Healthcare

Your people did nottrain to do paperwork.

Referrals, coding, claims and the forms nobody has time for. We build the systems that take that work, safely, and without your records leaving the country.

Operations, opened · An illustrative organization

Act oneFour systems

One patient, spreadacross four systems.

The referral, the record, the clinic call, the signed form: four places, none of which knows about the other three. All four are places someone on your team can already open, so we read them at that person's access level and write the result back into the same system.

Fax

Referrals in

EMRHL7

The clinical record

Calls and clinics

PDF

Paperwork

Referrals in

Reads
The referral letter, whoever sent it and however it arrived
Writes back
A structured referral on the right patient, triaged and dated

Read at the permission level of the person it works for

Act twoJoining the dots

The fax, the chart and the claimare one patient.

Your longest-serving admin already knows that. None of the four systems holding the pieces do.

Worked example · invented data

Reading what you already have…

FAX AND MAILTHE EMRCLINIC CALLSSCANNED FORMS

Nothing joined up yet

Already in a systemWas only ever in an email, a call or a PDF
The Canadian problem

The best models think down South.

This is the real blocker, and it is not a policy problem. Frontier models are served from data centres in the United States. Inference is not storage, but a prompt carrying a patient’s referral still crosses the border to be read, and for a great many Canadian custodians that single fact ends the conversation.

So the answer is not a better contract with an American endpoint. It is to stop moving the health information and move the model instead. That is what AI Space exists to do, and healthcare is the reason we built it.

None of this is legal advice, and we do not offer it. Your privacy officer decides what your organization may do; our job is to build a deployment they can say yes to, and hand them the documentation they need to decide.

Where it runs

Canadian regions, on infrastructure you can name in a privacy impact assessment.

What runs there

Open-weight frontier models, the ones you can actually host, kept current as new weights ship.

How you reach it

One OpenAI-compatible endpoint. Swapping the model under a live workflow is configuration.

When that is not enough

The same models run fully on-premises, inside your perimeter, on your hardware.

Where to go next

Every referral stops at a named clinician before it is filed, and the record is built for someone else to check.

The way in is the map: your departments, your systems, and where the paperwork actually sits.

Bring your privacy officer

Book a discovery call.

Start with the map of your organization, or with the one job that hurts. Measured in hours and money, and everything we build stays yours.

Loading form…