A team ships a brutal quarter on adrenaline and goodwill. To say thanks, the company adds a mindfulness app and a Friday fruit bowl. Three months later two of the best people quit, exhausted, and everyone is baffled, the perks were right there. That gap between the gesture and the cause is the whole subject of this piece: most workplace wellbeing spend treats the symptom in the individual while the cause sits in the design of the work.

The quick version

  • Psychological health at work is driven mostly by working conditions, workload, control, support, fairness, not by employees' resilience or perks bolted on afterwards.
  • Burnout is a recognised occupational phenomenon. The leading researcher, Christina Maslach, is blunt: it is usually a sign of a defective workplace, not a defective worker.
  • The scale is real: the WHO and ILO estimate that depression and anxiety cost roughly 12 billion working days a year globally, about US$1 trillion in lost productivity.
  • The highest-leverage moves are organisational and free: cut unreasonable demands, give people genuine control, make support real, and manage change with consultation.

The idea in depth

Start with the thing leaders most often get wrong: where wellbeing actually comes from. The dominant academic model of work strain is the Job Demands–Resources (JD-R) model, set out by Evangelia Demerouti, Arnold Bakker and colleagues in "The Job Demands–Resources Model of Burnout" (Journal of Applied Psychology, 2001). Its core claim is clean: every job has demands (workload, time pressure, emotional load) and resources (control, support, feedback, clarity). Demands drive the exhaustion side of burnout; missing resources drive the disengagement side. Strain is what happens when high demands meet thin resources for too long.

That framing is useful precisely because it is diagnostic. Stop asking "is this person resilient enough?" and start asking "what is the demands-to-resources ratio on this role, this month?" A team carrying punishing demands with good control and real support is in a very different place from a team with the identical demands but micromanagement and no backup, even though the workload number looks the same on a spreadsheet. Resources are the lever you can actually pull without lightening the work.

Burnout is about the job, not the person

The most important reframe in this field comes from Christina Maslach, who built the standard measure of burnout. In The Burnout Challenge (Harvard University Press, 2022), Maslach and Michael Leiter argue that burnout is best understood as a mismatch between people and their jobs across six areas: workload, control, reward, community, fairness, and values. Their headline is uncomfortable for any leader hoping to outsource the problem: burnout is usually a signal about the workplace, not a defect in the worker. Treating it as personal weakness, and prescribing yoga to fix it, both misses the cause and quietly blames the casualty.

You cannot meditate your way out of an unmanageable workload. The fix for a workplace problem is to fix the workplace.

Run the six mismatches as a checklist on the actual situation, not the person. Is the workload sustainable, or chronically over capacity? Do people have a say in how they work, or only what they're handed? Is recognition real? Is the team a source of support or friction? Are decisions seen as fair? Does the work connect to something people value? Each "no" is a lever, and most of them cost nothing but a manager's attention and a few changed defaults.

flowchart LR
  A(["High job demands
workload, pressure, emotion"]) --> C{"Are the resources
there to match?"}
  C -->|"Yes, control, support,
clarity, fairness"| D(["Engagement
energy, focus"])
  C -->|"No, thin or absent"| E(["Strain
exhaustion, disengagement"])
  E --> F(["Burnout, errors,
people leave"])
The same demands lead to engagement or burnout depending on the resources a leader provides. Leaders Loop

An honest limitation: none of this means individual factors are irrelevant, sleep, life events, and a person's coping resources all matter, and a leader is not a clinician. The point is one of leverage, not totality. The conditions a manager controls are the largest modifiable input, and the one the organisation is actually accountable for. Personal support still belongs in the mix; it just cannot substitute for fixing the work.

The evidence, and a practical framework to act on

Two things make this more than an opinion. First, the scale: the WHO's "Mental health at work" fact sheet (updated 2024), drawing on joint WHO–ILO estimates, puts the global cost of depression and anxiety at roughly 12 billion lost working days a year and about US$1 trillion in lost productivity. Second, the direction of the fix: the WHO's Guidelines on mental health at work (2022) recommend organisational interventions, changing working conditions to reduce psychosocial risk, alongside training managers to support their teams. The official guidance points at the work, not just the worker.

