Preparing people for AI means developing the initiative, structured thinking, and execution needed to turn possibilities into dependable outcomes—and designing work that allows those abilities to grow.
At Procter & Gamble, professionals working individually with AI matched the performance of two-person teams working without it on product-innovation challenges. AI-assisted proposals also combined technical and commercial perspectives more effectively across professional backgrounds. The findings, published in June 2026, show how access to assistance can change what people accomplish within a defined assignment. The Cybernetic Teammate, Organization Science
For executives, the implication extends beyond productivity. When tools expand the knowledge and execution support available to a person, workforce preparation must reconsider what that person needs to learn, how they approach unfamiliar work, and how their contribution is demonstrated.
Skill has always involved more than recall. AI brings renewed attention to the ability to recognize an opportunity, work out what it requires, and carry it through to a useful result.
That calls for a broader view of skilled performance, with initiative and a disciplined way of thinking alongside domain knowledge and practical execution.
For AI-enabled work, a useful working definition is this: skill is the demonstrated ability to turn intent into useful, dependable outcomes within a defined area of responsibility—by framing problems, connecting their implications, applying knowledge and tools, and carrying work through to completion.
Intent gives effort direction. A problem creates a reason to act; a possibility creates an opportunity to add value. Initiative turns that recognition into movement: seeking information, testing an approach, involving the right people, and persisting through uncertainty.
Structured thinking gives that movement coherence. A working mental map connects the desired result with the people, dependencies, decisions, risks, and operating conditions that shape it. Its value lies in revealing relationships: how a faster response changes review needs, how a new process affects another team, or how a promising idea will be supported after launch.
AI can help construct and challenge that map. People still need sufficient understanding to apply it, recognize missing context, and exercise the responsibilities they accept. Execution then tests whether their understanding holds in practice.
This creates room to recognize capabilities beyond familiar job descriptions. A person who understands customer concerns may contribute to process improvement when given analytical support and an opportunity to practice. Existing strengths provide a starting point; targeted learning can extend them. The appropriate scope and standard will differ by role.
Evidence from customer support illustrates that possibility. A study of 5,172 agents found a 15% average increase in issues resolved per hour with AI assistance, with larger gains among less experienced and lower-skilled workers and evidence of learning. Those results concern a particular deployment, but they give leaders reason to examine how support changes the path to proficiency. Generative AI at Work, February 2025
1. Give people meaningful outcomes and room to act.
Make the purpose of work clear, including the value to create and the constraints to respect. Give employees appropriate authority to identify problems, suggest improvements, and test ideas. Initiative needs time, access, and a legitimate route to action. A workforce expected to take ownership must have opportunities to exercise it.
2. Teach people to develop a working mental map.
Use real assignments to connect business value, stakeholders, workflow, knowledge, risk, and impact. Before building a solution, ask people to explain how their proposed approach would operate and what must be true for it to succeed.
For example, in an illustrative customer-service exercise, drafting a faster response is only part of the assignment. The employee also considers the applicable policy, the customer’s circumstances, the next team affected, and the point at which the issue needs a different owner. The learning comes from connecting those implications and testing them against feedback.
3. Put suitable tools, trusted knowledge, and clear rules inside the work.
Provide approved assistance and the business context needed to use it. Make permitted data use, review requirements, decision authority, and escalation routes easy to understand. NIST’s AI Risk Management Framework connects workforce training with defined responsibilities and policies for human–AI collaboration. NIST AI RMF 1.0, January 2023
Design the process so that obtaining context, checking a result, or seeking help is a practical part of execution. People should be able to act confidently within understood boundaries.
4. Develop understanding through execution and reflection.
Give people assignments that require them to explain choices, test assumptions, investigate failures, and improve the next attempt. Experienced colleagues can make their reasoning visible by walking through the consequences they considered and why they changed direction.
A small randomized study of 52 developers learning an unfamiliar programming library found lower immediate mastery with AI assistance on average, without a statistically significant time saving. Some patterns involving active comprehension were associated with better learning; those associations were not causal findings. The study does not establish long-term effects. Anthropic research, January 2026
The leadership implication is to design assignments that produce both a useful result and opportunities to develop the understanding required for future work.
5. Assess how people carry responsibility through to results.
Evaluate the outcome alongside the thinking and actions relevant to the role. Look for sound problem framing, appropriate use of assistance, collaboration, quality of execution, and the ability to respond when conditions change.
An assessment can include a missing piece of information, a changed constraint, or a flawed AI recommendation. Observe how the person recognizes the issue, adjusts the approach, and completes or appropriately escalates the work. Record the support available so that expectations remain grounded in the conditions of performance.
The executive task is to define what capable performance looks like for each role and establish the conditions in which people can develop and demonstrate it. The AI-Era Workforce Capability Framework provides the broader foundations; this requires applying them to assignments, support, and evidence of performance. AI-Era Workforce Capability
Business leaders must set outcome standards and provide opportunities. Learning teams must develop proficiency through practice. Technology and governance teams must make assistance usable and responsibilities clear. Managers must give feedback and reward initiative, collaboration, sound judgment, and follow-through.
As AI capabilities change, revisit which knowledge people need to retain, which work they can undertake with assistance, and which responsibilities require deeper expertise. Help them refine their mental maps through experience and evidence.
Workforce preparation should enable people to see possibilities, think through their consequences, and bring them into operation. That is how individual skill becomes enterprise capability.
For a role being equipped with AI, how will you develop and demonstrate its people’s ability to recognize an opportunity, think it through, and carry it into dependable operation?
Short, practical perspectives on AI-era operations, governance, and operating-model transformation.
Subscribe