Machine learning is becoming a buzz word for many organisations, with some not really knowing how and where they can best utilise it. With more companies installing IoT devices into their office spaces, there is an opportunity to use machine learning to further examine the rich datasets that are generated from these devices. Using data from devices such as sensors in combination with machine learning models, can help organisations better understand employee behaviour or when machines may need maintenance, when office spaces could need cleaning, or how meeting rooms could be used better in the future to optimise real estate. A lot of these algorithms are built into apps or the ‘Digital Twin’, so little manual work is needed to analyse this data. However, there is still scope to extract data from IoT devices and make more personalised machine learning models to inform workplace decisions. Below is some research that highlights some of the ways in which machine learning is being used or could be used in the workplace in the future:

1. Air quality prediction and forecasting

One of the ways in which machine learning is being used in smart buildings, is to estimate and forecast the air quality inside meeting rooms. Research indicates that a normal office environment can actually have rather high levels of CO2 which decreases our cognitive skill by up to 70%[1]. Furthermore, according to the US Environmental Protection Agency (EPA), air pollution if often between two and five times greater inside than it is outdoors. With the accurate prediction of air quality using historical data, it allows building managers to plan ahead, decrease the effects of harmful air pollutants on health as well as the costs associated, and create a cleaner and healthier environment. With machine learning, organisations would also be able to better understand the triggers for poor air quality conditions, allowing them to ensure these potential causes are identified and mitigated early on.

2. Predictive maintenance and cleaning

Machine learning is also being used for real time and historic room and building performance benchmarking, allowing organisations in the long term to minimise equipment downtime and costs associated with maintenance. Sensors can be installed on machines and can be connected to an IoT platform to stream data about the machine’s vital statistics in real time. Using the data, ML models can be built using historical data to predict failure, and managers can be alerted of future maintenance needs. Similarly, machine learning can also be used to predict when meeting rooms require cleaning, using data from sensors that can understand when rooms are more likely needed to be cleaned due to higher occupancy levels or learning from historical data that highlights previous cleaning habits.

3. Facial recognition and time management

Facial recognition software can be used to detect employees who are not working or who leave early from work and this will allow organisations to track their employees (in terms of utilisation levels) and help them improve on time management. Time management solutions company, EZHRM, based in India, launched software that allows companies to keep a record of the location of their employees, helping them to keep their staff organised, improve productivity and log pay roll more effectively. The software also arguably aims to help reduce company costs. For instance, by seeing how much their staff are actually working, this can help them to reduce overheads, so that they can focus on other areas of their business that will increase profits. However, some may find this as an intrusive form of data collection, and a global uptake of this solution is unlikely to be popular amongst many people, especially those that do not work full-time in the office, and therefore may be seen as ‘underutilised’ by their employers.

4. Facial recognition and meeting room attendance

Organisations are increasingly adopting facial recognition authentication based on machine learning technology for connection to meeting rooms in office spaces. These authentication devices can be equipped with a 360-degree camera which uses facial recognition to scan the room and identify meeting participants.  With microphones integrated, they can be used to pick up a participant’s voice and is able to transcribe every word spoken in a meeting in real-time. The voice recognition/transcription works with any language, which is helpful for companies where employees are working with teams that may be located in different regions across the globe.

5. Predicting space usage

Machine learning is also being used to forecast how often meeting rooms could be used. WeWork built a neural network where every time the model was fed a space layout, it began to learn the relationship between layout and usage [2]. Eventually it understood this relationship well enough that it could accurately predict how a layout would be used by their members before they began construction. They even tested this approach by providing the algorithm with layouts that it had not seen before, and the outcome was still fairly accurate. They estimated that the neural network is 40% more accurate than human designers in predicting how frequently a meeting room will be used by the building’s occupants. Building algorithms like these are powerful, as organisations can technically plan a space that fits the needs of employees before they actually occupy it. Despite this, the role of an architect is always paramount in making real estate decisions, however, models like these could still be used to make more informed and accurate space decisions.

References

[1] https://ehp.niehs.nih.gov/15-10037/

[2] Designing offices with machine learning – Ideas (wework.com)