AI Use Policy

Why this policy exists

We build AI-first software. AI is only useful if you can trust what it does with your data, so this page sets out how we design, test and run the AI systems we build for clients.

Humans in command

Every consequential action an AI system takes on your behalf has a clear approval path, with a full audit trail. AI accelerates judgement; it does not replace it.

Your data stays yours

• [State your training policy, e.g. client data is never used to train shared models without written consent].
• [State the data protection practices you have implemented, e.g. encryption in transit and at rest, role-based access].
• [State your IP and data-ownership terms: you own what you pay for].

Tested before trusted

AI outputs are evaluated against real examples before they go live, and against metrics agreed with you upfront: time saved, error rates, response times. [Describe your evaluation process]

Choosing models

We choose the AI models that fit your task, budget and data rules. [Confirm this is your approach, and list any providers you commonly use if you want to disclose them]

Being honest about limits

We assess feasibility, data readiness and expected impact before building anything, and we will tell you honestly if AI is not the answer. We start with one contained pilot and only scale what earns it.

What we will not build

[List the uses you will decline, e.g. systems designed to deceive people, unlawful surveillance, or decisions about people with no human review]

AI on this website

This website does not use AI to make decisions about you. Messages sent through the contact form are read by our team. [Confirm, or describe any AI tools you use to process enquiries]

Questions

If you have questions or concerns about how we use AI, write to [contact email].