Organisational digital twins introduce a fourth approach: learning by simulating. Instead of implementing a new incentive system and waiting to observe the consequences, managers can explore alternatives in a simulated environment. Instead of restructuring teams and hoping co-ordination improves, they can test different designs before making commitments. Instead of launching a major reskilling initiative, they can compare multiple approaches virtually.

The most important contribution of organisational digital twins may be the creation of a new mode of organisational search.

An organisational digital twin is not a dashboard, reporting tool, or analytics platform. It is a simulation environment built from firm-specific data and AI agents capable of approximating how organisational actors, processes and routines behav

Advances in data infrastructure and generative AI are making it possible to integrate these traces into a shared organisational memory. This creates the foundation for what I have called an organisational digital twi

Organisational charts reveal formal structures. But much of organisational life unfolds through informal networks, routines and patterns of co-ordination that are difficult to observe directly.

If codified selves represent individuals, organisational digital twins represent the interactions among individuals.

This raises questions that most organisations have never confronted. Can a firm continue using a digital representation of an employee after that employee departs? Should workers share in the value created by their digital twins? What rights should employees have over systems trained on their expertise? As AI becomes more capable of representing individuals, these governance questions are likely to become increasingly important.

A codified self emerges through a process of co-creation. The individual contributes expertise and data. The organisation contributes infrastructure, complementary assets, organisational context and technological capabilities. The resulting system is neither purely an employee asset nor purely a firm asset.

At the same time, codified selves create a new category of asset whose ownership is inherently ambiguous

At the same time, important limits remain. Many forms of human capital are tacit rather than articulated. Judgment developed through decades of experience, moral reasoning, intuition, leadership presence and context-sensitive decision making remain difficult to codify.

Workers may be able to extend their influence far beyond the limits imposed by time and attention.

Once aspects of expertise have been codified, they can be deployed across many interactions simultaneously. A senior engineer can help teams around the world without joining every meeting. A salesperson’s negotiation expertise can be made available across markets. A professor’s insights can reach far more students than would otherwise be possible.

The significance of codified selves lies in their ability to create separation between value creation and physical presence. Historically, organisations could benefit from an employee’s expertise only when that employee was available. Human capital and value creation were tightly coupled in both time and space. Codified selves change that relationship.

It is a representation of those aspects of human capital that become visible through organisational activity.

Importantly, codified selves do not simply codify knowledge. They can also capture procedural expertise, communication styles, reasoning approaches and some aspects of personality and interpersonal behaviour, provided those characteristics are expressed repeatedly through digital interactions

A top engineer’s approach to diagnosing technical failures, a consultant’s way of framing client problems, a professor’s teaching style or a founder’s approach to evaluating opportunities may all create value precisely because they are distinctive. Traditional AI systems tend to average out these differences. Codified selves seek to preserve them.

A codified self is an AI model trained on the articulated knowledge, skills, abilities and other characteristics of a specific individual in a specific organisational context. Unlike traditional AI systems that aggregate data from thousands or millions of people, a codified self attempts to preserve the distinctive characteristics of one person.

As individuals perform their jobs, they generate enormous volumes of digital traces: emails, presentations, documents, videos, meeting transcripts, software code, customer interactions and countless other artifacts. Embedded within these traces are not only facts and knowledge, but also aspects of how individuals reason, communicate, solve problems and interact with others. The concept of the codified self helps explain what happens when AI learns from these traces.

The traditional view of human capital rests on two assumptions. First, workers retain ownership of their human capital. Second, organisations can create value from that human capital only when workers are actively engaged in work. Artificial intelligence is beginning to weaken both assumptions.

Yet the study also uncovered an important constraint. Employees consistently rated responses as less helpful when they believed they came from AI, even when the response had actually been written by the chief executive. This was especially true if the question was of an inter-personal nature.

The results were revealing. Employees were not able to reliably tell whether they were interacting with the boss or with the AI. In other words, the Wade Bot passed the Turing Test with flying colours. Viewed narrowly, the finding suggests that AI is becoming remarkably effective at mimicking specific organisational actors

he possibility that organisations can deploy certain forms of human capital even when the individual who created that capital is not physically present. T

AI can increasingly learn from the digital traces of a specific individual and begin to represent aspects of that person’s expertise, reasoning and communication style