Policies that actually change anything

Most workplace mental-health provision is a poster, an app, and a number nobody has rung. This is a short account of what has evidence behind it, what does not, and the uncomfortable part in the middle.

Start with the uncomfortable part

No amount of wellbeing provision compensates for conditions that are making people ill.

If the workload is genuinely undeliverable, the hours are punitive, the pay does not cover rent, or a bullying complaint went nowhere — resilience training is not a neutral gesture. It relocates the problem into the employee, and people notice.

The single most useful question a board can ask is not "what support do we offer?" but "what about working here is making people unwell, and what are we prepared to change?"

What has support behind it

Manager training that is specific. Not awareness-raising. Training in what to say, what to do next, and what the manager's actual duty is. Trials of targeted manager training show effects on sickness absence, which is more than most interventions manage.

Genuine control over how work is done. Job control is one of the best-evidenced protective factors in occupational health, and it is a design decision rather than a benefit.

Predictable hours and predictable rest. Particularly for shift work and on-call.

A route to raise workload that changes something. The route mattering less than whether anything visibly changes as a result.

Adjustments that a manager can make without escalating each one. Most reasonable adjustments die in an approval queue.

A financial hardship route that is not the line manager. Money is one of the most common paths to a crisis and one of the hardest things to disclose upward. Money, debt and mental health →

Peer support and access-to-work style schemes, where they are resourced rather than voluntary-on-top-of-a-full-job.

What is mostly decoration

An EAP number nobody has tested. Ring your own. A striking number lead to a voicemail, a triage script, or a six-week wait for four sessions. That is not an argument against having one; it is an argument against counting it as provision without checking.

One-off resilience or mindfulness sessions. Not harmful. Not a substitute for anything, and resented when offered instead of addressing conditions.

Awareness days without a follow-through. A day of green ribbons in an organisation with no disclosure procedure teaches people that this is a communications exercise.

Anonymous surveys that produce no visible change. After the second one, response rates and honesty both fall, and you have made things slightly worse.

Things that are cheap and disproportionately useful

• Test your own crisis numbers annually. Helplines renumber and close. A poster with a dead number is worse than a blank wall.
• Write down who a disclosure may be shared with, and tell people. Predictable confidentiality raises disclosure more than encouragement does.
• Name a deputy for whoever owns this. A single point of failure on leave is the most common way a good process fails.
• Decide your postvention plan now. Nobody can write it on the day. After a death by suicide at work →
• Support the supporter. The manager holding a difficult situation is routinely offered nothing.

A caution about data

Do not build anything that scores individuals for suicide risk, and be sceptical of vendors offering it. Risk prediction is not accurate at the level of one person; a "low risk" flag is actively dangerous, and a "high risk" flag attached to an employee record is a discrimination problem with a person inside it.

Monitoring hours worked is legitimate and useful. Inferring someone's mental state from their keystrokes is not.