The rapid advancement of artificial intelligence is not just a technological shift; it’s a fundamental reshaping of the global workforce, creating a substantial upskilling challenge.
A recent McKinsey survey highlights a critical gap: While 75 percent of US workers anticipate their roles changing due to AI in the coming five years, only 45 percent have gone through a recent upskilling program.
This skills gap is unlikely to be closed by relying on traditional training programs. Such programs, typically structured as short-term, classroom-based events, have demonstrated limitations in creating lasting impact and a tangible return on investment. They often fail to provide the continuous reinforcement needed to embed new knowledge, and they do not foster the hands-on, real-world application that is crucial for genuine skill development.
More importantly, a conventional training approach often fails to address the deeper need for a cultural and behavioral shift in how individuals and teams think, decide, and collaborate in an AI-augmented workplace. To truly unlock the potential of AI, organizations should move beyond sporadic training that, particularly for technical skills, becomes obsolete before it even reaches the workforce. Consider an AI training program designed in January 2026 around one specific model: It would already be partially obsolete before completion, given that five major frontier AI models were released in February alone, alongside almost weekly feature updates. Instead, organizations should cultivate a culture of continuous learning and adaptation.
How can organizations build a rapid and adaptable learning culture in such a fast-moving environment? The answer lies in tapping into the more durable and innate ways that people actually learn—through role modeling and by collaborating with peers in the flow of work, creating a scalable training engine that continuously updates itself.
Role modeling
Neuroscience research on “mirror neurons” reveals that our brains are designed to learn by observing and imitating. In essence, we are inherently wired to mimic the behaviors we see in others, particularly in leaders.
In fact, McKinsey's own Influence Model for organizational change identifies role modeling as a critical component for embedding new behaviors and skills. When leaders and influential peers act as exemplars, their actions become a living curriculum, demonstrating the desired mindsets and capabilities in a way that is far more impactful than any training program.
We have adopted a “lighthouse” approach to accelerate AI adoption, intentionally selecting a small number of specialized client service teams to act as role models for the broader organization. These lighthouse teams are chosen based on a combination of factors, including notable record of AI impact, strong AI-enabled leadership, engaged AI champions (for example, team members eager to build capabilities), and the presence of high-quality compelling use cases. This ensures that early successes are both meaningful and highly visible, creating momentum beyond the initial teams.
In one lighthouse engagement, a team working on a procurement project in Europe reimagined its approach to long-tail supplier management. Equipped with new AI capabilities, the team identified approximately €80 million in value at stake for the client, while reducing manual sourcing and supplier analysis effort by roughly 90 percent. Crucially, the impact extended beyond the initial team: The team inspired others and shared their workflows, tools, and results with peer teams in their practice, helping spark similar use cases and driving broader adoption across their office.
Across lighthouse engagements, this model has consistently delivered both client impact and sustained capability building. Participants have demonstrated a roughly 35-percentage-point increase in AI proficiency and usage (to about 50 percent, up from about 15 percent), with gains persisting months after the sprint. Just as important, lighthouse client service teams act as multipliers—radiating their capabilities, sharing success approaches through internal channels and dashboards, and encouraging other teams to experiment with AI in their own contexts.
Peer-based learning in action
Research from the Harvard Business Review reveals that 55 percent of employees naturally turn to their peers first when they need to learn something new, and leading companies are actively fostering this tendency. By creating platforms for peer-to-peer coaching, knowledge sharing, and collaborative problem-solving, they build a dynamic learning ecosystem. These approaches not only facilitate the practical transfer of skills but also create a culture of shared ownership for development, increase employee engagement, and foster a supportive environment where continuous learning can thrive.
At McKinsey, we are putting this powerful principle into practice with our AI Fellows Program. This initiative identifies consultants who commit to both a personal upskilling journey and the responsibility to upskill their peers. The program begins with an intensive learning week to build a strong community, followed by a six- to eight-week deployment where fellows act as catalysts for AI adoption within their teams. They become embedded role models and peer coaches who drive change from within.
The impact on our client work has been immediate and substantial. One engagement leader noted that the program “allowed us to develop grassroots enthusiasm to implement a use case that generates $4 million to $6 million annually for the client—just from one use case.” Another leader was “wildly impressed,” stating that the fellow enabled work that “we otherwise wouldn’t have been able to do for our client in the time frame we had,” which in turn “changed the mindsets of client service team leaders; everything is now AI agentic.”
Internally, the effect is just as profound. The program creates a ripple effect of capability building, as one team reported: “Every week, the team learned something new from the fellow that they wouldn’t be doing otherwise.” To scale this impact, fellows have driven change across North America by apprenticing colleagues in local offices. Every office hosted in-person, hands-on “AI Arcade” sessions grounded in real-world use cases, increasing AI awareness, adoption, and confidence across the entire region. This model proves that by embedding learning into the flow of work and empowering our people to lead the change, we can build the AI-ready workforce of the future.
When we are responsible for upskilling a global workforce, we tend to default to broad-brush, enterprise-level programming. Yet, the continuous evolution of AI renders that traditional playbook insufficient. We need a new approach to capability building: creating mechanisms for organic, peer-to-peer knowledge transfer.
For learning leaders, the mandate shifts from delivering curriculum to catalyzing networks. By identifying and investing in our internal teams and providing structural support for them to coach their colleagues, we transform the organization into a self-updating learning engine. The organizations that lead in developing AI talent will be those that realize their most powerful learning asset is the collective, networked expertise of their own people.