Why AI Alone Won't Fix Learning
Scaling Expertise: How Skills Are Really Built in an AI World
Peter Zemsky
Founding CEO, Lexarius
April 11, 2026
12 min read
Share:
Most organisations already know what good looks like in leadership, sales, and communication. The challenge is not defining the right behaviours. It is executing them consistently under pressure.
​
This is where most learning approaches fall short. Content builds awareness but does not reliably translate into behaviour. Workshops can inspire, but the effect fades.
The real gap is not knowledge. It is execution in the moments that matter.
Skills are built through practice: repeated exposure to realistic situations, combined with high-quality feedback. Historically, this has been difficult to scale, relying on expert facilitation and coaching, often shaped by experienced thought leaders and L&D professionals. As a result, both the quality and consistency of practice have been hard to sustain.
​
What has changed is the ability to deliver this at scale. Advances in AI now make it possible to simulate real workplace situations, provide immediate feedback, and enable repeated practice. What was once constrained by time, cost, and access to expert coaching can now be delivered at scale.
But scale alone does not guarantee effective learning.
This is the approach behind Lexarius. By combining AI-powered role play with frameworks developed by experienced thought leaders and L&D professionals, it creates structured environments to practise the conversations and decisions that define performance.
​
The goal is not just to simulate activity, but to embed expert judgement into every interaction and every piece of feedback.
The Disruption of Learning
AI is transforming work in waves: first automating tasks, then reshaping processes, and now redefining entire functions. Work is becoming a continuously evolving system in which humans and AI are deeply intertwined.
​
As roles shift, the useful life of skills is shrinking, placing new demands on learning. Adaptation is no longer episodic; it must be continuous.
​
Yet most learning systems are not designed for this. They remain content-driven, focused on courses rather than capabilities. Engagement is low, particularly among time-constrained professionals, and the impact on real-world performance is difficult to measure.
​
The consequences are clear. Learning functions struggle to demonstrate return on investment, particularly for behavioural capabilities. When impact cannot be proven, investment declines—often when adaptability is most needed.
​
The result is a reinforcing cycle: faster skill obsolescence, slower learning response, and widening performance gaps.
How Humans Learn
To respond effectively, it is necessary to return to first principles: how do humans actually learn?
​
In most organisations, learning systems remain largely passive. Content is delivered, frameworks are explained, and completion is tracked. Yet this has limited impact on performance.
​
Learning happens through practice: repetition, experimentation, and high-quality feedback. The challenge is not access to knowledge, but the ability to apply it.
​
This leads to a more fundamental question: if learning must be grounded in practice, what exactly should we practise?
​
Across leadership, management, and execution, value is created through interaction—the ability to influence, align, challenge, and support others. Strategy is realised and performance managed through conversations.
If you want to improve performance, you must improve the conversations that produce it.
Improving performance therefore requires the ability to practise these conversations—refining judgement and receiving feedback on what works.
​
Traditionally, this has been difficult to scale. It relies on expert facilitation and carefully designed experiential environments that are effective but limited in reach.
A Human-Centric Way Forward
Meeting this challenge requires a step change in how capability is developed.
​
AI provides the foundation for this shift.
​
For the first time, it is possible to simulate realistic, high-stakes conversations at scale. Individuals can practise the interactions that define their performance—leading difficult discussions, managing stakeholders, or integrating new technologies into their work.
​
This changes the learning dynamic. Practice becomes continuous rather than episodic, tailored rather than standardised, and embedded in the flow of work.
​
It also changes what can be measured. Behaviour becomes observable, and progress can be assessed through demonstrated capability—how individuals respond, adapt, and improve over time. This creates a more direct link between learning and performance.
​
AI alone is not sufficient.
​
Without careful design, simulated interactions risk becoming superficial and feedback generic. Activity may increase, but capability may not.
The difference is not the technology. It is the quality of the underlying expertise.
Effective learning requires clear standards of performance, structured progression, and precise feedback. Experienced faculty and learning professionals understand how capability is built—how to sequence challenges, calibrate difficulty, and reinforce standards.
​
With this design, AI becomes a powerful engine for capability development—scaling not just access, but the quality of practice itself.
Human Agency in an AI World
Used narrowly, AI may drive efficiency while eroding human relevance. Used thoughtfully, it can strengthen performance, build confidence, and enhance the interactions that define work.
​
The real opportunity is not simply to scale practice, but to scale expertise. AI can simulate conversations, but it cannot define what “good” looks like. That requires deep domain knowledge, experience, and pedagogical craft.
Machine learning is accelerating. Human learning must accelerate with it.
That requires a step change in how capability is built—embedding expertise into every interaction, every piece of feedback, and every learning experience.
​
The organisations that achieve this will not just keep pace with AI. They will outperform because of it.
Peter Zemsky
Founding CEO, Lexarius
Peter brings distinguished academic leadership and global perspective from INSEAD, where he served as Deputy Dean and Innovation lead. At Lexarius, he is shaping a new category of experiential learning, combining rigorous research with AI to transform how organisations develop human capability.