Private AI that keeps your data where it is
Many teams want to use AI on contracts, case notes or research data, but cannot send that information to a third-party service. We deploy a language model on infrastructure you control and test it on one real use case.
Who it is for
- Law firms, accountants and advisers handling client documents
- Clinics and care providers with patient or service-user information
- Research groups with unpublished or restricted data
- Suppliers to the public sector with contract rules on where data may go
What stays in your hands
- The model runs on your servers or in your own private cloud tenancy
- Your documents are not used to train anyone else's system
- Access is controlled per person and logged
- Anything the AI wants to change needs a person to approve it first
The five-week pilot
The pilot answers one question: does this help with a real task of yours? You get a working system and a plain-language verdict.
Week 1: scope and data review
We agree one use case, look at the documents involved and decide where the system will run.
Week 2: deployment
We install and secure the model and the supporting search, and connect it to a small set of your documents.
Week 3: testing on your work
Your team tries real questions. We record what works, what does not and how often answers need correcting.
Week 4: access and approval controls
We add user access, audit logging and the human approval step for any action the system proposes.
Week 5: handover and decision
A usage guide, a short report and a recommendation: stop, extend or roll out.
Questions people ask
Does our data leave our environment?
The system is designed so that it does not. We document where data is stored and which connections exist, so you can check this yourself before and after the pilot.
Which models do you use?
Open-weight language models that can run on your own hardware. We choose and test the model against your use case during scoping, and tell you if a hosted service would suit you better.
What hardware do we need?
It depends on the model and the number of users. Options include a GPU server you own, a rented dedicated GPU server in the UK or a private cloud tenancy. We size this with you before you spend anything on hardware.
Does this make us compliant with data protection law?
No supplier can do that for you. Keeping data inside your environment makes your obligations easier to meet, and we document the system so your data protection lead can assess it.
What happens after the pilot?
You decide. You can stop with a working system and documentation, extend the pilot to more use cases, or move to ongoing support and a wider rollout.
Not sure it is worth it? Book a free scoping call. If a cloud AI service would meet your needs more cheaply, we will say so.