02 · Private, on-site AI
Your own AI. In your office. Off the internet.
Private AI gives you the capability of the big AI tools without sending a single byte of your business to someone else's servers. It runs on a machine in your office, or on private servers we manage for you: a model tuned to your work, with your knowledge base and retrieval built in, all talking to each other behind your own walls.
0%
of your data leaves your building
0%*
lower cost over 5 years vs per-seat cloud
~0 week
weeks to a working system
*Based on a team of 25 users replacing two $30/seat/month cloud AI tools with a private AI setup (one-time implementation + ~$400/month server cost), over 5 years with 10% annual price increases on cloud plans. Your numbers will vary.
/ How it works
Capable AI that never leaves your building, in about 3 weeks.
01
Week 1: we choose and tune a model for your work.
We start from a strong open model and tune it on your domain: your documents, your terminology, your tasks. The result performs like the tools you know, but specialised for what you actually do.
02
Week 2: we install it where you want it.
On a machine in your office, or as a private service on servers we run for you. Either way it's sealed off: your knowledge base and retrieval live alongside the model, and nothing is sent out to the public internet.
03
Week 3: your team is using it.
A familiar chat interface, connected to your internal documents and systems. It answers, drafts, summarises and searches across your files, privately, with no usage meter ticking in the background.
/ Cloud · Hybrid · Local
Cloud, hybrid, or fully on your own servers.
01 · Cloud
Maximum power, fastest to start.
The strongest models on the market, ready now. Best when raw capability matters most and your data rules allow it.
- Top models from OpenAI, Anthropic, Google
- Live in days, not weeks
- Pay for what you use
Data: processed in the provider's cloud
02 · Hybrid
Cloud power, your data kept private.
The strongest cloud models, but your documents and memory stay on your side. The AI answers with your context; outsiders never see what's yours.
- Big cloud models for the heavy lifting
- Your knowledge and memory kept private
- Resilient if a provider changes or goes down
Data: sensitive parts stay with you
03 · Local
Fully on your infrastructure. Nothing leaves.
Open models running inside your own walls. Total sovereignty and control, no recurring per-seat fees, no vendor lock-in.
- Open models: Llama, Mistral, Qwen, Gemma
- Zero data sent outside
- No recurring usage fees, no lock-in
Data: never leaves your building
Not sure which fits? That's the first thing we work out together, before any build.
/ The maths
Cloud bills climb. Yours stays flat.
Per-seat cloud subscriptions grow every year as your team and usage grow. Private AI is one predictable investment. The shaded gap is the money you keep.
/ Private vs cloud
What you get when the AI is yours.
Private AI
On your servers
Generic cloud AI
OpenAI, etc.
Data privacy
Usage fees
Speed at peak
Works offline
Ownership
Model stays put
/ Why it's better
Why private AI is worth owning.
01
Genuinely private.
Your contracts, client data and know-how never leave your walls. Nothing is sent to a third party, nothing trains someone else's model. For regulated or sensitive work, that's not a nice-to-have, it's the whole point.
02
Lower, predictable cost.
No per-seat subscriptions that climb every time you add a person. You pay to set it up and to run it, and the more your team uses it, the more sense the maths makes.
03
Tuned to you, not the average.
A model tuned on your work, with your knowledge base and retrieval wired in, gives sharper answers on your domain than a general tool, comparable to ChatGPT or Claude Sonnet on the work that matters to you.
04
It keeps working, on your terms.
No vendor can change the price, the rules or pull the plug. The system is yours: we set it up, document it, and can keep it healthy on a retainer if you'd like.
/ In their words
Where privacy and ownership matter most.
Our guest data never leaves the lodge now, and the whole team still has AI that knows everything about the property.
Our staff stopped digging through folders. They ask in plain language and the answer comes back in seconds, always from our own files.
/ FAQ
Questions about private AI.
- Is it really as good as ChatGPT or Claude?
- For general trivia, the biggest cloud models still lead. But for your work, your documents, your domain, your recurring tasks, a tuned private model with your knowledge base built in is comparable, and often more accurate, because it's specialised for you instead of everyone.
- Where does it run, and is my data safe?
- On a machine in your office, or on private servers we manage for you. Either way it's isolated from the public internet: your data stays inside your walls and is never sent to a third party or used to train another company's model.
- Is it cheaper than paying for AI subscriptions?
- Usually, once a team relies on it. There's a setup cost and a running cost, but no per-seat fee that grows every month. The more people use it, the better the economics compared with stacking up individual subscriptions.