When organisations discuss AI readiness, the conversation often moves quickly to skills. Employees need prompt training. Managers need AI literacy. Technical teams need model expertise. All of that is true, but it can still frame the problem too narrowly.
The larger gap is frequently between the way work is organised today and the way it needs to operate when AI can generate, classify, recommend and act. A workforce can complete dozens of training modules and still struggle to create value if roles, decisions, controls and incentives remain unchanged.
Exposure is not the same as replacement
The International Labour Organization’s refined 2025 index estimated that one in four jobs worldwide sits in an occupation with some exposure to generative AI. Its central conclusion was not that one in four jobs would disappear. Transformation was identified as the more likely effect because most occupations contain a mixture of automatable and distinctly human tasks.
That distinction matters. If AI changes a bundle of tasks rather than removing an entire job, leaders must redesign the bundle. Which activities move to the system? Which remain human? Which new checks, decisions and interactions appear? A job description written around the old workflow will not answer those questions.
Tool training is not capability building
Generic training can help employees understand what an AI system does and how to interact with it. But operational competence is role-specific.
A recruiter needs to know when an AI-generated candidate summary is incomplete or biased. A finance analyst needs to preserve the evidence behind a variance explanation. A manager needs to know when an automated recommendation requires challenge. A customer service employee needs a clear rule for when the system may draft, recommend or send.
The capability is therefore not simply “using AI.” It is performing a defined responsibility in an AI-enabled workflow.
Europe is turning literacy into an operating requirement
Article 4 of the EU AI Act requires providers and deployers to take measures that support the development of AI literacy among staff and other people using AI systems on their behalf. The European Commission’s guidance emphasizes context: the organisation’s role, the risks of the system, the knowledge of the people involved and the purpose for which the technology is used.
That is a useful management principle beyond compliance. A single introductory course cannot prepare every employee for every system. Literacy should follow the risk and the responsibility.
For leaders, the practical implication is clear: training should be linked to named use cases, real decisions and observable behaviours. Employees should know not only what the system can do, but how to verify it, when to escalate and what they remain accountable for.
The unit of AI transformation is not the tool. It is the workflow—and the human responsibilities designed around it.
Start with work, not roles
Many transformation programmes begin by asking which jobs will be affected. A more useful starting point is the flow of work.
Map the trigger, the information required, the sequence of decisions, the output and the exceptions. Then identify where AI could improve speed, quality, access or consistency. Only after that should the organisation redesign roles and learning.
This approach avoids two common errors. The first is automating a task that was unnecessary. The second is removing a junior activity without replacing the learning it provided.
Protect the formation of expertise
Entry-level work often contains repetitive activities that are attractive automation targets. Those activities can also help employees understand customers, systems and professional standards.
If AI performs the first draft, the first analysis or the first screening, organisations need a deliberate alternative for building judgment. That might include supervised review, simulations, rotations, annotated examples and earlier participation in complex cases.
The goal is not to preserve low-value work for its own sake. It is to preserve the learning that used to be embedded within it.
A practical five-part redesign
1. Define the outcome
Specify the customer, employee or business result the workflow must produce.
2. Separate routine from judgment
Identify which tasks are repeatable, which decisions require context and which consequences require human accountability.
3. Assign the human control points
State who verifies outputs, who approves action and who responds when performance deteriorates.
4. Build role-based learning
Train people on the actual system, data, risks, policies and exceptions they will encounter.
5. Measure adoption and quality together
Usage alone is not success. Track rework, error detection, decision quality, cycle time, confidence and business impact.
The leadership task
The AI skills gap is real, but it cannot be closed by training alone. Leaders must decide how work should change, which responsibilities remain human and how employees will develop judgment in the new system.
Organisations that treat AI as a software rollout may create pockets of productivity. Those that treat it as work redesign can build a repeatable capability.