AI and workplace analytics: what are we actually measuring?
The volume of data that organisations collect about their working environment is growing rapidly. Sensors measure occupancy and indoor climate; access systems record attendance; booking systems provide insight into the use of workstations and meeting rooms; and digital platforms contain information on communication and collaboration. Artificial intelligence enhances the ability to combine these various data sources, recognise patterns and investigate correlations..
This offers interesting possibilities. Data can help us use buildings more efficiently, tailor working environments more closely to actual usage, and gain a better understanding of the conditions employees need to do their jobs effectively.
At the same time, a fundamental question arises: when do we use data to better understand how the working environment functions, and when do we use that same data to interpret or assess how employees are performing?
From measurement to interpretation
In The Atlantic, Ellen Cushing (2026) describes the development of AI applications for employee monitoring. Whereas traditional monitoring primarily records observable activity, such as log-ins, computer usage or attendance, newer applications can, for example, analyse voice usage, facial expressions and digital communication, and attempt to draw conclusions from this about engagement, stress, mood or behaviour.
In doing so, AI goes a step further than existing workplace analytics. Data from access passes, booking systems and occupancy sensors can already be combined to reveal patterns in usage and behaviour. AI expands the possibilities for analysing and interpreting a wide variety of data. It thus makes it possible not only to analyse what is happening, but also to interpret what that behaviour might mean.
The risk of interpretation
The risk arises when a correlation that has been identified is interpreted as a causal link. Low attendance does not automatically mean low engagement, and low use of meeting spaces does not mean that there is little collaboration. This problem is not new, nor is it specific to AI. We are already familiar with this from workplace analytics: a low occupancy rate for a workspace does not necessarily mean there is no need for it. Perhaps that space is uncomfortable, noisy or poorly located.
AI does not alter this methodological principle. What AI does change is the scale on which data can be combined and patterns identified. This allows more and more complex correlations to become visible, but these do not yet constitute an explanation. The distinction between what we measure, the correlations we identify within that data, and the conclusions we draw from it therefore becomes even more important.
The opportunities of AI for workplace analytics
It is precisely in the combination of data sources that the potential of AI lies. Occupancy and booking data provide insight into usage and demand; employee surveys into perceived functionality, quality and support; and indoor climate data into the physical conditions.
AI can reveal correlations between these data sets that would otherwise be difficult to detect. For example, a model might identify that satisfaction with certain work zones declines when occupancy is high or noise levels are high. Patterns relating to presence, communication, interactions and the use of different workspaces can also be taken into account.
More advanced AI applications even attempt to infer stress, mood or engagement from signals such as voice usage, language or facial expressions. Such emotion recognition in the workplace is, in principle, prohibited under the European AI Act, with limited exceptions for medical or safety purposes (European Parliament & Council of the European Union, 2024).
This means that workplace analytics can go beyond simply recording what happens and help investigate the circumstances under which certain patterns arise. This can help organisations identify more precisely where the working environment effectively supports work and where there is room for improvement.
One key principle is essential here: use workplace analytics primarily to assess the performance of the working environment, not that of the employee. How often employees use certain workspaces is relevant information in this regard. Combined with data on availability, perceived functionality and quality, this can provide insight into how well the working environment supports the work.
It is not the data itself, but the purpose, the level of analysis and the interpretation that determine what we do with it.
From measurement to understanding
AI makes it possible to combine more data sources, recognise more complex patterns and explore new correlations. But ultimately, the value of workplace analytics does not lie in how much data we can collect or how many patterns AI can uncover within it. The value arises when these analyses help us to better understand how the working environment supports the work and where there is room for improvement.
This requires a clear distinction between measurement, correlation and conclusion. What have we actually measured? What patterns do we see in the data? And what sound conclusions can we draw from this?
The most important question, therefore, is perhaps not what AI can measure about employees, but what we actually need to measure in order to improve the working environment.
References
Cushing, E. (2026, May). The rise of emotional surveillance. The Atlantic. The Atlantic
European Parliament & Council of the European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union.
Nassau is happy to work with you to find the right solution
Every working environment requires a tailored approach. In a brief, no-obligation consultation, we’ll work with you to identify your organisation’s needs and explore the options that best suit them.
