Concentration, comfort and control: what the 2026 Hassell workplace report means for HSE
Page Published Date:
September 23, 2026
As AI moves concentrated work outside the office, psychosocial risk assessment needs to follow.
Hassell's Too Connected to Concentrate report surveyed 2,500 knowledge workers across Australia, Hong Kong, Singapore, the UK and the US, and found the workplace is being pulled in two directions at once: a slow physical reset in where people work, and a fast-moving digital disruption in how they work, driven by AI.
Read through an HSE lens, several of its findings map directly onto how psychosocial risk is now regulated in Australia.
The physical environment still shapes how people experience work
Hassell's data links basic building conditions, temperature, lighting, cleanliness, WiFi reliability, access to outdoor space, directly to how positively people rate their workplace overall. The report identifies a set of "critical" problems that are both common and strongly tied to poorer perceptions: buildings that are too hot or cold, slow lifts, and a lack of greenery or outdoor space among them. This is a useful reminder that environmental design sits inside the same risk picture as psychosocial hazards.
Safe Work Australia's Model Code of Practice: Managing Psychosocial Hazards at Work defines a psychosocial hazard as anything arising from the design or management of work, the work environment, plant, or workplace interactions, that may cause psychological harm. Poor thermal comfort, inadequate lighting and a lack of quiet space can contribute to fatigue, low job control and sustained low-level stress, which the Code requires PCBUs to identify and manage with the same rigour as physical hazards.
Where AI is starting to change the picture
Hassell found that more than eight in ten respondents believed they could use AI effectively almost anywhere with internet access, and reported using it in settings ranging from desks and meeting rooms to beds, sofas, bathrooms and public transport. Younger workers and frequent AI users were the most likely to work this way. For HSE teams, this signals work is becoming untethered from the controlled environments organisations have traditionally designed and risk-assessed.
This is not a hypothetical concern. The Work Health and Safety Amendment (Digital Work Systems) Act 2026, passed by the NSW Parliament in February 2026, made the use of AI and algorithmic systems in work allocation an explicit WHS duty, inserting a new obligation (s.21A) requiring PCBUs to ensure that work allocated by digital systems does not put workers at risk.
Sources differ on the exact commencement date of the operative provisions, so this should be confirmed against the current legislation before being relied on. It is a narrower provision than the picture Hassell paints, addressing algorithmic work allocation specifically rather than AI use in general, but it reflects the same underlying shift: regulators are beginning to treat the technology shaping how and where work happens as a live WHS consideration.
This is a NSW-specific reform; organisations in other jurisdictions should check their own state's current position rather than assume equivalent duties apply.
There is also an Australia-specific angle here, drawn from the report's charts rather than a stated figure. Of the four markets shown on Hassell's longitudinal graphs, Australia appears to have the lowest rate of full-time office attendance alongside the US and UK, sitting behind Singapore.
Read alongside the finding that AI use is spreading into homes, beds, cafés and commutes, this suggests a larger proportion of AI-enabled work in Australia may be happening outside any environment an employer has assessed, compared with markets where more work still happens in the office. That is a directional inference rather than a precise cross-market statistic, but it strengthens the case for extending psychosocial risk assessment beyond the physical workplace.
What people say would help
Asked what changes would most help them use AI well at work, Hassell's respondents pointed to more space for learning and mentoring, more space to focus, more privacy and more quiet, all ranked above requests for more technology.
Each of these sits within the hazard categories Safe Work Australia's Model Code already recognises, including job demands and job control. Giving people better conditions to concentrate and to seek support functions as a control measure for hazards regulators already expect to be managed, not only as a productivity measure.
The takeaway for HSE
As AI decouples work from any single location, HSE risk assessment cannot stay anchored to the desk. The environments where people are now doing concentrated, high-stakes work, including at home, on the move, and increasingly wherever a phone and a prompt will do, sit further outside the traditional office than most risk frameworks currently account for.
A practical starting point is a location audit specifically for AI-prompted tasks: where they happen, on what device, under what lighting and posture, and whether that setting was ever part of an existing risk assessment. Most organisations already have a remote-work policy. Few have tested whether it actually covers the places AI is being used.
