sponsorship that links AI initiatives to strategic outcomes and ROI and makes critical decisions on strategy, scope and resources.

• DWait for results to speak for themselves before involving executives

  Incorrect

  Without sponsorship the program cannot get the budget and decisions it needs to produce those results.

Executive sponsorship turns AI from a technical project into an enterprise priority. Sponsors from business and technology who connect AI to strategic outcomes can make the decisions, secure the resources and hold teams accountable for results.

• ASay nothing until rollout is complete, so that the change causes less anxiety

  Incorrect

  Silence lets rumors define the story and increases resistance; it does not reduce anxiety.

• BExplain openly what agents can and cannot do, when, and how roles change

  Correct

  AWS guidance is to communicate transparently about agent capabilities and limitations to set realistic expectations, and to position agents as teammates rather than replacements to reduce resistance.

• CPromise that no role will change in any way

  Incorrect

  Leadership cannot know every effect in advance, and a promise that later proves false destroys the trust the program depends on.

• DShare only the cost savings target with adjusters

  Incorrect

  Leading with savings alone confirms the fear of replacement and leaves out the information people need about their own roles.

People resist what they do not understand and what threatens them. Transparent, consistent communication about capabilities, limitations, timelines and the effect on roles, framing AI as a teammate, builds the trust that adoption depends on.

• ALeaders openly share their own failed experiments and rule out penalties for well-run ones that miss

  Correct

  AWS guidance is that leadership should support experimentation, that experiments are an opportunity to learn and that people are not punished for well-run experiments with undesired results, because knowing what not to do has value.

• BIncrease the experiment budget and replace the proposal form with a one-page template

  Incorrect

  Less bureaucracy helps in general, but the budget already goes unspent and staff name career risk, not paperwork, as the reason they hold back.

• CApprove only experiments whose business case shows a high probability of meeting targets

  Incorrect

  Screening for near-certain success confirms that misses are unacceptable and removes the experiments that would teach the most, including what not to do.

• DRun a firm-wide training program that shows staff what the approved AI tools can do

  Incorrect

  Training builds fluency, but the surveys show staff already understand the tools; it leaves the fear of career damage untouched.

Fear of failure is a cultural barrier that money, paperwork and training do not remove. When leaders openly share their own misses and make clear that a well-run experiment with a poor result is not a career risk, experiments become a way to learn rather than a gamble, without waiting for a new performance system.

• ATraining reserved for the data science team, who will then do all AI work

  Incorrect

  Concentrating knowledge in one team leaves the rest of the workforce unable to use or question AI, the opposite of enterprise-wide literacy.

• BA single one-hour video required for every employee once

  Incorrect

  A one-off generic video does not build role-relevant skills or sustained learning.

• CPostponing all training until the company's AI platform is fully built

  Incorrect

  Waiting for the platform delays literacy that is needed to identify use cases and adopt the platform once it arrives.

• DRole-based training tailored to each group's work with AI

  Correct

  AWS guidance recommends role-based training tailored to engineers, product managers, domain leads and compliance officers.

• EInternal hackathons that let mixed teams try AI on real problems

  Correct

  Organizing internal hackathons encourages experimentation and fosters innovation, building practical familiarity across the organization.

AI literacy grows through learning that fits each role and through hands-on practice. Role-based training and hackathons together build capability across the organization rather than in one specialist team.

• AA hands-on technical course for executives on training models

  Incorrect

  AWS's recommended education for senior executives covers AI's strategic potential, capabilities, business impact and the importance of data, not hands-on model building.

• BExecutive education on AI's potential and its data needs

  Correct

  AWS recommends educational initiatives to help senior executives understand AI's strategic potential, capabilities, business impact and the importance of data in generative AI projects.

• CDelegation of all AI decisions to the chief data officer

  Incorrect

  AWS guidance is to obtain executive sponsorship and form cross-functional governance committees that include business leadership; delegating to one member leaves the committee unaligned.

• DA proof of concept first, so its results educate the committee

  Incorrect

  AWS places executive education at the Envision stage, before proofs of concept; the committee would otherwise set priorities with the same mismatched views.

Leadership alignment depends on shared understanding. Targeted education for executives on what AI can do and what it needs, especially data, prepares them to sponsor and steer the program together.

• AA structured change program with impact assessments and a roadmap

  Correct

  The framework recommends a programmatic change acceleration approach: assess readiness through impact assessments and develop a change strategy and roadmap with leadership actions, communications, training and risk mitigation.

• BTechnology rollout first, then dealing with people issues as they arise

  Incorrect

  Treating people issues reactively raises resistance and slows adoption.

• CA single all-staff email announcing the new tools

  Incorrect

  One announcement does not manage the widespread changes to roles and ways of working.

• DLetting each manager decide whether to mention the tools at all

  Incorrect

  Inconsistent messages undermine adoption across the bank.

Organizations that apply a structured, integrated change process achieve higher adoption. Planning the people side alongside the technology makes AI rollouts stick.

• AA longer classroom course that covers more AI theory

  Incorrect

  More theory in the same format does not fix the gap between learning and use on the job.

• BNo training, so staff learn entirely by trial and error

  Incorrect

  Unsupported trial and error is slow and inconsistent.

• CMixed classroom, experiential and just-in-time training

  Correct

  The framework suggests combining virtual, classroom, experiential and just-in-time training and using immersion days.

• DTraining only for managers, who will pass it on to staff

  Incorrect

  Cascading through managers leaves most staff without direct learning.

People learn AI best through a mix of formats, including practice close to the moment of need. A blended program builds lasting literacy across the workforce.

• AForbid staff from reading the assistant's outputs

  Incorrect

  Staff need to read outputs to do their work; awareness, not avoidance, helps them notice problems.

• BRely only on the model's built-in safety training

  Incorrect

  AWS recommends layered defenses, including people who know what to look for.

• CTrain only the security team, since others do not need to know

  Incorrect

  The people using the assistant daily are best placed to notice manipulation.

• DTrain users and admins to spot and report prompt injection

  Correct

  Among mitigations for prompt injection, AWS recommends user education: training administrators and users to identify the signs of prompt injection and respond promptly.

AI-ready workforces understand the risks of the tools they use. Training users to recognize manipulation adds a human layer to technical safeguards.

• AKeep results confidential until the whole program is finished

  Incorrect

  Hiding early results wastes the momentum they could create.

• BStart ten new use cases at once to show ambition

  Incorrect

  Overextending risks failures without spreading the existing success.

• CReplace the program team to bring fresh energy

  Incorrect

  Replacing a team that is delivering results disrupts momentum.

• DShare measured results and celebrate the early wins widely

  Correct

  The Cloud Adoption Framework recommends celebrating early wins and building a change coalition to generate momentum for change.

Visible success builds belief. Sharing measured early wins helps staff see that AI delivers, which encourages adoption of what comes next.