How we handle your data, explained plainly.

Where it lives, who can see it, and what you keep at the end.

01 · Where your data lives

Your environment, your choice.

We fit the system to where your data already is, rather than moving it somewhere new.

01

In your own cloud account

Most systems we build run in your cloud account, under your logins and billing. You can see everything that is running and switch it off at any time.

02

On your own servers

Where data should not leave the building, we can run the system on-premise or on a local device, including models that run inside your network.

03

Short pilots, agreed in writing

If a quick pilot is easier to run on our side first, we agree that in writing, keep it separate from other clients, and move it to your environment at handover.

02 · What we commit to

Four things you can hold us to.

01

Access only to what the work needs

We ask for the narrowest access that lets the workflow run, start with read-only where we can, and give access back when the engagement ends.

02

Your data is not used to train models

We do not use client data to train or fine-tune models we reuse elsewhere. Where a hosted AI provider is involved, we use business terms with training turned off.

03

Confidential by default

We sign an NDA on request before we see anything sensitive, and we never name clients publicly without their permission.

04

You own the code and the system

The code, configuration and documentation are yours. At handover you get the repositories, credentials and a written guide, so there is no vendor lock-in.

03 · Day-to-day practices

Ordinary habits, applied every time.

None of this is unusual. What matters is doing it consistently, including on small projects where it is tempting to skip a step.

Passwords and keys kept in a secrets manager, never in code.Data encrypted in transit and at rest using your platform’s standard tools.Permissions based on roles, mapped to your existing user groups where possible.Every change reviewed before it goes live.Infrastructure written as code, so environments can be checked and rebuilt.Dependencies checked for known vulnerabilities as part of delivery.A person approves any automated action with financial or contractual impact.AI-assisted answers show their sources; automated actions are logged.
04 · Common questions

What security teams ask, answered plainly.

What data do you need access to?

Only what the workflow needs. We map this together at the start and use sample or anonymised data for early work wherever possible.

Do you use outside AI providers?

Often, yes. We tell you which provider would process which data before anything is sent, and use business terms with training turned off. If that is not acceptable, we use models that run inside your own environment instead.

How do you handle personal data?

We keep it to a minimum. Personal details are left out of AI prompts unless the task needs them, masked where possible, and handled under a written agreement covering purpose, retention and deletion.

Can our data stay in a specific region?

Yes. We can set up hosting and AI models in the region you need, or run everything on your own servers.

Do you hold security certifications?

Not at present. We are a small team, so we describe our practices plainly instead. We work inside your environment and controls, fill in security questionnaires honestly, and say clearly which controls are ours and which are yours.

What happens at the end of an engagement?

We hand over the code, documentation and credentials, remove our access, and delete any copies of your data we hold, then confirm this in writing.

How do we report a security concern?

Email [email protected] with “security” in the subject line. We reply within one business day and agree any disclosure timeline with you before anything is published.

Send us your security questionnaire. We fill it in completely and mark anything we cannot show evidence for as not in place, rather than giving an optimistic answer.

Bring your security team to the first call.We would rather design around your rules than guess them.

An AI Opportunity Audit covers where your data lives and what can and cannot leave your environment, before anything is built.

  1. Pick one workflowThe one that costs you most time.
  2. Working in 4 weeksOn your real data, with your team.
  3. Run it every dayWe keep improving it with you.