Consultation is not evidence because it happened. It is evidence because it showed you something you could not previously see.
On 2 October 2026, the People at Work platform closes. Organisational data will be removed and cannot be recovered.
Since 2020 it has been the free, regulator-endorsed psychosocial risk assessment survey in Australia, used by around 5,000 businesses and completed by more than 160,000 workers. New accounts closed on 1 June. The last new surveys could be launched on 1 July. After October, it is gone.
For a lot of organisations, that survey was the psychosocial risk process. Not a component of it. The whole thing.
Which makes this a useful moment to ask a question most organisations have never asked: was it working?
What the closure actually tells you
The regulators' stated reason is straightforward. Since People at Work launched, there have been changes in legislation, newly developed tools and updated research on psychosocial hazards, and the survey may no longer fully meet the needs of organisations seeking to explore psychosocial hazards in their workplace.
Read that carefully. A validated, evidence-based instrument, jointly funded by every Australian WHS regulator, has been retired because the ground moved underneath it.
If that can happen to the best-resourced tool in the country, it can happen to your engagement survey, your annual pulse check and the questions your managers have been asking in one-to-ones since 2019.
The instrument is not the process. It never was.
The duty was never the survey
Under section 35 of Victoria's Occupational Health and Safety Act 2004, employers must consult, so far as is reasonably practicable, with employees who are or are likely to be directly affected, when identifying or assessing hazards or risks, and when making decisions about the measures to be taken to control them.
Employees must be given a reasonable opportunity to express their views, and those views must be taken into account. Where employees are represented by a health and safety representative, that representative must be involved. Where agreed consultation procedures exist, they must be followed. Failing to consult is an offence, with penalties attached.
The Occupational Health and Safety (Psychological Health) Regulations 2025 do not prescribe how any of this is to be done. They point back to the existing duty.
That silence is not a gap. It is the point. The design of the process is left with the employer, which means the adequacy of the process is also left with the employer, and so is the burden of demonstrating it.
The removal of a survey platform changes none of that.
The question that separates the two
Two organisations can both consult. Both can produce a response rate, a report and a set of themes. And they can end up with completely different pictures of their own risk.
The difference is not effort. It is what the process was designed to do.
A process designed to confirm returns confirmation. A process designed to reveal returns things nobody wanted to hear.
So the question to put to your own process is not "did we consult?" It is "what did our consultation actually help us see?"
If the honest answer is that it confirmed what leadership already believed, you have not produced visibility. You have produced reassurance. We looked at why that distinction matters in Psychosocial Hazard Identification: Why Visibility Is a Governance Problem.
Four tests
These are the tests we apply to a client's existing consultation process before recommending anything. You can run them on your last round of consultation this week, with no new tooling.
1. The new information test
Take your last consultation output and mark every finding that leadership did not already believe before the process started.
If the answer is none, the process functioned as a confirmation exercise. That is not a failure of the workforce. It means the instrument only had the capacity to return what it asked about, and it only asked about what was already suspected.
2. The specificity test
Read your findings and ask whether they could belong to any organisation in your sector, or only to yours.
"Workload pressure in peak periods" could belong to anyone. "Rostering changes in November mean the intake team absorbs both the backlog and the new referrals, with no handover" belongs to one organisation.
Generic findings produce generic controls. Generic controls leave the actual exposure unmanaged, because the actual exposure was never described.
3. The disclosure risk test
Identify the hazard in your workplace that would cost someone the most to raise. Not the most serious hazard, the most expensive one to name.
Then ask whether your consultation process gives that person a route to raise it without that cost. If the only channels are a manager conversation, a team meeting, or a survey small enough that responses could be attributed, the answer is no.
This is the test most processes fail, and it is the one that determines the value of everything else. A process that surfaces only the safe hazards is a process that has told you about the safe hazards.
4. The traceability test
Pick one finding from your last consultation round. Follow it forward.
Can you show what was identified, what was decided, what control was implemented, who owns it, and when it was reviewed? Can you show what was fed back to the people who raised it?
If the trail stops at the report, consultation happened and nothing followed. Under section 35, views must be taken into account. A finding that went nowhere is difficult to describe as having been taken into account, and impossible to evidence.
It also has a cost you will pay next round. People who raise things that go nowhere stop raising things.
What to do instead of replacing the survey
The instinct after October will be to find another survey. That is the wrong first move, because it repeats the assumption that caused the problem: that an instrument is a process.
A defensible identification process draws on more than one source.
The records you already hold. Incident reports, complaints, grievance files, turnover by team, absence patterns, exit interviews, and the free text nobody reads. Most organisations are sitting on more evidence than their survey will ever produce.
Structured conversation, not open invitation. "Let us know if anything's a problem" is not consultation. Specific questions about specific work, asked of the people doing it, are.
A channel that carries no professional cost. Confidential by design, not confidential by promise. This is what makes the disclosure risk test passable.
Design for the exposures your instrument does not name. Every framework has blind spots. If your workforce includes people whose exposure your framework was not built to describe, that is precisely where a survey will return nothing. We wrote about one example of this in Cultural Load Is a Psychosocial Hazard, Not a Diversity Issue.
None of this requires a platform. It requires a process someone has designed on purpose.
Before October
Two practical things, if People at Work is in your history.
Export your reports before 2 October 2026. After that date organisational data is removed from the platform and raw data is not available for download. That is your psychosocial assessment history and there is no recovery option.
Then run the four tests above against what you exported, before you decide what comes next. The reports are more useful as evidence of how well your process worked than as evidence of your risk.
Test your own process
Our self-check walks through whether your consultation is identifying the full range of psychosocial hazards, or only the ones that were easiest to see.
Download: Are You Covered? A Plain Language Self-Check for Employers
No form. No email required.
Working out what comes after People at Work? Book a conversation with Burriin.