ai services & data safety
AI Services & Data Processing
Last updated: 15 August 2026
Data we map before building
Before any AI workflow is built we identify the systems involved, the categories of data processed, the purpose of the automation and the third-party providers required. Automating an unmapped process just moves the risk somewhere less visible.
Output accuracy
AI systems produce probable output, not guaranteed output. They can be confidently wrong. We test against your real material before launch and tell you where the failure modes are, but no AI feature we build should be treated as an authority on its own.
Human review
Anything consequential — money, legal commitments, medical or safety-related content, or a final answer to a customer that cannot be retracted — stays reviewable by a person. We document where review applies before the workflow goes live, and we will not remove it to make a demo look smoother.
Third-party models and providers
AI features depend on third-party model providers whose availability, pricing, behaviour and terms can change without notice. A model update can change output quality on a system that worked yesterday. We name the providers a project depends on before launch and monitor for changes that affect you.
Transparency
Where a customer is interacting with an AI system rather than a person, that is disclosed. This reflects the transparency expectations of the EU AI Act, and it is also simply how we prefer to work: a customer who discovers they were misled about talking to a machine is a customer lost.
Sensitive data
We avoid sending sensitive data to AI tools unless there is a defined business need, a suitable provider configuration and a written agreement with the client. Do not send payment details, government identifiers or health information through the general quote form on this site.
Retention and access
For document, voice, email or CRM automations we document what enters the workflow, what each tool can access, where records are stored and how long they are kept. Project-specific data-processing terms are agreed before any production workflow handles client or customer data.
This page is a practical summary of how we work, not legal advice. Project-specific data-processing terms are agreed in writing before any production AI workflow handles client or customer data.