Private AI · Built and supported in Perth
Your team is already using AI.
Where are the documents going?
Nobody approved it. There is no policy. But the work is faster with it, so client files, patient notes, financial records and tender pricing are being pasted into free chatbots — and transmitted to servers you do not control, in countries you did not choose.
OnPrem removes the question entirely. We build private AI systems that run on hardware in your office. Same capability. Nothing transmitted, nothing retained by a third party, nothing offshore.
No obligation. We will tell you if you don’t need us.
The part nobody wants to check
You cannot ban this. You have already tried.
Prohibiting AI does not stop it — it just stops people telling you about it. The tools are free, they work on a phone, and they genuinely save an hour on a bad afternoon. Every firm we talk to believes their exposure is smaller than it is, because the people creating it have no reason to mention it.
Pasted in chunks to find the relevant material before a deadline.
Dictated detail submitted to produce a structured clinical record.
Uploaded to explain a variance, carrying the client’s whole position.
Submitted for polish — the exact material that decides whether you win.
Each of these is a disclosure of confidential information to a third party, usually processed overseas. Under the Australian Privacy Principles you generally remain accountable for personal information after it has been handed to an overseas recipient — being unaware it happened is not a defence, and in most cases you will have no way to prove what was sent or what became of it.
Two ways to answer it
The usual approach
Policy, training, and hope
- — Write an AI policy. Circulate it. File it.
- — Depend on every staff member following it on their worst, busiest day.
- — Rely on a vendor’s terms of service, which they can change.
- — Have no record of what was disclosed if you are ever asked.
- — Accept that the productivity gain goes to whoever ignores the policy.
The architectural approach
Make the disclosure impossible
- — The AI runs on a machine inside your office.
- — Requests never touch the internet. It works with the connection unplugged.
- — No third-party account, no terms of service to depend on.
- — No per-message cost, so nobody is discouraged from using the safe option.
- — Staff get the productivity gain, sanctioned, with nothing leaving the building.
Who this is for
Firms that hold something they cannot afford to leak
The obligation differs by profession, but the exposure is identical: useful AI requires real detail, and real detail is exactly what you are bound to protect.
Law Firms
Confidentiality and privilege are not policies you can update
Read more →Medical & Allied Health
The turnover exemption does not apply to you
Read more →Accounting & Advisory
Tax file numbers carry their own regime
Read more →Engineering & Resources
Client NDAs rarely contemplate a third-party AI service
Read more →What we will not tell you
This does not suit everyone, and we will say so.
A private system runs open models on hardware with real limits. Large models run well; the very largest run slowly or not at all. If your work genuinely needs frontier reasoning on novel problems, a commercial service will do it better, and we would rather you heard that from us than found out after committing.
If two people occasionally use AI for non-client work, you do not need this. A clear policy and a paid business subscription is a proportionate answer, and it is a cheaper one.
The case for on-premise is strong when a whole team would use it daily on confidential material, when a single disclosure would be a notifiable breach, or when a client contract simply does not permit the data to go anywhere. If that is not you, we will tell you in the first conversation.
Common questions
- What actually is an on-premise AI system?
- A dedicated machine that sits in your office and runs capable open AI models locally. Your team uses it much like any AI chat tool — through a browser on your own network — except the request never leaves the building. There is no account with an overseas provider, no per-message fee, and no copy of your documents on someone else’s server.
- Is it as capable as ChatGPT?
- For the work that fills a professional’s day — summarising long documents, drafting from your own material, extracting structured data, answering questions across your files — the open models available now are genuinely strong, and closer to the commercial services than most people expect. For frontier reasoning on novel problems, the large commercial models still lead. We will tell you which of your use cases fall on which side before you spend anything.
- How much does it cost?
- It depends on how many people use it and what you need it to do, so we scope it rather than list a price. What we can say is that it is a capital purchase with a support arrangement, not a per-seat subscription that grows every time you hire — and it does not increase when your team uses it more.
- What if we decide it is not right for us?
- Then we will tell you that ourselves. For some firms a written AI policy and a paid business subscription is a proportionate answer, and we would rather say so than sell you hardware you do not need. The assessment conversation costs you nothing either way.
- Do we need IT staff to run it?
- No. We configure it, install it, and support it. From your team’s point of view it is a page in the browser. If you already have an IT provider we work alongside them rather than around them.
No obligation
Find out whether this fits your firm
Tell us how your team is using AI today and what you cannot afford to have leave the building. We will tell you honestly whether an on-premise system is the right answer — including when it is not.