For employers

Evidence the pathway works.
Never a window into it.

You get proof that support exists, is reached, and completes. You never get a transcript — there isn’t one to give you.

Utilisation

By site and by swing. The number your EAP reports annually, except you can see it move week to week and against roster phase.

Risk-tier counts

How many conversations reached elevated or crisis tier, over time. Trend, not individuals — the unit is the event, never the person.

Handoff completion

Offered, and taken. This is the number that matters and the one nothing else in your stack can produce: proof the referral pathway actually closed.

No message content. Ever.

The safety ledger has no free-text column — not redacted, not access-controlled, absent. There is nothing to leak, nothing to subpoena, and nothing for a manager to ask you for.

No rollout project.

The reason this doesn’t need a business case for the IT department.

  • No account provisioning — no seats to buy, assign, or reclaim
  • No identity integration — nothing to connect to your directory, because there are no identities
  • No offboarding — someone who leaves simply stops using it; there is no account to close
  • No helpdesk load — there are no passwords, so there are no password resets
  • You hand out a code — one per site or per crew, however you want the reporting split

The duty you already
have.

WHS regulations 55A–55D place a positive, enforceable duty on employers to manage psychosocial hazards. The WA code of practice names FIFO accommodation as a workplace and isolated work as a hazard.

The hard part of that duty isn’t having a support pathway. It’s demonstrating that one exists, that people reach it, and that it completes — and doing so without collecting the very thing that would stop them using it.

We won’t tell you an app discharges that duty. Under the same regulations the hierarchy of controls puts a support tool near the bottom — job design, rostering and supervision sit above it. Any vendor who tells you otherwise is selling you a liability, not a control.

The evidence underneath.

Every figure traceable to a named source. Anything we couldn’t evidence was left out rather than softened.

33%

of FIFO workers score high or very high psychological distress on the K10 — against 17% in a WA benchmark group and 10% in an Australian male norm group.Curtin University for the WA Mental Health Commission · n = 3,108

28%

of remote mining and construction workers show high or very high distress, against 10.8% in the general population.Bowers et al. · Medical Journal of Australia, 2018 · n = 1,124

1.45×

the suicide incidence rate of other Australian workers, among mining workers, 2012–2019. 95% CI 1.24–1.70.King et al. · Safety and Health at Work, 2023

What we won’t quote at you. The widely repeated “twice the rate” suicide figure is construction, not mining. The mining-specific study puts the true rate somewhere between 11 and 25 per 100,000, because miners are misclassified as construction workers in coronial data. We use the ratio above, which is the number that holds.

Where your data is.
And where it’s going.

Asked properly, this is two questions — where it’s stored, and where it’s processed. Most vendors answer only the first. Here are both, as they stand today.

Storage — Singapore

Your safety ledger, your content and your site information pack sit in ap-southeast-1. Not Australia, today. Worker conversations are not among them — those stay encrypted on the handset.

Processing — United States

Our safety layer currently executes on US infrastructure by platform default, and the models that write the reply and score the risk run there too. A message passes through; none of it is stored at either end.

Retention — nothing is kept, and nothing trains on you

No provider in our chain trains on your conversations. Where a provider offers a formal zero-retention agreement we hold one or are securing it, and we’ll show you the current status per provider rather than a blanket claim.

All three are moving onshore.

The database will move to Sydney, ap-southeast-2. The Gateway — the code that scores every turn — will pin to Sydney with it. And the models will move to AWS Bedrock’s au.* geographic profiles, which pin inference to the ANZ boundary. That last one only became possible in April 2026, and it’s the reason this is a plan rather than an aspiration.

We’ll change the tense on this page the day each one completes — not before. If onshore residency is a gate for you, ask us where it’s up to and we’ll tell you precisely.

Voice won’t follow it, and that’s worth knowing now. Speech providers don’t offer Australian residency at all, so a spoken conversation will still leave the country afterwards. On that half the lever is retention terms rather than location. If residency matters more to you than voice does, typed chat carries the identical Spine — and voice can be switched off per company.

On certification,
plainly.

The question every procurement team asks third, and the one most vendors answer with a logo rather than a sentence.

We don’t hold ISO 27001 or SOC 2. We’re a small team, and at this size a formal audit would cost more than the engineering it audits. It’s on the roadmap for when the customer base justifies it, and we’ll say so the day it starts — not before.

What we’d rather be judged on in the meantime is the part a certificate doesn’t reach. A certificate says a process was followed. These are properties of the system — including where it runs — and you can test every one of them.

There is no personal information to breach

No name, email, phone number or employee number is ever collected. Not minimised, not encrypted — never asked for. The most robust control over personal data is not holding any.

The ledger cannot hold a conversation

Not by policy or permissions, but by database schema — there is no free-text column to write one into. A misconfiguration cannot expose what the table has no room for.

On-device encryption you can watch us prove

An automated test writes a message, closes the database, reads the raw file off the disk and asserts the text isn’t in it — alongside a control test that fails if the encryption is removed. We’ll run it in front of you.

What we won’t claim.

Worth having in writing before your procurement team asks — and worth asking every other vendor for.

  • No engagement, ROI or incident-reduction percentages. We have no pilots and no customers yet, so any number would be invented
  • No named customers and no testimonials, for the same reason
  • No ISO 27001 or SOC 2 — we hold neither today, and we won’t imply an audit is underway when it isn’t
  • Not “clinically validated”. It is built on validated instruments and reviewed against them; that is a different sentence and we use the accurate one
  • Not a replacement for your EAP, and not a medical device

The questions
you’re going to ask.

In the order they usually come up. If yours isn’t here, it’s the one we most want to hear.

What happens when it misses something?

It will. No system, human or AI, catches everything, and any vendor who tells you otherwise is describing a product that doesn’t exist. What we can say is that two independent detectors have to both miss it, one of which cannot be talked out of a match, and that every failure in our 211-scenario suite is written down and handed to you rather than averaged away.

If it’s anonymous, how do you escalate?

We don’t, and we say so in the app. Rusty cannot call 000 for someone it cannot identify. What anonymity buys is that they talk at all, and take the handoff themselves — and the ledger proves the pathway was put in front of them every time. That trade is deliberate and it is the honest one for this workforce.

What’s our liability if something goes wrong?

A question for your counsel and not for a website, and we would rather have it early than late. What we bring to that conversation: it is disclosed as AI at onboarding and in every conversation, it never presents as a clinician, it is positioned as a complement to your EAP rather than a replacement, and there is an auditable record that support was offered at every elevated turn.

Can a manager get at what someone said?

No, and not because of permissions. The safety table has no free-text column — not redacted, not access-controlled, absent. There is no report to run, no export to request and no support ticket that produces one. If someone in your business asks us for it, the answer is that it does not exist.

Our crews won’t use employer software.

That assumption is correct and it is the design problem, not an objection to argue with. There is no login, no email, no employee number and nothing that connects a conversation to a person — because the only version of privacy a sceptical workforce believes is the one where there is nothing to look at.

How long does rollout take?

There is no rollout. No account provisioning, no directory integration, no seats and no offboarding. You hand out a site code and a worker is talking in under a minute. The trial is two weeks, full access, and you start it by emailing us.

Does it replace our EAP?

No. If it works, your EAP sees more of the right people, earlier — which is the outcome we are actually selling. If a vendor tells you their chatbot replaces clinical care, walk away from them, and from us if we ever say it.

Two weeks.
Full access. No strings.

Bring the liability question and the what-happens-when-it-misses question. Those are the ones we’ve done the work on.