Explains and summarises.Does not assess and does not write policies.Review before use.

That sentence sits under the AI assistant on every screen. This page explains what it means, where the model runs and how to switch it off.

What it will and will not do

Will

  • Explain a requirement in plain words: What does clause 6.1.3 ask for?
  • Summarise a long record or a set of findings
  • Say why an auditor might ask for something
  • Point you to the step or screen where the work happens

Will not

  • Decide whether a control is effective or a risk is acceptable
  • Mark anything as done or compliant
  • Write your policies or procedures
  • Speak for your certification body

Those judgements belong to you, your consultant and your auditor. An assistant that wrote your policies would also make it harder to show the auditor that you understand them.

Where it runs

The model runs on an EU endpoint. Your instance sends the question, and the context you are looking at, only when you ask.

You ask a questionIn the help corner, on the screen you are working on
Your instanceAdds the context from that screen. Hosted in [EU region]
Model endpointScaleway, Paris, France. Returns the answer to your instance
  • Model[model name and version]
  • Used for training[confirm from Scaleway terms; state plainly]
  • Retention at the provider[confirm]
  • Conversation history[confirm: stored in your own instance]
  • Who can switch it onAn administrator in your organisation

Off is a normal setting

Everything else in Hisland Comply works without the assistant. Your administrator can switch it on or off at any time. When a Hisland consultant runs your instance, it starts switched off.

Book a demo

The help corner with the boundary sentence shown at the bottom of the panel. The same panel in dark mode.
The boundary sentence at the foot of the help corner. Reshoot with an AI follow-up answer

Questions about the assistant

Can it read all our records?
It sees the screen context you send with a question, and [describe the index scope precisely once E06 is built].
Why not let it draft our policies to save time?
Policies only work if the people who follow them recognise their own company in them. We would rather help you write a short policy that is true than hand you a long one that is not.
Can we use our own model endpoint?
[decide]