Legal
On this website
The chat demo is scripted. It is labelled “AI demo (scripted)”, follows a fixed script written by people and makes no calls to any AI model. Nothing you click or type in it leaves your browser.
The writing on this site was written and edited by people. [If you use AI to draft content, say so here.]
Images and illustrations were drawn for this site and are not AI-generated. [Update if that changes.]
In the agents and automations we build
Every system we deliver follows the same rules, written into the project scope:
People know they’re talking to AI. Customer-facing agents introduce themselves as AI assistants and offer a person at any point.
People approve what’s expensive to undo. Refunds, payments, deletions, contract changes and first contact with a new customer wait for a named person’s approval. Clients can make these rules stricter, never looser without a written review.
No advice we’re not qualified to give. Agents do not give medical, legal or financial advice; those questions are handed to people.
Logged and reviewable. Every run is logged so a person can see what the system did and why. Logs are kept for [30 days] unless the client asks otherwise.
Your data stays yours. We use model providers’ business terms that exclude training on client data, and we never train our own models on it. We tell clients which providers process their data before we start.
Tested before and after launch. Each agent is evaluated against real cases before it goes live, and accuracy is reported weekly afterwards.
Models and providers
We choose models per project after testing them against the client’s own evaluation set. Providers we commonly evaluate include [OpenAI, Anthropic, Google and open-weight models hosted in the client’s cloud]. Naming them describes our practice; it doesn’t imply a partnership or endorsement.
Questions or concerns
If you think one of our systems behaved in a way this page doesn’t describe, write to [ai@yourdomain.com]. A person reads every message.