Private AI for lawyers, accountants, and clinics. Nothing leaves the building.
For practices where confidentiality is the job. If you’re the partner, the practice manager, or the office manager who keeps the front desk running — this is built for you. We deploy local AI systems — running on a server in your office, with zero traffic to OpenAI, Anthropic, or any third-party cloud. PIPEDA-clean by design. We’ll sign a data processing agreement before the first call ends.
Cloud or on-premise — your call.
Most of our work runs in the cloud. Some of it can’t. The first thing we figure out on a call is which build fits your business — not the other way around.
Cloud build
n8n plus Claude or OpenAI APIs. Fastest to ship, lowest cost, easiest to scale. Right for: marketing automation, content workflows, lead routing, public-facing chatbots, anything that doesn’t touch privileged or regulated information.
- check2–4 week typical build
- checkFrom $500 (FAQ chatbot) to $10K (multi-system)
- checkLow monthly hosting for servers + AI tokens, or self-host
On-premise build
Local LLMs (Llama, Mistral) running on a server in your office. n8n running locally. Zero network traffic to any third-party AI provider. Right for: legal, healthcare, dental, accounting, financial advice, and anything else where privileged client data lives.
- check4–8 week typical build (hardware shipping + setup)
- check$5K–$10K build + $2K–$3K one-time hardware (yours forever)
- checkNo API fees ever — zero cost-per-token, zero usage caps
- checkAir-gappable: works without internet if you want
Practices where confidentiality is the job.
If your regulator can ask where your client data lives, we build for that question.
Law firms
Intake forms, conflict checks, document summarisation, drafting assistants. Solicitor-client privilege never leaves your office.
Dental & medical practices
Recall reminders, intake automation, insurance claim drafting. PHI stays inside your office, on your network.
Accountants & bookkeepers
Tax season inbox triage, receipt classification, client-question drafts. Client financials never touch a U.S. server.
Financial advisors
Meeting summary drafts, KYC document handling, client communications. Compliant with provincial securities frameworks.
Creatives with private client work
Photographers, therapists, coaches. Client info and creative IP stay in your studio — we never train models on your work.
Anyone with a regulator
If a provincial body, college, or federal regulator can ask where your client data sleeps tonight, this is your build.
What “on-premise” actually means.
Plain English version — you don’t have to be technical to evaluate this. We explain it in your first call so you can take it back to your IT person or your regulator.
A server in your office, not a cloud account
A small dedicated machine (typically a Mac Mini or equivalent, ~$2,000–$3,000 one-time) sits in your office. Plugged into your existing network. We install and configure it. Your IT person, if you have one, can audit it any time.
Open-source models, not OpenAI
We deploy Llama, Mistral, or comparable open-source LLMs. They run entirely on your hardware. No API calls to OpenAI, Anthropic, or anyone else. No request ever leaves your network unless you explicitly route it to.
Workflow engine, also local
n8n — the workflow tool we build on — runs on the same server. Your automations live inside your office. They can read your Outlook, your practice management software, your file shares — whatever you give them access to — without ever phoning home.
Air-gap option
If you want, we configure the server with no internet access at all. Updates happen on a schedule we set together. Truly air-gapped: nothing in, nothing out, except by your hand.
PIPEDA, in plain terms.
We’re not lawyers and we don’t pretend to be. But we’ve built this product specifically so you can answer your regulator’s questions confidently.
PIPEDA-aligned by architecture
On-premise deployment means client personal information never crosses a border, never reaches a third-party processor, and never enters anyone’s training data. The architecture itself addresses most of PIPEDA’s 10 fair-information principles.
Data processing agreement on request
We sign a written DPA before any build starts when you need one. We’ve drafted them for cloud builds, and they’re even simpler for on-premise builds — there’s nothing for us to process.
Provincial health-privacy laws
PHIPA (Ontario), PHIA (New Brunswick), PHIA (Nova Scotia), HIPA (Saskatchewan), HIA (Alberta) — the on-premise architecture is built to work within any provincial framework. We’ll work with your privacy officer to validate specifics.
Audit trail
Every prompt, every response, every workflow run is logged locally with timestamps and user attribution. Your records, your retention policy, your discoverability if a regulator ever asks.
What an on-premise build actually costs.
All numbers up front. No discovery-call gotchas.
Build
One-time. Includes deployment, your workflows, training, documentation. You own all of it.
Hardware
One-time. Your asset, not ours. Typically a Mac Mini or comparable. Sits in your office.
API costs
No per-token, per-request, per-user charges — ever. The model runs on your machine, not someone else’s cloud.
Ongoing support options — security patches, model updates, and tweaks — are covered in your proposal. Most clients add it; it’s not required.
Questions practices ask before they buy.
Where exactly does my client data live?
Do I need to be technical?
What if the LLM hallucinates a fact?
Will you sign a DPA?
Can my paralegal use it from home?
How long does an on-premise build take?
When something breaks, we’re there.
Four things we’ll never do.
We never put your data on someone else’s GPU.
We never sign you to a SaaS lock-in.
We never build without a DPA in writing.
We never let the model phone home.
Built by LogicPros — 21 years in Atlantic Canada tech · Microsoft, CompTIA, Meta certified · Service NB alumnus.
Built for the work that can’t leave the building.
Book my free audit20 minutes. We tell you honestly whether on-premise is the right call for your practice.