Platform / Data privacy

Keep control of your data.

Choose where SmoothOperator.ai runs, which models it may use beyond its native private one, and what, if anything, leaves your environment.

Deployment options

Choose the boundaries your work requires.

Every implementation starts with one assessment that covers where your data is stored, which models can reach it, where inference runs and which systems are connected. We then agree a managed cloud, a cloud you control, or your own local infrastructure, based on what the application needs. The controls are set before anything is released.

Native Private

Nothing leaves. The whole platform runs inside your perimeter, on your own infrastructure or fully air-gapped. Private models do all the work, and no prompt or document crosses your network boundary. Each customer has an isolated environment with separate knowledge, configuration and records.

Clear View Hybrid

The best model for each task, on your terms. Use frontier models where they add value and private models everywhere else. One governed gateway decides what goes out, to which provider, and records every call. External providers run on your own keys, only where your policies and compliance rules allow.

Anonymized Hybrid

The model sees the work, never the client. External models receive the substance of the task with identifiers replaced by consistent tokens inside your environment. Results are mapped back locally. The mapping key stays inside your environment.

Anonymization layer

Use external models without exposing sensitive data.

Many organisations aren’t allowed to send client or personal data to external AI providers, whether because of a contract or their own policy. The anonymization layer replaces names, financial figures and other identifiers with consistent tokens inside your environment. The external model works on the tokenized data, and the results are mapped back locally. This is often the quickest first project.

See how the anonymization layer works
  1. 01 / Inside your environment

    Replace identifiers locally.

    Before anything crosses your perimeter, names, figures, account numbers and references are replaced with consistent tokens. The structure, relationships and meaning are preserved while the original identifiers stay inside your environment.

  2. 02 / External model

    Work with tokenized data.

    The external model analyses the tokenized dataset. It can compare, rank, summarise and reason across the work without receiving the original identifiers.

  3. 03 / Back in your environment

    Map the results back locally.

    The answer is restored inside your perimeter, with tokens mapped back to the names, figures or references they represent. The mapping stays inside your environment and is not shared with the external model.

Control and oversight

See what the AI can access and what it did.

You decide what the AI can access and do: its permissions and scopes, the tools it is allowed to use, and where approval is required. Every result goes to a named reviewer along with its evidence. The execution trace and audit log remain available to your auditors and administrators.

Your knowledge stays yours

Your knowledge is never shared.

The knowledge you bring is recorded and versioned for your organisation only. It is never shared with other customers and never used to train anyone else’s model.

Responsibilities

Clear ownership of every approval.

Legal, privacy, security and regulatory approvals stay with the decision-makers you name. We go through the relevant compliance requirements with you, identify any gaps, and agree responsibilities, evidence and acceptance criteria before release.

Tell us where your data has to stay and which models you need, and we’ll propose the deployment that fits.

Discuss your data requirements ↗

Have a business challenge AI could solve, and want a production-grade solution to match?

Tell us about your vision for a solution so we can accelerate your time-to-market.

Let’s talk ↗