Private AI infrastructure for professional firms
Private AI for firms that can't afford to leak.
Magpie runs private, dedicated AI clusters for professional firms. One tenant per cluster — yours. Nothing is logged, nothing is used for training, and the rules are set by the firm in a one-page policy document.
An Australian company · Australian data residency available as an option
01 · The problem
The problem is already inside the firm
Fee-earners are already using AI on client work — often through personal accounts, on personal terms of service. Most firms have no policy, no visibility, and no good answer when a client, an insurer, or a regulator asks which tools touched which files. Australian law societies have published guidance on AI use in legal practice, and the question lands on the managing partner's desk.
Personal accounts
Client documents flow through consumer AI tools the firm doesn't control, on data terms nobody at the firm agreed to.
No visibility
Without a firm-run service there is no record of who used what, on which matter, under which rules — nothing to show a client or an auditor.
An expected answer
Regulator guidance on AI in legal practice is now published and specific. "We told staff to be careful" is not a governance position.
02 · What Magpie does
What Magpie does
We set up and run a dedicated AI cluster for your firm, then hand you the controls. Your team keeps using the AI tools it already knows — pointed at infrastructure the firm governs.
A dedicated, single-tenant cluster
Your firm is the only tenant. No other customer's traffic shares your hardware. Nothing is logged, and nothing is used to train any model.
A one-page policy document
Which model runs, the data-handling rules, and per-person quotas — set by the firm in a single plain-language document. When you want a rule changed, that page changes.
A dashboard in plain sight
A practice manager opens a browser and sees who used what, what it cost, and which models are running. No terminals, no jargon.
A weekly usage report
One page each week: seats used, volume, cost, peak times — the evidence pack for deciding whether the service earns its place.
03 · The pilot
The way in: a two-week pilot
Deliberately small. One team of 10–25 people, two weeks, then a decision with evidence in hand — not a slide deck.
-
1
Set up with you
We stand up your firm's cluster and draft the one-page policy document with your practice manager in a single working session.
-
2
Two weeks of real work
Your pilot team works on the cluster with its own keys. You watch the dashboard and receive a weekly usage report.
-
3
A day-14 decision meeting
We sit down with the two weekly reports and the policy document and decide together: continue, or don't.
-
4
Walk away whole
If the answer is no, there is no lock-in and no auto-renewal. You leave with a full export of your configuration, policy, and keys.
The pilot is free. It runs on a smaller open model — enough for your team to judge how the machinery works on real work. If you continue, we price the model tier that fits your firm, with the full cost breakdown in hand before you decide anything.
04 · Said straight
What we won't claim
-
I
We won't claim to be cheaper.
A dedicated private cluster is not cheaper than a vendor seat plan, and we won't pretend otherwise. The case is control and confidentiality — a firm that can say exactly where its AI runs, what it keeps, and who sets the rules.
-
II
We won't overstate the privacy.
The honest claim is this: a dedicated single-tenant cluster, nothing logged, nothing used for training, the policy set by you. Australian data residency is available as an option for firms that require it.
-
III
We won't pretend to be a big vendor.
Magpie is a small Australian company. Support is business-hours, the pilot terms are written down, and the exit export is real — because in this profession, overclaiming is how you lose trust.
05 · The founder
Who's behind Magpie

Magpie Net is founded and run by Jack Pertzel, a Melbourne-based founder. Before Magpie he founded and ran Holistic Help AU, a registered NDIS provider — a sector where client confidentiality is a legal obligation — and built the company's own invoicing and operations software when vendor tools fell short.
Magpie works with a small number of firms at a time, personally. The person on the call is the person who sets up your cluster.
06 · Next step
Book a 20-minute call
Twenty minutes to see whether a pilot fits your firm — what's on your files today, what a private cluster would change, and what the two weeks would look like. If it's not a fit, you'll know quickly.
Also in the family
Contribute compute
Magpie Net also operates an open testnet where individuals contribute spare computing capacity to a shared network — the other half of the family. Learn more about contributing compute.