I wanted to share with you some thoughts and ideas about digital twins in the specific and rarely discussed context of cognitive work. But before that, and to avoid any misunderstanding, let’s start at the beginning and explain what a digital twin is before talking about what it could become.
So, before considering digital twins as applied to work, expertise, or organizations, a detour is necessary because the term is now sufficiently broad to encompass just about anything, from enhanced dashboards to anthropomorphic projections that are more narrative than engineering.
If we do not begin by clarifying what the concept of digital twin actually covers, everything that follows is likely to be read either as well-intentioned science fiction or as the resurrection of old HR promises that were never kept.
In short:
- The digital twin was born out of an operational need: to control complex physical systems without interrupting them, using an accurate and up-to-date digital representation.
- It differs from models, simulations, and dashboards in that it is continuously linked to reality via data flows, serving to act rather than explain.
- The uses of digital twins are already established in industry and concern critical environments, where they are mainly used to reduce decision-related risks.
- There are descriptive, predictive, and prescriptive digital twins, but their value depends above all on their integration into an observation-decision loop.
- To consider applications in cognitive work or organizations, we must first ask ourselves what we are trying to control and how this will affect the system concerned.
A concept born out of an operational problem
The digital twin was not born out of a desire to represent the world, and even less out of a desire to make it intelligible to decision-makers, but out of a very prosaic problem: how to control complex and critical physical systems without being able to stop them, dismantle them, or learn from mistakes.
It was in this context that the concept took shape in the early 2000s during work related to NASA systems. The challenge was to maintain a digital representation that was sufficiently accurate and up-to-date to reduce uncertainty in operational decision-making.
The digital twin is therefore inherently a control tool, not a knowledge tool. It does not seek to explain reality, but rather to enable action to be taken without disrupting its functioning. This industrial origin is not anecdotal, as it still influences how the concept is understood and used today.
What a digital twin is not
Much of the current misunderstanding stems from confusion between digital twins and older or simpler concepts.
A model describes a system based on assumptions, often deliberately simplifying reality to make it easier to manipulate. A simulation evolves this model in hypothetical scenarios in order to explore possible behaviors. Finally, a dashboard aggregates observed indicators to make a past or present situation visible, without claiming to directly influence its evolution.
The digital twin is distinguished by the fact that it relies on a continuous link with the real system, via operational data flows. It does not start from a theoretical state but from an observed state, and is not designed to illustrate or explain, but to be used in a decision loop.
Without this living link to reality, talking about digital twins is at best a misuse of language, at worst a conceptual illusion.
Stabilized uses due to constraints
It is tempting to present digital twins as an emerging technology still in search of use cases, which is factually incorrect. Their uses are already widely established in industry, critical infrastructure, energy, transportation, and complex asset management.
Twins are used to design products without multiplying physical prototypes, to monitor equipment in continuous operation, to anticipate failures, or to test operating scenarios without exposing the real system. Predictive maintenance is a good example of this logic: the digital twin makes it possible to assess the actual condition of a piece of equipment, project its likely evolution, and decide when to intervene.
What is striking when observing these uses is their simplicity, even their classicism. They concern constrained, critical, and necessarily costly systems, where decisions must be traceable and justifiable. The digital twin is a tool for reducing decision-making risk, not a lever for transformation.
A typology of digital twins
A distinction is traditionally made between descriptive, predictive, and prescriptive twins. This typology helps to clarify levels of maturity, but it should not obscure the essential point.
Regardless of its level of sophistication, a digital twin only exists when it is part of a loop connecting observation, interpretation, and decision-making. The question of whether it merely informs the decision or goes so far as to recommend actions is less about the nature of the twin than about the choices made around it in terms of automation, governance, and responsibility.
Bottom line
This first article therefore has no other ambition than to establish a framework for the future and avoid any potential future misunderstandings. What characterizes a digital twin is neither its ability to produce a narrative nor its claim to replace human labor, but rather the level of demand it imposes: cost, fragility, dependence on reality, and integration into a control system that is only valuable if it preserves the observed system.
If we then want to talk about twins applied to expertise, cognitive activities, or organizations, we will have to accept shifting the question. Not “can we represent?”, but “what are we trying to control, and with what consequences for the system being observed?”.
A digital twin is never interesting for what it represents, but for what it allows us to decide without degrading what it claims to improve.
To answer your questions…
A digital twin is a digital representation that is continuously linked to a real system through operational data. Its primary purpose is to support decision-making without interrupting or degrading the operation of the system being observed. Unlike a simple visualization, it is designed to act in a control-oriented manner, not to explain or describe reality.
A model is based on assumptions, a simulation explores theoretical scenarios, and a dashboard describes a past or present situation. The digital twin differs from these in that it has a living link to reality. Without continuous data flows and decision-making use, talking about digital twins is a misnomer.
The term has become very broad and is sometimes used to refer to analytical or narrative tools that have no operational link to reality. This confusion obscures the industrial origins of the digital twin and its limitations, fueling unrealistic expectations and science fiction-like discourse.
They are already widely used in industry, energy, transportation, and critical infrastructure. Their uses are straightforward and well defined: design, monitoring, failure prediction, and predictive maintenance. Their main purpose is to reduce decision-making risk in costly and constrained systems.
Before any transposition, it is necessary to clarify what we are seeking to control and with what consequences. A digital twin is only valuable if it is part of a demanding and responsible decision-making loop. The key issue is not representation, but the impact of control on the observed system.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
People Centric Operations: adapting work and operations to knowledge workers
People Centric Operations 2.0: how AI is reinventing knowledge work at scale
Big Data : why you’ll love to hate it
Big Data: A Revolution That Will Transform How We Live, Work, and Think
How analytics help to make better fact-based HR decisions
Figures and ROI : please admit when you have no clue
