Domain 4 review: leadership and change

The second half of Domain 4 is about people and scale: Task 4.3, leading enterprise-wide change and building an AI-ready workforce, and Task 4.4, scaling AI from pilots to enterprise-wide deployments. Andre invited Kwame Asante, the developer who had built the BookVet booking app, to this session, because Kwame had watched several technology changes at Cascade Ridge succeed or stall. "The software was never the hard part," Kwame said. "Getting people to change how they work was."

Task 4.3: sponsorship, champions and teams

Skill 4.3.1 asks you to secure executive sponsorship and leadership alignment, and to identify and empower AI champions. A sponsor is a senior leader who owns the initiative's success, removes obstacles and commits resources. Champions are respected people within teams, often not managers, who try new tools early, help colleagues and carry feedback back to the project. At Cascade Ridge, Dr. Kaur has agreed to sponsor the planned call-note pilot at the emergency hospital, and two experienced technicians have volunteered to act as champions on the night shift.

Skill 4.3.2 is about cross-functional teams with clear accountability. The guide names business, technical, legal and compliance members. An AI initiative run only by technologists tends to build something nobody uses; one run only by the business tends to underestimate data and risk work. Each team needs a named accountable lead and an agreed way of making decisions.

Communication and cultural barriers

Skill 4.3.3 is about transparent communication that addresses workforce concerns about timelines, outcomes and role changes. People's first question about AI is usually "what happens to my job?", and vague reassurance makes the fear worse. Good communication is specific and honest: what the tool will and will not do, when changes will happen, how roles will change, what training is available, and how people can raise concerns. If roles will change significantly, say so early.

Skill 4.3.4 asks you to recognize cultural barriers and apply leadership interventions. The guide names risk aversion, resistance to change and fear of failure.

Workforce capability and new roles

Skill 4.3.5 asks you to choose workforce development approaches that build AI literacy. The guide gives examples: proofs of concept, hackathons, training programs and responsible AI training. Different groups need different depths. Executives need enough literacy to make sound investment and risk decisions; front-line staff need practical skill with the tools they use and knowledge of when not to trust them; technical teams need deeper skills; everyone needs responsible AI basics.

Skill 4.3.6 asks you to identify roles that should shift from manual work to human oversight and collaboration with AI. The principle is to let AI handle what it does well (volume, speed, pattern finding, first drafts) while people handle what they do better (judgment, empathy, accountability, exceptions). At Cascade Ridge, the no-show pilot plan has reception staff moving from phoning every client to working a prioritized list from the model and handling the conversations that need a person. Their job will change shape; it will not disappear.

Task 4.4: scaling from pilot to enterprise

Skill 4.4.1 names an iterative set of phases: envision, experiment, launch and scale. Envision defines the goal, the value and the success measures. Experiment tests feasibility and value on a small scale with real users. Launch puts a governed version into production for a defined scope. Scale extends it across sites, functions or customers while monitoring results. Each phase ends with a decision based on evidence, and the process repeats as new use cases join.

Skill 4.4.2 recommends scaling methods that start with short-term wins and build toward enterprise-wide deployment. A quick win that shows measurable value builds credibility, funding and skills for harder initiatives. The trap is choosing quick wins that lead nowhere; the best ones also build reusable foundations, such as a cleaned dataset, a governance process or a shared integration.

Skill 4.4.3 is about establishing an AI center of excellence (COE) and cross-functional collaboration mechanisms. A COE is a small central group that sets standards, shares reusable components and lessons, provides expertise to business teams and supports governance. Organizations choose between a centralized model, where the COE does most AI work; a federated or hub-and-spoke model, where business units build with COE standards and support; and a decentralized model with light coordination. As maturity grows, many move from centralized toward hub-and-spoke.

Skills 4.4.4 to 4.4.6 complete the task. Continuous feedback and success metrics track value over time, not just at launch. Moving from experimental to production-grade adds requirements that pilots often skip: security review, monitoring, support arrangements, documentation, model operations, cost controls and a named owner. And evaluating multiple factors during scaling, such as performance at higher volumes, differences between sites, support capacity, cost at scale and business continuity, protects the organization from a rollout that works in one clinic and fails in eleven.

Practice scenarios for leadership and scaling

Kwame brought examples from previous technology changes at Cascade Ridge, and Noor turned them into exam-style items. Each tests one skill from Task 4.3 or 4.4.

Communication that works

Task 4.3 questions about communication often offer a choice between messages. The strongest options share four features. They are early: people hear about changes before rumors fill the gap. They are specific: they say what the tool does and does not do, and how roles will change. They are two-way: they invite questions and show what changed in response. And they are consistent: leaders say the same thing in town halls, team meetings and written updates, and their actions match their words. Options that delay communication until everything is final, or offer general reassurance without facts, are usually distractors.

Andre wrote a short communication plan for the no-show pilot as practice. Reception staff would hear about the pilot from their clinic manager two weeks before it started, with a one-page explanation of what the prioritized list would look like and an honest statement that their work would shift from phoning everyone to handling the most important conversations. A champion at each pilot clinic would collect questions weekly, and Andre would answer them in a short written update. After four weeks, staff would see the early results and be asked what to change before any wider rollout.

Choosing an operating model

Questions about centers of excellence often turn on fit. A centralized model suits an organization at an early stage with scarce AI skills, because it concentrates expertise and keeps standards consistent, but it can become a bottleneck. A hub-and-spoke model suits an organization with several business units that want to move at their own pace, because it combines central standards and shared platforms with local delivery. A decentralized model gives business units the most freedom but risks duplication and inconsistent governance. When a scenario describes a central team overwhelmed by requests from many units, the usual answer is to move toward hub-and-spoke, keeping standards central while building capability in the units.

Andre noticed that his own role had shifted during the autumn in exactly the way the guide describes. He had started as a manager who scheduled staff and ordered supplies; he was becoming the person who decided where AI would help, who would oversee it and how its value would be shown. That change, from doing the work to directing and overseeing it, is what Task 4.3 means by shifting roles toward human oversight and collaboration with AI.

A practical test for any scaling plan is to ask what would happen in the first week at a new site if the system stopped working. If the answer is that nobody would know what to do, the plan is not yet production-grade.