Legal
AI policy
Last updated 18 August 2026
Ductective uses AI to generate its answers. This page says how, what it is being built to refuse, and where it can be wrong.
We use AI
Ductective is being built on large language models. The answers you will see are generated, not hand written by a person, and not looked up from a fixed list of questions.
We say this plainly because you should know what you are reading before you act on it.
What the AI is allowed to say
It is being built to answer from documentation it can point to, and to name the document and the page behind each claim.
Where it cannot point to a source, the answer is withheld rather than shown. Where the question falls outside the documentation it has, it is being built to say so rather than fill the gap.
What it will refuse
Gas and combustion, live electrical work, and refrigerant handling are hard refusals by design. It points to standard procedure instead of walking you through the work.
It is not a general purpose chatbot, and it is not a substitute for your training, certification, or judgement.
Where it can be wrong
Generated text can misread a table, carry a claim further than the page supports, or cite a passage that does not apply to the unit in front of you.
That is why the citation is part of the answer. Read the page before you act, and follow your own procedure for anything you work on.
Your data and AI
Waitlist signups are not used to train models. The email, role, and equipment fields exist so we can tell you when early access opens and decide what to build first.
If that changes before launch, we will say so here and in the launch message.
Questions about any of this: [CONTACT EMAIL]