E6.1Included
Requirements
What the model needs to do, how many people use it and where the data must stay.
Detail E6 — Private AI
AI that keeps your data inside the business.
Some data shouldn't go to an outside AI provider at all: client files, health records, legal documents, unreleased designs. Open-weight models can run on infrastructure you control, in the UK or EU, so nothing leaves. We help you choose a model, set it up and connect it to the tools your team uses.
- Price
- On request
Fig. E6.1 — A private set-upNot to scale
Your teamChat · Documents · Workflows
Open-weight modelP1On infrastructure you control
Your dataNever sent to an outside AI provider
Notes — When it makes sense
Who needs private AI.
- 01Businesses handling sensitive client data, like legal, financial, medical or HR work.
- 02Contracts or clients that forbid sending data to outside AI services.
- 03Teams with high, steady usage, where running your own model can cost less than paying per use.
- 04Anyone who wants to keep working if an outside provider changes its prices or terms.
Specification — The trade-offs
What to weigh up honestly.
- Privacy and controlYour data stays where you decide, and the model doesn't change unless you change it.
- CapabilityThe best open-weight models are very capable, but the largest commercial models still lead on the hardest tasks.
- Running itSomeone has to keep it running and updated. We set that up and can look after it.
Table E6 — Scope
What a private set-up includes.
E6.2Included
Model choice
Open-weight models such as Llama, Mistral, Gemma, Qwen or OpenAI's gpt-oss, tested on your own tasks.
E6.3Included
Infrastructure
A set-up you control, on your own hardware or private cloud servers in the UK or EU, sized for your usage.
E6.4Included
Your documents
Search over your own files, so answers draw on your information without it leaving.
E6.5Included
Access for your team
A chat interface and connections to your tools, with sign-in and permissions.
E6.6Included
Security and updates
Access controls, logging and a plan for updating the model and software.
General notes — FAQ
Questions about private AI.
Is a private model as good as ChatGPT?
For many everyday tasks, like drafting, summarising and searching documents, a good open-weight model does the job well. For the hardest reasoning tasks, the top commercial models are still ahead. We test on your tasks so you can judge for yourself.
What does it cost?
It depends on the model size and how many people use it. There's a set-up cost and a running cost for the hardware or servers. We'll compare it with paying a commercial provider before you decide.
Can we mix private and commercial AI?
Yes, and many businesses should. Sensitive work goes to the private model; everything else can use the commercial tools your team already likes.
Fig. E6.2 — Next step
Need AI that keeps data in-house?
Tell us what kind of data you handle and what you'd want the model to do. We usually reply the same working day.