How we use AI
The rules every AI system we build follows, and what AI does on this website. Last updated 22 October 2026.
In the AI we build for clients
AI is optional in the Custom Systems we build. When a system uses it, the AI runs at an outside provider the client approves, never on the client's server. Whatever the system, it is set up to these rules before it goes live:
- It says what it is. An assistant tells the people it talks to that it is an assistant, not a person.
- It answers from approved content. It draws only on content the client has approved, and hands difficult questions to a person.
- A person reviews what matters. Anything sent to customers or published is approved by a person. An assistant that acts prepares the booking, the reply or the update, and a person confirms it.
- Its limits are written down. What the AI may and may not do, including when it hands over to a person, is agreed with the client in writing before we build.
How data is handled
When a system uses AI, requests go from your system to the provider that runs the model, through our connection or your own provider account. We tell you exactly what leaves your premises. We store usage records (time, model, amount used), not the content of your conversations. We only keep conversation content if you ask us to while we troubleshoot something, and only for a short time. We'll tell you which AI providers process your data, and we sign a data processing agreement with every client. If you use your own AI account, your provider's terms apply to that account.
On this website
The Ask Glint assistant on this site is a demonstration with pre-written answers. Nothing you click in it is sent to us or stored.
When we add an AI assistant that takes enquiries on this site, we will update this page and our privacy policy first, to say what it collects and how it is used.
Questions
Ask us anything about how an AI system we built works, or how we would build yours: info@gridandglint.com. See also Custom Systems.