AI Box · Private AI deployment
Private AI, designed to run inside your own environment.
AI Box is Ramerion’s concept for an in-company AI deployment: an assistant intended to be configured around your internal workflows and approved knowledge, and to run within your facility or private infrastructure.
The final architecture — and exactly where data may flow — is agreed with you during technical discovery.
- Intended to be configured around your workflows and approved knowledge
- Access designed to follow the permissions you already use
- Intended for your facility or private infrastructure, subject to the final architecture
01In plain language
An AI assistant configured for your organisation, intended to run in an environment you control.
- What it is
- A private AI environment for a single organisation, designed to be configured with that organisation's approved documents, permissions and workflows.
- Who it is for
- Organisations that want AI assistance but need it to stay close to their own systems and governance — engineering, operations and documentation-heavy teams in particular.
- What it is not
- A public chatbot, or a substitute for qualified judgement. Its outputs are answers and drafts for people to check.
02How it fits together
People, approved data and a private AI environment — inside one boundary.
Your teams ask questions, the AI environment is intended to work only with the sources you authorise, and the results come back as governed outputs for people to review. Where that boundary sits is decided with you.
01 Company users
- Engineering teams
- Operations & maintenance
- Administrators
02 Authorised data sources
- Approved document libraries
- Engineering records
- Internal knowledge bases
03 Private AI environment
- Access & permissionswho can ask, and what they may see
- Retrieval over approved dataintended: only the sources you authorise
- Model runtimemodel choice agreed per deployment
- Logging & governancerecords and policies you define
AI Box
04 Governed outputs
- Responses for people to review
- Draft documentation
- Usage and audit records
Human review. Outputs are intended to go back to the people who asked, to be checked before anything is relied on.
03Where it could help
Potential use cases
These are directions for a deployment, not a list of shipped features. What an individual AI Box does is agreed during discovery.
Internal knowledge retrieval
Potential use caseAsk plain-language questions across approved internal material — standards, procedures, project records — with responses designed to point back to the source documents so they can be checked.
Engineering documentation assistance
Potential use caseDraft, summarise and cross-reference engineering documentation such as method statements, inspection notes or handover packs, with an engineer reviewing every output before it is used.
Organisation-specific workflows
Potential use caseAssistants shaped around one particular process — finding your way around a project archive, preparing a maintenance briefing, triaging internal requests — designed to work within the permissions your teams already have.
04Deployment
Decided with your team.
No two environments are the same, so none of the decisions on the right are made in advance. Each one is worked through with your technical team during discovery, and written down before anything is installed.
- Infrastructure
- Where the system runs — hardware in your facility or private infrastructure you control — and the capacity the agreed use cases need.
- Model choice
- Which model or models suit the use case, the available hardware and your licensing requirements.
- Access controls
- How users sign in, and how the deployment respects the permissions already set on your data.
- Integrations
- Which document stores, databases or engineering systems it may read from, and in which direction data may move.
- Logging & governance
- What is recorded, who can review it, how long records are kept and how usage policies are applied.
- Updates
- How model, software and security updates are delivered, tested and approved in your environment.
- Support
- Who operates the system day to day, and what support Ramerion provides alongside your IT team.
05Privacy & data flow
The principles we design to.
Designed for private, in-environment deployment
The intended deployment places the AI environment inside your facility or private infrastructure, rather than on a shared public service.
Your chosen environment as the boundary
A core design goal is for customer data to remain within the environment you choose. Any connection beyond it is identified, discussed and approved with you during discovery.
Designed around what you authorise
The deployment is intended to work with the data sources you approve, under the permissions you set — not everything it could technically reach.
Boundaries in writing
Deployment and data-flow boundaries are defined during technical discovery and documented, so your security and IT teams can review them.
What this page is — and what it is not
This page describes
A concept and the design goals behind it: where an AI Box is intended to run, how it is intended to treat your data, and what gets decided with you.
It is not
A security certification, a compliance attestation or a guarantee of isolation. Specific security and privacy properties are confirmed for each deployment, in writing, once its architecture has been reviewed.
On AI output
Responses can be incomplete or wrong. They support — and never replace — responsible engineering review, sign-off, and the codes and standards that apply to your work.
Start with discovery
Discuss an AI Box deployment.
Tell us about the workflows, data sources and constraints you have in mind. The first step is a technical conversation about whether — and how — a private deployment fits.