Posted on March 31st, 2026

As AI continues to evolve, a clear shift is emerging.

For years, progress has been driven by pre-training — feeding models vast amounts of internet-scale data to build general intelligence. But as we move toward more advanced applications, particularly in physical AI and specialised enterprise use cases, a new priority is taking centre stage:

Post-training.

This shift has important implications — not just for AI developers, but for Learning & Development professionals.

The Difference: Pre-Training vs Post-Training

To understand the shift, it helps to clarify the distinction.

Pre-Training

Pre-training is about breadth. AI models are trained on massive datasets to learn general patterns: language, concepts, common knowledge, and broad problem-solving. This is equivalent to giving someone a strong academic foundation.

Post-Training

Post-training is about depth and application. It uses experience-based data to refine a model for a specific context: real-world interactions, task-specific scenarios, decision-making patterns, and domain expertise. This is equivalent to developing professional capability through experience.

Why Post-Training Is Now Critical

Two major trends are accelerating this shift.

1. Physical AI: Learning Through Simulated Experience

In robotics and physical systems, general knowledge is not enough. A robot may understand "what a tool is," but that does not mean it can use it correctly, adapt to changing environments, or execute tasks safely and efficiently. To solve this, organisations are using digital twins, simulations, and human-guided interactions. These generate large volumes of experience data, which are then used to refine models through post-training.

2. Vertical AI: Specialisation Through Real-World Data

In the digital world, organisations are discovering something similar. General AI models are powerful — but not always optimal for specific tasks. Instead, companies are taking base models, feeding them domain-specific interaction data, and post-training them for specialised performance. In many cases, these focused models now outperform more general systems within their domain.

The Underlying Shift: From Knowledge to Experience

Across both physical and digital AI, the same principle is emerging:

Performance is no longer driven by general knowledge alone — it is driven by structured experience.

This reflects a broader trend in AI known as the next phase of the "bitter lesson" — where systems improve not through handcrafted rules, but through scalable learning from experience.

Why This Matters for L&D

This shift should feel familiar. Because it mirrors a long-standing challenge in organisations: training provides knowledge, but performance depends on experience.

L&D teams often invest heavily in content, courses, and frameworks — yet still face slow time-to-competence, inconsistent performance, and over-reliance on experienced individuals.

The missing layer is the same one AI is now prioritising: structured, transferable experience.

The Link to Expert Performance Modelling

This is where Expert Performance Modelling becomes critical. Your top performers do not just know more — they recognise patterns others miss, make faster and better decisions, apply judgement in context, and adapt under pressure.

This is post-training in human form. But in most organisations it remains tacit, is rarely structured, and is difficult to scale.

Expertise as a Service (ExaaS): Building the Post-Training Layer for Organisations

The ExaaS approach addresses this directly. It focuses on capturing how experts actually think and perform, structuring that into decision models and performance frameworks, and turning that into reusable, scalable assets.

In effect, it creates the organisational equivalent of post-training data. As highlighted in the broader AI shift: the tacit knowledge, decision frameworks, and workflows of top performers are exactly the type of "experience data" needed to refine intelligent systems.

Digital Twins: Where It All Comes Together

When structured correctly, this expertise can be deployed through digital twins, decision-support systems, coaching frameworks, and performance tools.

These function as the bridge between human expertise and AI capability — not replacing people, but scaling how they think.

The Strategic Implication for L&D

The role of L&D is evolving.

From delivering knowledge — to structuring and scaling expertise.

This means moving beyond content creation, focusing on how performance actually happens, capturing experience and not just information, and enabling faster, more consistent capability development.

Final Thought

AI is teaching us something important. General knowledge is no longer enough. Whether in machines or organisations, performance improves when experience is captured, structured, and reused.

For L&D, this presents a clear opportunity: to become the function that turns expertise into a scalable asset.