For the "what do I actually look at" question, the UK Health and Safety Executive's Management Standards are the most usable framework going. They name six areas of work design that, managed badly, drive stress: demands, control, support, relationships, role, and change. Treat these as a quarterly risk review, give them the same seriousness you'd give a physical hazard. Where are demands unrealistic? Where do people lack control? Is support genuine or nominal? Are relationships toxic anywhere? Is anyone's role chronically unclear (a classic, fixable resource gap, see roles, responsibilities & decision rights)? Is change being done to people or with them?

mindmap
  root(("Psychosocial
risk areas
HSE"))
    Demands(["Workload &
pace"])
    Control(["Say over
how work is done"])
    Support(["Real backing
& resources"])
    Relationships(["Conflict,
bullying"])
    Role(["Clarity of
responsibilities"])
    Change(["Managed with
consultation"])
The HSE's six areas, a free, structured checklist for the conditions a leader controls. Leaders Loop

Note one limitation on the numbers: headline figures like "12 billion days" are global estimates with wide error bars, useful for scale and direction rather than precision. Resist the temptation to quote a tidy "every £1 returns £5" ROI line you saw on a slide; those vary enormously by source and intervention. The defensible claim is simpler and stronger: psychosocial risk is large, measurable, and substantially within management's control.

A worked example

Take a customer-support team, call it the Harbour desk. (Illustrative scenario; figures are illustrative, not from a real company.) After a product launch, ticket volume doubles. Headcount doesn't. The manager, wanting to look supportive, brings in a wellbeing webinar and a subscription to a meditation app. Six weeks later, two senior agents resign, average handle time has crept up, and the survey scores are worse than before the "wellbeing push."

Run the JD-R and HSE lenses over it and the cause is obvious. Demands roughly doubled; resources didn't move. The agents had no control over the queue, no extra support, and a role that had quietly expanded into firefighting. The webinar treated exhaustion as a mindset problem in people who were, in fact, drowning in a workload problem.

Now change the inputs the manager actually controls. She caps the daily ticket target and reroutes overflow rather than letting it pile silently onto individuals (demands). She gives agents authority to resolve refunds up to a threshold without escalation (control). She protects a daily 30-minute "no-tickets" block and pairs newer agents with a buddy (support). She makes the temporary nature of the surge explicit and consults the team on the staffing plan (change, fairness). None of this needs a new budget line. Within a few weeks the numbers settle, not because people got more resilient, but because the demands-to-resources ratio came back into range. The wellbeing app was never the lever. The work was.

Frequently asked questions

Isn't wellbeing really HR's job, or the individual's?

HR can build the policies and the safety net, and individuals own their own habits, but the day-to-day psychosocial conditions are set by line managers, because they control workload, deadlines, who does what, and how change lands. The WHO guidelines reflect this by recommending manager training as a specific intervention. The single biggest wellbeing input most people have is their direct manager's choices.

Aren't perks and EAPs a waste of money, then?

Not a waste, but the wrong end of the problem if they're all you do. An Employee Assistance Programme, counselling access, and flexibility are genuinely valuable as support and a safety net for people in difficulty. They just can't offset a job that is structurally unmanageable. Use them as the floor, not the strategy, and never as a substitute for fixing demands and control.

How do I spot a problem before someone burns out?

Watch leading signals, not just lagging ones. Cynicism creeping into a once-engaged person, withdrawal from optional collaboration, a quiet drop in quality from someone reliable, and rising sickness absence are early reads. A simple, anonymous pulse on the HSE areas, or the items in a people-analytics survey, makes the invisible visible. The cheapest signal is honest 1:1s where it's safe to say "I'm underwater."

If I cut demands, won't performance drop?

That assumes demands and output are the same thing, they're not. Exhausted people make more errors, take longer, and leave, all of which cost performance. The JD-R point is that you protect output by keeping demands and resources in balance, not by maximising demands. Sustainable pace usually beats heroic sprints over any horizon longer than a quarter.

What's the single most powerful thing I can do?

Give people real control over how they do their work, and protect them from demands they have no say over. Control is the resource the research keeps returning to, and it's the one managers most often withhold by reflex. Autonomy within clear boundaries does more for psychological health than almost any perk you can buy.

Related in the Toolkit

Where to go next